<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by JaviPas on Medium]]></title>
        <description><![CDATA[Stories by JaviPas on Medium]]></description>
        <link>https://medium.com/@javipas?source=rss-c6b6b89f5e8------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*hyT-BHPzepfgS4V1.jpeg</url>
            <title>Stories by JaviPas on Medium</title>
            <link>https://medium.com/@javipas?source=rss-c6b6b89f5e8------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 05:20:59 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@javipas/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Put a HAIS in Your Life]]></title>
            <link>https://medium.com/@javipas/put-a-hais-in-your-life-cdeaed5bb2b6?source=rss-c6b6b89f5e8------2</link>
            <guid isPermaLink="false">https://medium.com/p/cdeaed5bb2b6</guid>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[claude]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[mac-studio]]></category>
            <dc:creator><![CDATA[JaviPas]]></dc:creator>
            <pubDate>Tue, 06 Oct 2026 05:39:44 GMT</pubDate>
            <atom:updated>2026-10-06T05:39:44.734Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1vl6LPgoF9JVNmV9RQwX1Q.jpeg" /></figure><p>In October 2012, I published an article titled “<a href="https://javipas.com/2012/10/04/pon-un-nas-en-tu-vida/">Put a NAS in Your Life</a>.” Back then, I thought those little home servers were wonderful. I still use this kind of device heavily today: as you may know, <a href="https://javipas.com/2025/01/07/asi-me-he-montado-mi-propio-google-fotos/">I built my own NAS</a> with Unraid, among other things, to have a local, private alternative to Google Photos.</p><p>Lately, though, I’ve wanted something that would give me one more option: the ability to run local AI models. The idea is simple: a box in the living room, like the NAS I have now, that can do everything the NAS does — and also run a local AI model, so I can ask questions about my data and devices without sending anything elsewhere.</p><p>That’s where a new term comes in. I think it describes these machines rather well:</p><blockquote><em>HAIS — Home AI Server</em></blockquote><p>The idea is to have a machine at home, in your living room, acting as a server for local models. In practice, it could do much more. It could be your work computer, which you access over remote desktop or macOS screen sharing. And, like a NAS, it could also run download, media, or home automation services such as Home Assistant.</p><p>Tools like Unraid make this easier on machines dedicated to that purpose. But perhaps we won’t need Unraid at all: AI itself could configure the machine and get all those services running smoothly.</p><h3>There are options out there</h3><p>Machines that can do this have been around for a few years. Last year, in fact, I wrote about the machine I would have liked to have at home: <a href="https://javipas.com/2025/06/17/montarme-mi-propio-chatgpt-cuesta-12-000-euros/">a Mac Studio M3 Ultra with 512 GB of unified memory</a> and 819 GB/s of memory bandwidth. At the time, it cost €12,000. That was already a lot, but the memory shortage later made those machines unavailable, while the less expensive configurations quickly went up in price.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7TBUwjWrvFKiTRx9ZER__Q.jpeg" /></figure><p>A few days ago, Apple introduced the Mac Studio M5 Ultra, among other products. The version with the “bad” SoC — a 30-core CPU and 64-core GPU — and 256 GB of memory, with 1.2 TB/s of bandwidth, <a href="https://www.apple.com/es/shop/buy-mac/mac-studio">costs €11,050</a>. The chips are more powerful, but they aren’t why the price has nearly doubled. That’s down to the memory. We probably won’t be seeing the memory prices of a couple of years ago again anytime soon. For now, that train has left the station.</p><p>There are other options, of course:</p><ul><li><strong>Nvidia DGX Spark and its lookalikes:</strong> Dell GB10, ASUS Ascent GX10, Acer Veriton GN100 AI Mini, HP ZGX Nano AI Station, and others. They all use the GB10 Grace Blackwell chip and 128 GB of “unified” LPDDR5X memory. They look good, except for one thing: memory bandwidth is 273 GB/s. Price: €4,800, if you can find one.</li><li><strong>AMD Ryzen AI Max+ 395 / Ryzen AI Max+ Pro 495:</strong> These machines, led by the Framework Desktop, can also act as local AI servers thanks to their 128 GB of “unified” LPDDR5–8000 memory. A newer model even has 192 GB of LPDDR5X-8533. The problem is the same: memory bandwidth. It’s 256 GB/s, or 273 GB/s on the 495. Price: €3,500, if you can find one.</li><li><strong>One or more GPUs:</strong> An RTX 5090 is wonderful, with nearly 1.8 TB/s of bandwidth — unmatched for now — but it has “only” 32 GB of GDDR7 memory. You can use older GPUs such as the RTX 4090, with 24 GB and 1.01 TB/s, or run several cards in parallel to pool their memory. In practice, that means splitting the model across them. Either way, they use a huge amount of power, make a lot of noise, and generate a lot of heat.</li></ul><p>It’s a great solution for some use cases, but not an efficient one, and certainly not an elegant one. The price is <strong>astronomical</strong>. We’re back to the meme that an RTX 5090 <a href="https://videocardz.com/newz/rtx-5090-meme-is-about-to-become-reality-regular-geforce-rtx-5090-cards-now-hit-5099">costs $5,090</a>.</p><p>There are probably other options, such as an RTX 6000 Pro, with 96 GB of GDDR7 memory and a price tag of around €20,000. But that’s roughly where the HAIS market stands today, and it points to a clear conclusion.</p><h3>Macs are excellent HAIS machines right now</h3><p>If local AI models are your HAIS priority, the Mac Studio — or a reasonably powerful Mac mini — looks like the most well-rounded machine available today. And you don’t necessarily have to mortgage your house to get one; there are some interesting alternatives.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-53eZV9ITmJMktT7Lt-zQg.jpeg" /><figcaption><em>The table speaks for itself. </em><a href="https://x.com/0xIlyy/status/2092317832071512545">https://x.com/0xIlyy/status/2092317832071512545</a></figcaption></figure><p>The image in that <a href="https://x.com/0xIlyy/status/2092317832071512545">earlier tweet</a> gives a rough picture of the market — it leaves out AMD, but still. I also came across <a href="https://x.com/yume_arasaki/status/2092767101039898681">a post</a> by @yume_arasaki with a useful analysis of the options and, more importantly, some essential advice: “Buy enough memory to fit the model.”</p><p>In other words, if you want to run Qwen3.8–27B, which uses around 20 GB of video or unified memory, you need a machine with 32 GB of RAM. This user says the operating system alone uses 8–12 GB; I’d say it’s less. The same goes for other models: if you want to run larger ones, that’s fine, but be realistic and invest in the model that’s actually worth it for you.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3dlBVvGaNWWqe_HlUTo9yQ.jpeg" /><figcaption><a href="https://x.com/yume_arasaki/status/2092767101039898681">https://x.com/yume_arasaki/status/2092767101039898681</a></figcaption></figure><p>The <a href="https://pbs.twimg.com/media/HQsAgitbIAAPM_k?format=jpg&amp;name=4096x4096">full-size version</a> of that comparison is genuinely useful if you’re trying to decide which Mac to buy for this kind of work. It would be great to have 256 or even 512 GB of unified memory and never worry about whether a model will fit. But spending more than €10,000 is hard to justify for almost anyone.</p><p>The sensible rule still applies: buy enough memory for the model you actually plan to use.</p><p>What’s interesting is that Apple is winning a game it didn’t even know it was entering. When the company introduced the M1 chips in 2020, we were amazed by their performance per watt. But unified memory turned out to be a bet on the future, even if Apple probably didn’t know it at the time.</p><p>Unified memory lets the CPU and GPU share memory dynamically. Apple has also designed its chips so that memory bandwidth keeps getting more impressive. I made this point in 2025, when I wrote that <a href="https://javipas.com/2025/06/17/montarme-mi-propio-chatgpt-cuesta-12-000-euros/">building my own ChatGPT would cost €12,000</a>: when you want to run a local AI model, two things matter more than anything else:</p><ol><li>A lot of video or unified memory.</li><li>As much memory bandwidth as possible.</li></ol><p>Other factors affect the experience, too. Prefill, for example, wasn’t optimal on the M3 Ultra, but it <a href="https://x.com/wei_wang/status/2093724610441236648">seems set to improve considerably</a> on the M5 Ultra thanks to the new Neural Accelerators <a href="https://www.apple.com/es/newsroom/2026/08/apple-introduces-new-mac-studio-with-m5-max-and-m5-ultra/">built into each core</a>.</p><p>And then there are neat tricks like connecting multiple Macs — even different models — over Thunderbolt to pool their unified memory and build small (or not so small) AI clusters capable of running larger, more powerful models. That comes from <a href="https://exolabs.net/">Exo Labs</a>, which has been working on this for a while and has now <a href="https://x.com/exolabs/status/2092320487019880735">partnered with Apple</a> to bring the option to Macs.</p><h3>What makes a HAIS a HAIS?</h3><p>It’s worth spelling out what I think these machines should be. To start with, a HAIS shouldn’t be your gaming PC, drawing 600 watts with the GPU running at full tilt just to serve AI models. No.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kVfAob16t02Tt3a-mds4lQ.jpeg" /></figure><p>A HAIS should be on 24/7 and be especially efficient and quiet. That’s why Mac Studios and Mac minis are such good options: for now, nothing beats them on that combination.</p><p>It should also have a substantial amount of video or unified memory. But, as that person on Twitter said, don’t get carried away: look for a machine with enough memory for the model to fit. You don’t need a 256 GB Mac Studio M5 Ultra to run models like Qwen3.8–27B, which actually perform very well. A 32 GB unified-memory machine is almost enough, and a Mac mini M5 Pro could be a good fit.</p><p>Expandable storage matters, and that’s the Mac’s weakest point. But it’s not a big problem: a good external SSD solves it, and there are also alternatives <a href="https://www.amazon.es/dp/B0CTWSD8X6?tag=Incognitosis-21&amp;linkCode=osi&amp;th=1&amp;psc=1">such as DAS units</a>. You’ll also want excellent connectivity to your local network; a 2.5 GbE or 10 GbE port gives you plenty of room to work with.</p><p>Beyond that, the rest is largely taken care of. macOS, Linux, and even Windows are all well equipped to run AI apps and services, and to make good use of Docker and other tools. That lets you put the machine to work in all sorts of ways. You’ll need to think about backups, of course, but that goes without saying.</p><h3>What’s missing from this picture?</h3><p>I don’t want to get too deep into the more advanced details of running local models, but inference depends on a lot of things. Here’s a quick overview.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4myC738kDW1IktgQ1Qq3pQ.jpeg" /><figcaption><em>That DGX Spark sure is cute.</em></figcaption></figure><h4>Prefill (“How long until my AI answers?”)</h4><p>This is what happens before the first token appears. Say you ask a model to analyze a 30,000-token document and answer a question about it. The model has to process your prompt and the entire document to generate the KV cache. Prefill involves huge matrix multiplications that can run in parallel and depend much more on raw compute power.</p><p>This is where the M5’s new Neural Accelerators can drastically reduce <a href="https://machinelearning.apple.com/research/exploring-llms-mlx-m5">time to first token</a> (TTFT) — the time between sending a request and seeing the model produce its first token. It’s great if your machine can “write” at 50 tokens per second, but that number is a little misleading if it takes a long time to process the prompt and its input.</p><p>The RTX 5090 is spectacular here, with more than 20,000 CUDA cores and around 3,350 TOPS, compared with the DGX Spark’s 6,144 CUDA cores and roughly 1,000 TOPS. Apple doesn’t publish a directly comparable TOPS figure. In any case, compute power matters especially during prefill; during token generation, memory bandwidth becomes much more important.</p><h4>Decoding (“How fast does my AI write?”)</h4><p>When the model answers, it generates tokens like a stochastic parrot, as you probably know. This is where memory bandwidth takes center stage. It largely determines the tokens-per-second “speed” people talk about when comparing AI models and hardware.</p><p>You can get a rough estimate with a simple calculation: divide memory bandwidth by the actual size of the model weights in memory. If the model uses 20 GB, that gives us:</p><ul><li>DGX Spark: 273 ÷ 20 = 13.7 tokens/s</li><li>M5 Ultra: 1,200 ÷ 20 = 60 tokens/s</li><li>RTX 5090: 1,792 ÷ 20 = 90 tokens/s</li></ul><p>Other factors come into play, but this calculation is useful as a rough guide. More memory bandwidth generally means higher speed. A very large model will still take a while to generate tokens because it has so much information to process every time it produces one.</p><p>There are other topics we could cover — quantization, KV cache, software ecosystems — but I don’t want to make this too heavy. These are the factors I think matter most.</p><h3>The best argument against a HAIS costs €20 a month</h3><p>Let’s be honest: these machines are a major investment. That’s a lot of money upfront for a machine you may or may not use enough to justify the cost. Especially when there’s one simple truth:</p><blockquote>Cloud models are much better at many things right now.</blockquote><p>They’re more capable, more accurate, and much faster. They respond quickly, can handle huge prompts and documents several megabytes in size, and do it without making you notice the effort. The best frontier models in the cloud are still ahead in overall capability, context, multimodality, and ease of use.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*c4c3Kw5-LJggOpWTUJlyBA.jpeg" /></figure><p>You can only get close by spending a fortune on hardware with enormous amounts of unified memory, so you can fit very recent open-weight models once they’ve been quantized.</p><p>Cloud AI also works brilliantly, to be fair. And if you’re not going to use a local model all day for all kinds of tasks, you may be wasting your money. Many people would say you’re better off paying €20 a month for ChatGPT Plus, Claude Pro, or another service. We know these tools work very well — really, really well.</p><p>Inference prices have generally been coming down, and competition among AI companies is fierce. Chinese companies in particular offer huge, capable models at laughably low prices. That makes cloud AI a very attractive option.</p><p>But it has drawbacks. The biggest is privacy. People ask ChatGPT, Claude, and Gemini all sorts of things, but may not realize that their data is handled by OpenAI, Anthropic, and Google, respectively. With a cloud service, your prompts, documents, and responses have to leave your machine to be processed on someone else’s infrastructure.</p><p>Companies use <a href="https://openai.com/security-and-privacy/">encryption</a> and access controls, and let you limit how your data is used. But you still have to trust their policies, systems, and data-retention practices. With a truly local model, that dependence goes away: your data can stay entirely within your own network.</p><p>Local models address that concern because your queries don’t leave your machine. You can ask about health, finances, or anything else — even things you might feel a little guilty asking about — without anyone seeing those conversations. They stay between you and your model.</p><p>The cloud gives you superpowers today, but only by sending your data beyond your machine. A local model requires a painful upfront investment, but it gives you a level of privacy and control the cloud can hardly match.</p><p>Spending €3,000, €5,000, or €10,000 on a machine for local models is hard to justify when €20 a month gets you access to frontier models that are usually better right now. But comparing a HAIS with a ChatGPT subscription tells only part of the story. A HAIS doesn’t have to replace ChatGPT. It can coexist with it; the two aren’t mutually exclusive.</p><p>Perhaps the future won’t involve choosing between local and cloud AI. We’ll probably use both: cloud models for tasks that need the most capable systems in the world, and local models for tasks where privacy, availability, marginal cost, or permanent access to our own data matter more.</p><p>That, I think, is how a new product category starts to take shape.</p><p>Today we have the Mac Studio, DGX Spark, Framework Desktop, and various mini PCs with Ryzen AI Max chips. None was designed exactly for this purpose, but they’re all starting to resemble the same idea: a small, efficient box that stays on all the time, with plenty of memory and enough power to become the digital brain of your home.</p><p>That’s what a HAIS is to me.</p><p>With a HAIS, the possibilities multiply. You could keep a model available to every device in your home — with AI configuring and managing them for you — build a RAG system around years of documents and photos, transcribe audio, sort images, control home automation, run private agents for routine tasks, or use it as a private family assistant. And with all of that, you stay in control. It’s a seriously powerful idea.</p><p>NAS devices did something similar twenty years ago. At first, they seemed like a geek’s indulgence. Then they began storing our photos, movies, backups, camera footage, services, and half our digital lives. I think AI could follow a similar path. And a HAIS wouldn’t have to replace a NAS: it could eventually absorb it.</p><p>In 2012, I wrote “Put a NAS in Your Life.”</p><p>Perhaps it’s time to put a HAIS in yours.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cdeaed5bb2b6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Tesla Software Experience: Almost Brilliant, Occasionally Infuriating]]></title>
            <link>https://medium.com/@javipas/the-tesla-software-experience-almost-brilliant-occasionally-infuriating-5b90568f6491?source=rss-c6b6b89f5e8------2</link>
            <guid isPermaLink="false">https://medium.com/p/5b90568f6491</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[tesla]]></category>
            <category><![CDATA[software]]></category>
            <category><![CDATA[ux]]></category>
            <dc:creator><![CDATA[JaviPas]]></dc:creator>
            <pubDate>Tue, 02 Jun 2026 16:12:41 GMT</pubDate>
            <atom:updated>2026-06-02T16:12:41.478Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vEb_nZ_9_Wx8s1w9pSeWPA.jpeg" /></figure><p>I love my Tesla. Every time I drive it, I feel like I’m floating on a little cloud. Like gliding. That is the feeling I get when driving this marvel. If you ever want to call me a fanboy of something, call me a Tesla fanboy, because I am absolutely fascinated by a car that, in my (admittedly limited and subjective) opinion, leaves the rest of the automotive world behind.</p><p>An example: thanks to my work, I recently had the chance to drive a €65,000 electric car that, theoretically, was better than my Tesla Model Y in every possible way. It was more powerful, sure, but in everything else it felt tacky to me. The interior had details that reminded me of a cheap commercial van and plasticky buttons that I found horrible, and for example the screen and its UI/UX felt quite modest… to put it politely.</p><p>This is something that has been happening to me lately every time I get into other cars, no matter how recent they are: they feel tacky, cheap, almost from another era. I have embraced minimalism, and I love that zen-like calm that surrounds me every time I step into my little Tesla. I used to defend physical buttons and having things to press and turn, but I have become a devotee of that 16-inch QHD (2,560 x 1,440) screen that acts as the absolute control center of the car.</p><p>You don’t need anything else.</p><p>Or do you?</p><p>Well, honestly, a little bit.</p><p>This is a post I’ve wanted to write for a while because, despite all its virtues, Tesla is missing some important things in the software department. Those shortcomings, mind you, are intentional, because Tesla’s software philosophy is what it is — but that doesn’t mean I can’t criticize it, ask for solutions, or suggest some of my own.</p><p>So here we go.</p><p>Before anything else, a disclaimer: this is a personal opinion based on my personal experience with my Tesla Model Y Standard (bought in March’26). It is possible that I am wrong in some areas because there are better ways of doing things, so if you have comments, I’ll be delighted to read them, learn something, and enjoy my beloved Tesla even more.</p><p>Mr. Musk, this one is for you.</p><h3>A Computer on Wheels</h3><p>I’ve heard people say Teslas are basically tablets on wheels, but I would say they are actually computers on wheels. That touchscreen, which admittedly looks like a tablet glued onto the dashboard, is an absolutely essential part of the user experience, but what you don’t see is that the screen is just that: a screen.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pITke-iea7mlExaIjNE00g.jpeg" /></figure><p>What actually does most of the work is a combination of two boards. The first is for the MCU (Media Control Unit, with an AMD SoC in the case of HW4-equipped cars), and the second is the Autopilot Computer, with proprietary AI chips working alongside the external cameras. The photo is bad, but it’s the one I managed to take during <a href="https://www.xataka.com/movilidad/he-probado-modo-fsd-supervisado-tesla-he-descubierto-cosa-supervisar-agotador-que-conducir">my supervised FSD tests</a> in Amsterdam: in reality, that whole “book” is closed, and the two boards are separated by a central cooling plate.</p><p>But of course, you never really notice any of that, because all you see is this sort of “iPad” where everything works wonderfully well.</p><p>There are not too many details about the software foundations of that tablet used to manage infotainment, but we do know Tesla uses a custom Linux kernel — the <a href="https://github.com/teslamotors">TeslaMotors repository on GitHub</a> is pretty outdated, oh well — together with a Qt interface heavily adapted to the needs of the car. Of course, that infotainment layer is one thing, and the real-time operating system (RTOS) that controls the “serious” functions of the car is something entirely separate.</p><p>But as I said, that part is invisible to users, and that is perfectly fine.</p><p>And from there, cool things start happening.</p><h3>The Cars That Never Grew Old</h3><p>First of all, there are the Over-The-Air (OTA) updates, which taught an entire industry that the car you bought not only does not get worse with time, but actually gets better — like good wine.</p><p>The magic of updates means Teslas can fix problems and add features, with varying degrees of success (I’ll come back to this later). Cars used to have software — if you could even call it that — that was basically frozen forever. Tesla taught an entire industry that there was a much more powerful way to do things.</p><p>You may remember that in 2017, during Hurricane Irma, Tesla unlocked around 15 extra kWh of battery capacity (about 50 km / 30 miles of additional range) via software for owners of Model S 60/60D vehicles in Florida, so they could evacuate. It was a positive gesture, but it also made it clear that Tesla was artificially limiting the range of some vehicles.</p><p>The level of control Tesla has over these cars is also absolute, and that means, for example, that if you buy a used Tesla and the previous owner had paid for FSD, you may not get to keep it. <a href="https://observer.com/2020/02/tesla-software-updates-jailbreak/#:~:text=The%20Model%20S%20was%20rigged%20with%20such,version%20and%20removing%20the%20two%20$8%2C000%20options.">Tesla disables</a> it, because if you want it, you have to pay for it again.</p><p>Still, OTA updates are simply wonderful.</p><p>I think we’ve had three of them so far in our Tesla, and while none has been especially dramatic, knowing that your car will work better and gain new little features is fantastic.</p><p>Updates, by the way, have to be downloaded <a href="https://www.tesla.com/ownersmanual/model3/es_es/GUID-A5A60CB3-7659-4B08-B2FD-AFD12C2D6EE1.html">over Wi-Fi</a> (only critical patches can be downloaded via LTE), and the process can take quite a while. We don’t have Wi-Fi coverage in the parking space of our tiny bourgeois mini-resort, so whenever we get an update notification, we use mobile tethering while driving and let the update slowly download in the background.</p><p>The process is slightly annoying because many times <a href="https://www.reddit.com/r/TeslaLounge/comments/1j6lwll/software_updates/">the download speed shown on screen stays at 0 B/s</a>. It feels like nothing is happening, but somewhere I read that during those moments the updates are actually being decompressed in parallel.</p><h3>The GPS That Removes Battery Anxiety</h3><p>Then, of course, there is the GPS navigation system. It is an absolutely central part of the experience because it is either always in the foreground, or sitting quietly in the background waiting for you to bring it back up.</p><p>And here, things get bittersweet, because it is wonderful and absolutely brilliant at some things — and frustrating at others.</p><p>For example, it is extraordinary as an integral part of the charging experience. It doesn’t matter where you are going: the car is always prepared to make sure you don’t get stranded, and it is not only cautious, but also constantly trying to optimize time.</p><p>Let me explain.</p><ul><li>If I tell the Tesla that I want to go to <a href="https://en.wikipedia.org/wiki/Soria">Soria</a>, it suggests a route and tells me that I need to stop for X minutes at Supercharger Y in order to arrive in Soria with Z amount of battery left. You can tell it that you want to arrive with more than Z (say, 50% instead of 10%), but of course that means spending more time at the charger.</li><li>… and Tesla tries very hard to make sure you spend as little time charging as possible. To do that, it prefers you to arrive at the Supercharger with a low battery level, because charging is much faster that way. Charging from 10% to 50%, for example, is much quicker than charging from 50% to 90%. So even if you sometimes feel like the car is cutting it close — <em>“wow, arriving with just 10% feels a bit risky”</em> — Tesla actually knows you’ll get there just fine and then recharge in no time.</li><li>But there is more to it than that. Tesla also takes weather conditions and route characteristics into account. It knows you are crossing mountain passes, driving faster than usual, or dealing with snow, and it constantly recalculates how much battery you will have left both at the Supercharger and at your destination, adjusting the route if necessary.</li></ul><p>All of this makes the famous “range anxiety” almost disappear.</p><p>At least in my case.</p><p>I suppose there are people who have managed to get stranded in their Tesla, but my guess is that it happened because either 1) they got lost, or 2) they stubbornly tried to squeeze every last kilometer out of the battery. Which are, more or less, the same reasons why you can end up stranded in a gasoline car.</p><p>So Tesla has managed to design a system that feels almost magical, because everything is calculated and anticipated so that you barely have to worry about anything.</p><p>The downside is that it focuses primarily on Tesla’s own Supercharger infrastructure. You can enable an option to show third-party chargers, but we have it activated and, honestly, during the trips we’ve taken I don’t think the car has ever shown us one. That is where apps like ABRP (A Better Route Planner) or Electromaps come in, but they are separate apps you need to use on your phone.</p><h3>Where Is My Waze?</h3><p>If I had to say one thing I dislike about Tesla, the first thing that would instantly come to mind is this: I wish I could use Waze in the car. It’s the navigation app I’ve used for years almost automatically whenever I drove anywhere. It didn’t matter if I already knew the route — I used it to avoid (or at least reduce) traffic jams and, of course, to get speed camera alerts.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JjB-v-nJ4j6wpP9tH4RmqQ.jpeg" /><figcaption>A phone mount attached to the Tesla screen is the easy and cheap option if you want Waze in the car.</figcaption></figure><p>I think Waze is absolutely brilliant, but I cannot use it in my Tesla. At least not natively, because Tesla follows an extremely closed and proprietary philosophy. It is impossible not to think of Apple here, because they basically do the same thing with their ‘<em>take it or leave it’ </em>approach.</p><p>Tesla has its own maps and GPS navigation system, and it simply does not want to give users the option to use alternatives natively.</p><p>It is a real shame, but there is not much you can do about it.</p><p>In fact, the current workarounds are all a bit mediocre.</p><ul><li>You can buy a phone mount and place your phone there. That is the option we chose because it is the simplest and cheapest. There are plenty of MagSafe-compatible mounts, or you can buy a case that adds support, but in my case I picked a simpler mount, which lets me place the phone horizontally above the Tesla screen. The downside is that Waze in landscape mode is worse than portrait mode, but for long trips at least it does the job.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Z4ExdPpDlU4Rx6QKb6cApQ.jpeg" /><figcaption>Enhance S3xy Dash</figcaption></figure><ul><li>You can buy an aftermarket dashboard screen like the Enhance S3XY Dash, which is genuinely cool, but <a href="https://www.enhauto.com/pages/dash">costs around €500</a> once you buy it together with the Commander (which is also pretty neat). I know someone who bought one, but I’m not sure whether they use it for Waze. I should ask.</li><li>You can also “hack” the Tesla a little to get Waze onto the main screen. This works by mirroring your phone and projecting Waze inside Tesla’s web browser. I tried it and honestly, I was not impressed. Resolution drops noticeably and there were occasional disconnections. There are several apps for this, and they are easy to find, but I cannot even remember which one I tested.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KX5m8sR3CJiw20XmCAX6Uw.jpeg" /><figcaption>This is Tesla Android miniPC. Interesting, but the project is in an unstable state.</figcaption></figure><ul><li>And finally, there are little mini-PC devices that do basically the same thing, but through hardware instead of software. With them, the Tesla screen becomes a sort of external monitor where you can run Android apps like Waze properly. Again, there are several options out there, and I was tempted to try one called <a href="https://teslaandroid.com/">Tesla Android</a>. You can buy it prebuilt or even assemble it yourself using something like a Raspberry Pi 4. The project <a href="https://www.reddit.com/r/teslaandroid/">seems abandoned</a> and unsupported, though, so be careful.</li></ul><p>Ideally, Tesla would stop making life difficult and simply offer support for both CarPlay and Android Auto. Rumors about this have been floating around for quite a while, but honestly, I don’t believe them.</p><p>And that is a real tragedy because, while I think Tesla’s navigation system is decent, I don’t love it.</p><p>Every now and then it makes some pretty strange routing decisions, and roads I know by heart suddenly become artificially and unnecessarily longer routes — perhaps because the maps are not entirely up to date, or because the algorithms are still not quite polished, at least around here.</p><p>But there is another problem.</p><p>If you want traffic alerts and speed camera warnings, you need to pay for Tesla’s Premium Connectivity subscription. I already <a href="https://javipas.com/2026/04/27/programando-una-app-para-evitar-pagar-la-conectividad-premium-de-tesla/">talked about this elsewhere</a>, and while it comes with extra benefits, it costs $10 a month, and I simply do not drive enough to justify paying that much — so I miss out on features that I could get back instantly if Tesla supported Waze.</p><p>Which, of course, is unlikely to happen.</p><p>Because if it did, a lot of people would stop paying those $10 a month.</p><p>This is a company, dear readers, and companies exist to make money. So I doubt we will ever see native support for CarPlay or Android Auto. My only hope is that, just as Tesla eventually added apps like YouTube Music and Spotify, maybe one day they will also add support for Google Maps or Waze.</p><p>But for now, all we can do is wait. What a pity, Mr. Musk.</p><h3>Apps: They Do Exist</h3><p>Although GPS navigation dominates the screen experience, Tesla also offers a selection of apps that you can use directly from the touchscreen. It is a curious catalog: limited, very limited — and also highly debatable — but it exists.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Cu-0_wiO_pk1moTMK0-Tzw.jpeg" /><figcaption>On the lower left part of the screen, play controls…</figcaption></figure><p>For example, I like that there is native support for Spotify and Apple Music, and especially for YouTube Music, which is what I use because I also happen to subscribe (<em>me? </em><a href="https://subscriptocracy.com/book/"><em>paying for subscriptions?</em></a><em> who would have thought?</em>) to the Premium version.</p><p>I also love small details, like the mini music player shown below the car visualization. Weeks after getting the Tesla, I discovered that you can actually swipe that mini player away by dragging it to the left side of the screen. When you do that, real-time consumption information for the trip appears.</p><p>It’s cool.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ycHqCxcQt4aKll2l8lojDw.jpeg" /><figcaption>… and if you swipe back in the opposite direction, the playback controls return.</figcaption></figure><p>As I said, there are some more apps, although quite a few of them feel much more suited to a phone or tablet than to an in-car infotainment system. I’m talking about Phone, Calendar, Messages (SMS), Zoom (<em>Zoom? Really? Why not Teams, Meet, or even FaceTime?</em>), or Bluetooth. The web browser is more interesting, and although I have barely used it, it actually works very well.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Lt7Gp19ROViV3aK26B96Lw.jpeg" /><figcaption>Source: <a href="https://www.youtube.com/watch?v=hHDNPxoX1Xk">The Interface Cars</a></figcaption></figure><p>And then, of course, there are the entertainment-focused apps. Tesla offers a “Cinema” mode so you can enjoy Netflix, YouTube, Disney+, or Twitch, although naturally you can only use it while parked. It can be useful during charging stops, for example, but if I’m honest, I have used it exactly zero times so far. I’m not saying it is bad to have it, mind you — only that during those stops everyone usually ends up checking their phone for a while, and it probably isn’t the ideal moment to watch a movie or an episode of a show.</p><p>The Arcade app is for me a little more interesting option here, because it offers a small collection of mini-games that you can also enjoy while parked. Our use of it has also been almost nonexistent, although my wife, for example, has played Sudoku a few times, which apparently is surprisingly hard even at relatively easy levels (and she is basically Sudoku-woman).</p><p>The kids have briefly tried some of the games too, but again, for me this whole section feels fairly irrelevant. Tesla briefly supported Steam on the Model S/X, which was striking at the time, but that stopped being supported a while ago. Even more intriguing is the possibility of <a href="https://www.reddit.com/r/TeslaLounge/comments/1sp8s55/tesla_gaming/">playing cloud games</a> from Tesla’s browser by connecting a controller via Bluetooth (or by cable through the USB port in the glovebox). I am not subscribed to Xbox Game Pass Ultimate, so I haven’t been able to try it, but honestly I doubt I would get much use out of it.</p><p>Still, options are always nice.</p><p>There are also some more curious apps, like Caraoke (which we have never used), in addition to radio, which in newer models only works through an internet connection. My wifeuses it from time to time, but I am not much of a radio person, I’m afraid.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*czaf3n1zzt23OjPv7HEteg.jpeg" /></figure><p>What I do find rather silly is the whole Toybox thing. Sure, it does no harm, but I’m genuinely surprised that Mr. Musk went a bit overboard with the quirky ideas here. Especially things like the fart sounds (or “Emissions,” as Tesla amusingly calls them), which are admittedly funny to show off to your kids’ friends, but not much more than that.</p><p>There are other gimmicks too: taking silly photos of passengers (not very good ones), a drawing app worse than Paint, Romance Mode (a fireplace animation on the screen), or Mars Mode, which turns the navigation map into the surface of Mars. They are mildly entertaining for the occasional laugh, but that is about it.</p><p>I wasn’t planning to talk much about Tesla’s mobile app for Android — the one I use — because technically it is separate from the car itself, but I do think there are both really nice ideas and missed opportunities here.</p><p>I love how well it works as the car key.</p><p>The moment you approach, it recognizes you, unlocks the car, and configures everything according to the driver profile.</p><p>I also love being able to remotely control things like climate settings or locking and unlocking the car, and I especially like that it always tells you where the car is parked (although in underground parking garages that may not help much).</p><p>Still, I think it could provide much more information about charging sessions, energy consumption, and charging costs, for example. I also think the data from Tesla’s Energy app should be available directly in the mobile app.</p><h3>Long Live the Dashcam</h3><p>Among Tesla’s software features there are also a couple of tools entirely dedicated to making use of the car’s cameras — especially the exterior ones. The Camera app simply lets you view the same cameras that activate while reversing, so outside of that it is not particularly useful.</p><p>The much more interesting one is Dashcam, which actually serves several purposes. The first is automatically recording video clips during unusual situations — hard braking, accidents, sudden steering corrections — although you can also manually trigger recordings, for example by honking the horn.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XY5xjRw8Smsdrn5wZTR-cw.jpeg" /></figure><p>By default, Dashcam <a href="https://www.tesla.com/ownersmanual/modely/es_es/GUID-3BCC07CE-5EA2-4F40-99D1-27690898FF3C.html#:~:text=Active%20la%20%3Cem%3EDashcam%3C%2Fem%3E%20tocando%20Controles%20%3E%20Seguridad%20%3E%20%3Cem%3EDashcam%3C%2Fem%3E%20.">records videos</a> to the USB drive connected to the glovebox port (Tesla includes one) and continuously overwrites older footage with new recordings. It is a brilliant idea, but of course, if something important happens, you need to remember to remove the USB drive before the clip you care about gets overwritten.</p><p>The second feature is the famous Sentry Mode, which allows the car to monitor its surroundings. People often use it when parking in places where they want to know if someone is trying to tamper with the car — theft, vandalism, things like that. But the car can also activate lights and display warning messages if it detects people lingering suspiciously close.</p><p>It works autonomously and for free if all you want is video recording, but if you want to view the cameras live from your phone, you need Premium Connectivity. We played around with it during the first few days and I think it is genuinely well done, although it does consume battery (somewhere between 2% and 5%, <a href="https://www.reddit.com/r/TeslaModelY/comments/1kr6kxx/power_consumption_of_sentry_mode/">according to Reddit users</a>).</p><p>Maybe I will not use those features much, but I think they are genuinely fantastic. And I also think it is fantastic how well resolved and integrated they feel within the overall user experience. Here, the feeling is very similar to using an iPad: usability feels almost perfect, and execution does too. It is one of many examples showing that experience matters. Tesla has been doing this for years, and it has clearly learned, iterated, and refined all these little details.</p><h3>Monitoring the Tesla</h3><p>And if there is one app that really shows attention to detail, it is the one called “Energy,” which lets you monitor the energy consumption of every trip you make with your Tesla. It is really, really good.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*rIySrLraLqGUEuEHeZbM1g.jpeg" /><figcaption>Source: <a href="https://www.youtube.com/watch?v=hHDNPxoX1Xk">The Interface Cars</a></figcaption></figure><p>You can monitor both driving sessions and energy usage while the Tesla is parked, and the options here are fantastic — even more so the visualizations. The app provides a huge amount of information. For example, it lets you see where during a trip you consumed more or less energy than projected, and why (wind, rain, speed), and among other things you can configure routes for comparison.</p><p>There are also some curious interface choices. For example, the car shows remaining battery either as a percentage — like a phone — or as estimated range in kilometers. Tap once and it switches from percentage to kilometers. Tap again and it switches back. Why not display both at the same time? There is plenty of space on the screen, but Tesla seems convinced that showing only one of the two is somehow the better choice.</p><p>And here is something I genuinely do not understand: The data seems to stay trapped inside the Tesla. You cannot properly consult it from Tesla’s mobile app, even though that feels like the obvious thing users would want. The mobile app is quite good for some things — turning on climate control so the car is nice and cool before you arrive, for example — but it surprises me that it does not expose all that information you would naturally expect to be available from your phone. Data nerds are probably very frustrated by this.</p><p>But fortunately, there are solutions. Or rather, several solutions. I know of three, although there are plenty more. The most popular is probably <a href="https://www.tessie.com/">Tessie</a>, which is paid but seems to be the most mature and complete option for people who do not want to complicate their lives. <a href="https://about.teslafi.com/">TeslaFi</a> is also paid, although it looks somewhat rougher around the edges. And then there is the absolute geek paradise if you enjoy tinkering as much as I do: <a href="https://github.com/teslamate-org/teslamate">TeslaMate</a>.</p><p>TeslaMate is an open-source project that takes advantage of Tesla’s API — available to all owners — to expose all the data hidden inside your car. And believe me, the amount of detail is astonishing. Especially when it comes to energy consumption and trip information.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CRt-MgS31W3phoQk-N8WAQ.jpeg" /></figure><p>Getting it running takes a bit of effort, but it is not terribly complicated. After that, the amount of information TeslaMate gives you is simply overwhelming. At first, I mainly wanted to track charging costs and energy usage, but the level of detail for every trip is remarkable. Here is a partial example from one of our drives to Soria.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-cVdnwn6F9t2baXtwu84tg.jpeg" /></figure><p>Personally, I think the project is spectacular. But here something happened that I already knew would happen: Eventually, the data bored me. There is simply too much of it, and I am not <em>that</em> curious. I still keep it running in the background, collecting data and showing it whenever I want, but the initial fascination eventually collided with reality: I do not really need to know the slope of kilometer 123 on the road to Soria, or whether my consumption at that exact moment was X or Y.</p><p>It is fun as a geeky hobby, but that is about it.</p><h3>The Risks of Putting Everything on a Screen</h3><p>There were a few more notes I had written down for this post, and in particular one important topic: how potentially dangerous it is to drive a car where so many controls live exclusively on a touchscreen. Especially when some of the most critical functions are only accessible through that screen.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FVK5sThD4GBI%3Fstart%3D52%26feature%3Doembed%26start%3D52&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DVK5sThD4GBI&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FVK5sThD4GBI%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/5293fe8fbb362efd43e1719d88c936f8/href">https://medium.com/media/5293fe8fbb362efd43e1719d88c936f8/href</a></iframe><p>I admit I used to be a bit of a hater of the automotive industry’s obsession with touchscreens. And Tesla, as far as I know, was one of the companies that really pushed that trend forward — so this was something that worried me before buying the car. Do you know how long that concern lasted?</p><p>Five minutes.</p><p>One of the things that initially felt strange was changing gear from the screen. The car is automatic, of course, but there is no traditional gear lever or physical selector anywhere in sight. At first, I even kept instinctively reaching with my right hand to where a gear stick would normally be. The system works like this: you press the brake, and then on the left side of the screen a small graphic appears. From there, you can either swipe the car icon upwards to go forward, or downwards to select reverse.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oAbyN28AQs_famCNjVRhzw.png" /></figure><p>It is surprisingly intuitive, and my doubts quickly disappeared, because in reality the strange part is not how you change gears — it is that you almost never have to think about it. Most of the time it only matters when parking, and even then there is a software option that tries to detect your intention when maneuvering, and it works surprisingly well. Not perfect, though. In my garage space there is not much light, and I think that affects it, because it does not always interpret the driver’s intention correctly.</p><p>That detail reassured me quite a bit, but there is another element that is slightly more delicate: climate control. That control lives in the lower-left part of the screen, and turning it on or off, or changing the temperature, is not exactly trivial.</p><p>The temperature controls are two small arrows on either side of the numeric value. You can tap them repeatedly or slide your finger left or right over the number to adjust it. But to actually hit the right spot and control it properly, you need to look at the screen. And at least for me, I do not feel comfortable interacting with it too much while driving because I often miss and have to look again.</p><p>The solution, however, is simple: <a href="https://www.youtube.com/watch?v=YWNCjR4DEi0">Tesla’s voice control</a> works very well for this. A quick “turn climate control on/off” or “set temperature to 23 degrees” is enough for the car to understand and execute the command. And if it does not get it right, you just rephrase it and try again.</p><p>Something similar happens with the volume controls on the lower right side. That is one of the few things I have changed manually only twice. If I want to adjust volume, I usually use the scroll wheel on the left side of the steering wheel, which feels much safer because I do not need to look at the screen. The position of the touchscreen control simply is not ideal — it forces you to look away more than you should.</p><h3>Grok Is Terrible</h3><p>Speaking of voice control, this is a good moment to talk about what is probably the biggest disappointment in Tesla’s software for me: Grok. It is interesting, because although my experience with Grok has been terrible, some English-language reviews seem to partially agree with me, even if they also find some interesting aspects that I personally have not seen yet. Maybe your experience is different, but this is mine.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*w-YN8N4STjF1DIN3mCjIFg.jpeg" /></figure><p>I do not fully understand the approach behind xAI’s chatbot. It feels like it does not really want to compete with the best systems, and instead just keeps releasing new versions that are mainly used inside X (formerly Twitter). There does not seem to be a clear enterprise focus, which is where the money is right now, and although the model itself is probably not bad, it feels oddly irrelevant.</p><p>I have used it very, very little, but like many others, I was curious when it was included in Tesla’s software. The potential here is huge, because while the built-in voice assistant already works reasonably well in a basic Alexa-like way, this could go much further.</p><p>The problem is that it simply does not. Not only that — it is actually a mess. Maybe this is because I am using it in Spanish, but almost every interaction ends up in completely incoherent conversations where I ask something and Grok responds with something else entirely.</p><p>During a trip to <a href="https://en.wikipedia.org/wiki/Biarritz">Biarritz</a>, we wanted to stop near Hondarribia on the way back, and when we asked Grok for recommendations, it did not even recognize <a href="https://en.wikipedia.org/wiki/Hondarribia">Hondarribia</a> as a real place. It kept changing the name repeatedly. When we finally managed to get it to understand what we meant, it still could not recommend anything useful, and we ended up laughing at how absurd the whole thing was. It has happened a couple more times since then, and eventually we stopped using it altogether because it simply did not make sense.</p><p>There are two takeaways here.</p><p>First, I genuinely cannot understand how Tesla and xAI allowed Grok to reach this stage in a production car. Second, this is probably the worst version of Grok we will ever see. It can only improve from here, and it will improve, so I assume that in a year or two this will all be solved.</p><p>At that point, the in-car assistant could become exactly what we expect from a proper AI voice system: something that can do almost anything. Because once voice becomes a primary interface for controlling a car, it will matter even more when these systems are properly polished. At that point, the lack of physical buttons will feel much less relevant, because voice will take over much of the interaction inside the car.</p><h3>Autopilot and the Future</h3><p>In our old 2010 Ford Kuga, I was a huge fan of cruise control. Being able to automate that part of long highway drives felt like magic. So you can imagine how I feel about Tesla’s Autopilot.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6pTgxNBiAC_FviaeWXVWqg.jpeg" /></figure><p>Even though I only have the basic driver assistance features, I find them genuinely impressive. Lane detection can be annoying for some drivers, but I actually find it helpful. I even like it when the car “complains” if I change lanes without signaling, gently resisting and vibrating the steering wheel as a kind of reminder to behave properly on the road.</p><p>From there, there are two main features.</p><p>The first is traffic-aware cruise control, which not only maintains speed but also adjusts it according to traffic conditions, keeping a safe distance and accelerating or braking as needed.</p><p>The second is autosteer, which in addition to the above keeps the car centered in its lane. This feature still amazes me, and I have used it quite a bit on longer trips, but I have to say it is not completely polished. We have already experienced a few well-known “phantom braking” events, where the car suddenly slows down sharply because it thinks something ahead is suspicious. That is quite unsettling, because it is not something you are actively controlling.</p><p>On top of that, if you want to overtake another car, you need to take over manually by using the indicator. It is a bit clunky in those situations because you are constantly switching the system on and off. If you did not have to do that, you would essentially have something very close to supervised FSD on highways.</p><p>Autosteer can also get confused by road markings, especially those indicating exits, so as I mentioned <a href="https://www.xataka.com/movilidad/he-probado-modo-fsd-supervisado-tesla-he-descubierto-cosa-supervisar-agotador-que-conducir">when I tested</a> supervised FSD in Amsterdam in April, this requires not just attention — which is mandatory — but a slightly higher level of attention than usual.</p><p>You can relax, but you cannot forget that the machine is not actually driving alone. That said, I do not think human drivers will remain the main actors for very long. I would say that in 10 or at most 20 years, all cars will have at least a non-supervised autonomous driving mode. And frankly, it will probably be safer than human driving, which is still full of errors, distractions, and impulsive decisions.</p><p>It is only a matter of time.</p><p>And honestly, it is fascinating to be living through this transition period — this “proto-autonomous driving” era. It feels like something I will be able to tell my grandchildren one day: “I drove cars that almost drove themselves… and sometimes did weird things for no reason.” And they will probably ask me: “Grandpa, how did you ever dare drive those primitive machines?”</p><h3>Tesla Needs an App Store</h3><p>I am probably missing a few things, but one thing became clear to me from the moment I started using my Tesla: the software is brilliant, but also frustrating. There are questionable decisions when it comes to the catalog of apps and features, but above all there are absences. Missing pieces of software that feel like they should obviously be there — but are not.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*V1QhcmRlDgCb4mkHfTKGEQ.jpeg" /><figcaption>Concept design. Via: <a href="https://electrek.co/2022/02/16/tesla-app-store-concept-is-so-real-you-can-almost-touch-it/">Electrek</a></figcaption></figure><p>For me, the clearest example is Waze, but the same could be said for other apps that users might want to run in their Tesla either while driving or while parked. Two quick examples already mentioned: ABRP and Electromaps, as well as charging network apps like Endesa or Zunder.</p><p>I understand that there are safety requirements for apps that can be used while driving, but I also think that problem is already solvable on Tesla’s side. There is an obvious solution to all these software gaps: Tesla should build its own app store. Its own App Store.</p><p>Apple has done extremely well with that idea, and I do not understand why Tesla has not adapted it to its own ecosystem. It would open up the software ecosystem in a controlled way, and it would also give Tesla an additional revenue stream, which never hurts. There have been <a href="https://www.teslarati.com/tesla-official-api-app-store/">rumors about this</a> for a while, but nothing concrete has ever materialized that makes me think Tesla is seriously planning it.</p><p>My only hope regarding Waze is that Tesla eventually decides to integrate something similar natively. But I doubt it will happen, because it would likely reduce the incentive to pay for Premium Connectivity. Would I pay €10 a month just to have Waze and traffic alerts built in?</p><p>Maybe yes, and that is the point. I would rather have more options than fewer, and right now Tesla is artificially limiting them.</p><h3>Conclusions</h3><p>I am probably forgetting a few things, but this is already one of the longest posts I have ever written. The idea was to share my impressions of Tesla software three and a half months after buying the car, from the perspective of someone who has been writing about technology for more than 25 years, both <a href="https://javipas.com/">personally</a> and <a href="https://www.xataka.com/autor/javier-pastor">professionally</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HMbDDX3-T-5KQrvHEeVvNA.jpeg" /></figure><p>My conclusion is that Tesla software is both extraordinary and flawed. It is not that I am an expert in infotainment systems, but among everything I have tried, Tesla is clearly a step ahead in my opinion. For me, the biggest limitation is still the lack of support for CarPlay and Android Auto, although maybe that will change in the future.</p><p>But despite all of that, I end where I started: With all its strengths and weaknesses, one thing is clear to me.</p><p>I love my Tesla.</p><p><strong>P.S.</strong> If you are thinking about buying a Tesla and <a href="https://ts.la/javier284252">want to use my referral code</a>, great — both you and I will get 1,000 km of Supercharger charging credit. Cheers.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5b90568f6491" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Anthropic is the new Apple]]></title>
            <link>https://medium.com/@javipas/anthropic-is-the-new-apple-f6eed71e66c9?source=rss-c6b6b89f5e8------2</link>
            <guid isPermaLink="false">https://medium.com/p/f6eed71e66c9</guid>
            <dc:creator><![CDATA[JaviPas]]></dc:creator>
            <pubDate>Fri, 06 Feb 2026 13:38:58 GMT</pubDate>
            <atom:updated>2026-02-06T13:39:52.741Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xf9Ti5FvL2qEPhCGEwimug.jpeg" /></figure><p>In 2006, Apple launched the “<a href="https://en.wikipedia.org/wiki/Get_a_Mac">Get a Mac</a>” campaign — better known to most people as the “I’m a Mac” campaign, thanks to the line that opened many of the ads released over the following years.</p><p>Those ads were brilliant because Apple mocked Microsoft in such a friendly, disarming way that even Microsoft fans probably enjoyed them. The campaign leaned hard on the stereotypes of each platform, but did so with such charm that it was hard to feel offended. It’s funny, in fact, how kindly the Mac (Justin Long) treated the PC (John Hodgman): many viewers ended up sympathizing more with the poor PC, while the Mac sometimes came across as just a bit smug.</p><p>Which, of course, perfectly matched the image we all had — and still have — of Macs and PCs. The ads were fantastic, truly, and you can watch <a href="https://www.youtube.com/watch?v=0eEG5LVXdKo">the entire series</a> in one nearly 40-minute session. I’ve chosen this particular one for a very simple reason: Gisele Bündchen appears in it. If anyone wants to criticize me for that, they’re more than welcome to do so. But come on — it’s genuinely funny, and Gisele Bündchen is in it. Enough said.</p><p>There’s much more, of course. You can find additional information on the <a href="https://www.youtube.com/watch?v=0eEG5LVXdKo">Apple Fandom wiki</a>, and plenty of thoughtful analyses elsewhere, such as those from <a href="https://www.youtube.com/watch?v=0eEG5LVXdKo">WorkingPsychology</a> or <a href="https://www.youtube.com/watch?v=0eEG5LVXdKo">The New York Times</a>.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2Fs01yqqbpuI0%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3Ds01yqqbpuI0&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2Fs01yqqbpuI0%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="640" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/98aa4bb2a5060b6572374b99bbe05eb0/href">https://medium.com/media/98aa4bb2a5060b6572374b99bbe05eb0/href</a></iframe><p>Like almost every ad Apple has ever made, these commercials were designed to convey a single, crystal-clear idea.</p><p>Apple was the superior choice.</p><p>Apple was (and is) a better company than Microsoft. Macs were superior. Apple’s software was superior. Even its philosophy of life was superior. I’ve written before about how <a href="https://www.youtube.com/watch?v=0eEG5LVXdKo">Apple has been an absolute master</a> at using advertising as a vehicle for its self-proclaimed superiority. It did so magnificently, and those campaigns are a big reason Apple enjoys the image it has today.</p><p>What’s particularly interesting is that just a few years earlier, Apple had already launched another epic, legendary campaign:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YRXa20f08cnb6ruWv8r_eA.jpeg" /></figure><p>“Think Different.”</p><p>That campaign urged us not to be sheep in the Wintel herd, not to be part of the faceless masses, but to stand out, to be original, to think for ourselves. It was a wonderful message — right up until it stopped being convenient for Apple. Because not long after, Apple achieved the exact opposite of what it had preached: a world where nobody thought or acted differently. Everyone with the same Mac, the same iPhone, obediently wearing Apple’s jersey. Sheep again — just in a different flock.</p><p>Fast-forward a few years, and we arrive here. A genuinely remarkable ad, in my opinion:</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FFDNkDBNR7AM%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DFDNkDBNR7AM&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FFDNkDBNR7AM%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/03f8176eadb8378a60f08f61bf59b97c/href">https://medium.com/media/03f8176eadb8378a60f08f61bf59b97c/href</a></iframe><p>In it, Anthropic plays Apple. The idea is strikingly similar, and the final message of the campaign is blunt and powerful:</p><p><strong>Keep thinking.</strong></p><p>Interesting, isn’t it? From “Think Different” to “Keep thinking.” It’s a counterintuitive and rather beautiful idea. The ad encourages you to keep thinking, not to surrender your mind to AI, not to let machines replace human thought. And yet, of course, what it really wants is for you to stop thinking — so you’ll rely exclusively on their AI.</p><p>Apple wanted us to stop being different so we could all be the same, united under the Apple banner. Anthropic wants us to keep thinking so that, in the end, we stop thinking — but this time with their AI, Claude.</p><p>There are many parallels between Apple and Anthropic. Like Jobs’ company before it, Anthropic has adopted a position of moral superiority. <em>We</em> care about AI alignment — the <a href="https://www.anthropic.com/news/claudes-constitution">“Constitutional AI” document</a> is a perfect example. <em>We</em> believe in <a href="https://www.anthropic.com/news/anthropics-responsible-scaling-policy">responsible scaling</a>. <em>We</em> are <a href="https://www.anthropic.com/transparency/system-trust-reporting">transparent</a>. <em>We</em> release open standards for everyone, <a href="https://www.anthropic.com/transparency/system-trust-reporting">like MCP</a>.</p><p>In short: <em>we are superior.</em></p><p>And because they see themselves as morally superior, they feel free to mock the competition in exactly the same way Apple once mocked Microsoft. They’ve done so recently with an absolutely fabulous advertising campaign — so fabulous, in fact, that I’m embedding all four ads simply because I want to, and because I can.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FDe-_wQpKw0s%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DDe-_wQpKw0s&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FDe-_wQpKw0s%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/8a891c777e8b6fb741cd5dd78a1763bb/href">https://medium.com/media/8a891c777e8b6fb741cd5dd78a1763bb/href</a></iframe><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FFBSam25u8O4%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DFBSam25u8O4&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FFBSam25u8O4%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/7efef111ab743b9d0ad25d5fc876d91d/href">https://medium.com/media/7efef111ab743b9d0ad25d5fc876d91d/href</a></iframe><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2F3sVD3aG_azw%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3D3sVD3aG_azw&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2F3sVD3aG_azw%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/d4476d7f8e87a942a66c6aad59844fb0/href">https://medium.com/media/d4476d7f8e87a942a66c6aad59844fb0/href</a></iframe><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FkQRu7DdTTVA%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DkQRu7DdTTVA&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FkQRu7DdTTVA%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/363d4637558d8bc042d55d5bb1c58692/href">https://medium.com/media/363d4637558d8bc042d55d5bb1c58692/href</a></iframe><p>The first thing that caught my attention was that these were human ads. They’re not AI-generated; they use real actors — or at least I think they do, which these days is no small assumption. The second thing is how incredibly well executed they are, especially the clever idea that the human assistant always pauses briefly before answering, and always responds with what English speakers call <em>sycophancy</em> — which translates quite nicely as flattery.</p><p>AI models flatter you. “What a great question, Javier.” “What an interesting thought, Javier.” They win you over with conversational compliments. And here Anthropic — whose own model also does this, albeit perhaps a bit less — twists the knife to stab its biggest rival, OpenAI.</p><p>They never even mention OpenAI by name. They don’t need to. It’s immediately obvious who they’re talking about. Not just because of the assistant’s tone, but because Anthropic is criticizing exactly what OpenAI has just announced: <a href="https://openai.com/index/our-approach-to-advertising-and-expanding-access/">ads are coming to ChatGPT</a>.</p><p>Frankly, this was inevitable. Many people have no desire — or ability — to pay for AI, so advertising is the obvious way to complement a subscription model. Don’t want to pay? Fine. You can still use ChatGPT — but you’ll see ads. OpenAI has taken a lot of criticism for being the first major player to do this, but I’m convinced they’ll all follow eventually. Anthropic included.</p><figure><img alt="https://x.com/claudeai/status/2019071113741906403" src="https://cdn-images-1.medium.com/max/1024/1*lxN67-s5T5w5zlhRG0KXug.jpeg" /></figure><p>Still, because ads are the “dirty,” user-hated tactic, OpenAI becomes the villain, while Anthropic — wrapped in its moral high ground — gets to play the good guy. The pure one. The cool one. They proved it with the campaign that debuted yesterday, launched with the message: <em>“Ads are coming to AI. But not to Claude. Keep thinking.”</em></p><p>It’s a powerful message. We are superior. We won’t run ads (even though they’d be very convenient), because, once again, we are superior to OpenAI and everyone else.</p><p>Unsurprisingly, the message seems to have irritated Sam Altman. Last night, <a href="https://x.com/sama/status/2019139174339928189">he posted a heated response</a> admitting that the ads made him laugh, but complaining that they were “dishonest.” They are — but that’s how advertising works in the U.S. You can roast a competitor and get away with it, as long as it’s funny and relatively harmless. Defamation simply isn’t part of the equation.</p><p>If I were Sam Altman, I’m not sure I would have responded at all. His post feels like another case of <em>if it stings, it’s because it landed.</em> I probably would’ve gone with “don’t feed the troll” and moved on — but that’s easier said than done when the provocation is this good.</p><p>The reality is that Anthropic is increasingly positioning itself as the prestige brand in AI: <em>we have a great, exclusive model; we don’t need ads.</em> OpenAI, meanwhile, is aiming for the masses — the “plebs,” almost — those who can’t afford to pay and will have to swallow the ads.</p><p>The analogy with Mac vs. PC, or iPhone vs. Android, is obvious. You don’t just buy a Mac or an iPhone because they work well; you buy them because they make you feel superior, because they place you above the masses.</p><p>Anthropic seems to be chasing exactly that same effect. Use our AI and keep thinking — and that will make you superior. You’ll rise above the unthinking crowd.</p><p><em>Chapeau</em>, Anthropic.</p><p><em>Chapeau</em>.</p><p><em>Disclaimer: this is a lightly edited translation of my </em><a href="https://javipas.com/2026/02/05/anthropic-es-la-nueva-apple/"><em>original post</em></a><em> at Incognitosis.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f6eed71e66c9" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[When AI Becomes Essential]]></title>
            <link>https://medium.com/@javipas/when-ai-becomes-essential-cb528d3c8c8f?source=rss-c6b6b89f5e8------2</link>
            <guid isPermaLink="false">https://medium.com/p/cb528d3c8c8f</guid>
            <category><![CDATA[vibe-coding]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[JaviPas]]></dc:creator>
            <pubDate>Fri, 23 Jan 2026 13:57:02 GMT</pubDate>
            <atom:updated>2026-01-23T13:57:02.060Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*II6Nfcbp0OasP9VU7pGCtw.jpeg" /></figure><p><em>This is a translation from </em><a href="https://javipas.com/2026/01/22/cuando-la-ia-se-convierta-en-imprescindible/"><em>the original</em></a><em>, in Spanish, at </em><a href="https://javipas.com/"><em>Incognitosis</em></a><em>.</em></p><p>Xavi is a close friend of mine. An old friend. One of those friendships that never really goes away, even if you don’t see each other all the time. We’re both pretty nerdy, and back in the day, whenever we met up, we’d talk about geeky stuff and play a mini championship of <em>Match Day II</em> on his Commodore 64, or mess around with some other retro game on one of the many old machines he owns.</p><p>Lately we don’t do any of that.</p><p>We only talk about AI and about how much it blows our minds to code things without having any idea how to code.</p><p>Both of us are doing things we never imagined we’d be capable of, and we’re both absolutely fascinated and ecstatic and worried and terrified by what’s coming. For us, and for everyone else. Out of our conversations usually comes an idea for a post on Incognitosis, and when it doesn’t, those ideas pop up later in our WhatsApp chats. This article, in fact, comes directly from something he said to me a few days ago:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QlMPB1UPSOVTcwpeckez3A.jpeg" /><figcaption>(Sorry, english readers, this is the original screenshot)</figcaption></figure><p>I’m sure this has happened to you more than once. You’re using ChatGPT, or Claude, or Gemini, fully immersed. You’re in the zone. Coding, writing, thinking out loud. And suddenly: boom. The chatbot informs you that you’ve reached your usage limit and that you’ll be able to continue in X hours.</p><p>And you just sit there, blocked and irritated.<br> “What do you mean, what?!”</p><p>It completely kills the momentum. Sure, you can jump to another chatbot and sort of continue there, but it’s never quite the same. The context is missing, the flow is gone, and the conversation loses its meaning.</p><iframe src="https://cdn.embedly.com/widgets/media.html?type=text%2Fhtml&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;schema=twitter&amp;url=https%3A//x.com/hackerrank/status/2013499794736763174&amp;image=" width="500" height="281" frameborder="0" scrolling="no"><a href="https://medium.com/media/56f4dc4c6c8fea0b7c6fc1d4ec5a97f3/href">https://medium.com/media/56f4dc4c6c8fea0b7c6fc1d4ec5a97f3/href</a></iframe><p>AI companies know this perfectly well. They enforce these limits very deliberately, with the usual goal in mind: pushing you to subscribe to a paid plan where those limits magically disappear. On top of that, you never know exactly when you’re about to hit the wall. If I did know, for example, I’d start summarizing the conversation into a prompt as I go. That way I could reuse it later, in another session or even in another chatbot, with all the context intact. But no. The warning always comes out of nowhere, without notice. Ambush-style.</p><p>IIf you search for “<a href="https://x.com/search?q=claude%20usage%20limit&amp;src=typed_query&amp;f=top">claude usage limit</a>” on X, you’ll find a long stream of people complaining about exactly this. One of the tweets I embedded even includes a video that’s particularly unsettling: a scene from <em>Breaking Bad</em> where Jesse is, if I remember correctly, going through withdrawal. Or something very close to that. Either way, the analogy is uncomfortable — and disturbingly accurate.</p><p>We’re becoming addicted to AI.</p><p>My feeling is that, little by little, AI is turning into something essential in our lives. Just like computers once did. Like the internet. Like smartphones. Take any of those away and we feel lost, isolated. Asking ourselves, “Okay… now what?”</p><p>For many people, this still isn’t the case. They ignore AI for now. They see it as something that exists, sure, but doesn’t really do anything for them, so they don’t bother using it. Which makes sense: they don’t yet understand what it can do for them today, let alone what it might do tomorrow. I’ve said it before, and I’ll say it again: my gut feeling is that <a href="https://javipas.com/2024/10/02/la-ia-va-a-cambiarlo-todo-o-no-o-si/">AI is going to change absolutely everything</a>. And when that happens, it will become almost unavoidable. We’ll depend on it for nearly everything, just like we depend on electricity or the internet today. Cut either of those off at scale and for a prolonged period, and let’s see who manages to survive the chaos.</p><p>As Xavi pointed out in his message, that’s already what’s happening. We’re quietly building a total dependency on AI. We criticize it, we know it’s flawed, but we use it more and more because, well… it’s amazing. It’s convenient. Look how cool it is that I can generate Studio Ghibli–style images without knowing how to draw. Or build apps without really knowing how to code. Or do darker things too — because yes, those exist and will exist. AI is just a tool, and like any tool, it can be used for good or for harm.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OFE8Iqm9rklJl-cvj-VWTQ.jpeg" /><figcaption>Oopsie. Sorry again.</figcaption></figure><p>Xavi and I kept chatting on WhatsApp, and he summed it up perfectly. The same AI that’s allowing us to do things we simply couldn’t do before is also becoming a threat. Because when something does everything for you — even if you pay for it — you get comfortable. You rely on it. It feels practical, logical, reasonable. Why walk to Soria if you can drive? Why do a multiplication by hand if you have a calculator? Why use this app if I can just ask AI to build one tailored exactly to what I need?</p><p>The problem is that all of this makes you a little more useless. A little lazier. A little dumber. You think less. You outsource more. And suddenly it’s very easy to spend your entire day scrolling through TikTok or Instagram or whatever fresh hell they’ve come up with. Until we slowly turn into what <em>WALL-E</em> portrayed so brilliantly.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2Fs-kdRdzxdZQ%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3Ds-kdRdzxdZQ&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2Fs-kdRdzxdZQ%2Fhqdefault.jpg&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/ddf9894c8759919fa44e5d2fc17c8d34/href">https://medium.com/media/ddf9894c8759919fa44e5d2fc17c8d34/href</a></iframe><p>Google already made us lazier. Worse at remembering things. Why bother, if it remembers everything for us? That’s why I once wrote an article called <a href="https://javipas.com/2017/03/07/quedemonos-sin-saberlo/"><em>Let’s Not Know</em></a>, which started like this:</p><blockquote><em>Some time ago a well-known friend+ recommended a curious movie to me: </em><a href="https://www.imdb.com/title/tt1791682/">While We’re Young</a><em>. Ben Stiller and Naomi Watts meet a young couple, Adam Driver and Amanda Seyfried, who seem pretty laid-back. One night they go out to dinner, a doubt comes up in the conversation, and Ben Stiller makes the gesture we’d all make: he pulls out his phone and says, hold on, I’ll look it up. Then Adam Driver goes all guru-like and, with a serious and thoughtful face, says to him:</em></blockquote><blockquote><em>— No. Let’s not know.</em></blockquote><blockquote><em>The phrase is brilliant, and it leaves Stiller and Watts stunned. Which is how we’d all be if someone stopped us right now from turning to our phone or smartwatch to check that piece of information we don’t remember and don’t even need to remember anymore. We don’t need to, because for that we have these little gadgets that make us a bit lazier every day.</em></blockquote><p>With AI, this effect will be multiplied by a thousand. We’ll become more comfortable than ever. More passive. We’ll forget what little we still know — which, thanks to Google, wasn’t much to begin with. And there we’ll be, happily chatting with our AI assistant on our phone, blissfully free from thinking or remembering anything at all, until one day we ask something like:</p><p>“Wow, I really enjoy talking to you. This is great. Oh — all the lights in my house just went out. How did you fix that again?”</p><p>And the chatbot replies:</p><p>“Claude usage limit reached. Your limit will reset at 7:00 AM.”</p><p>That’s a bad, dear reader.<br>Really bad.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cb528d3c8c8f" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How I Built the Perfect Immich TV App (Without Writing a Single Line of Code)]]></title>
            <link>https://medium.com/@javipas/how-i-built-the-perfect-immich-tv-app-without-writing-a-single-line-of-code-4aac3108b51d?source=rss-c6b6b89f5e8------2</link>
            <guid isPermaLink="false">https://medium.com/p/4aac3108b51d</guid>
            <dc:creator><![CDATA[JaviPas]]></dc:creator>
            <pubDate>Tue, 09 Dec 2025 11:16:26 GMT</pubDate>
            <atom:updated>2026-04-28T09:32:26.878Z</atom:updated>
            <content:encoded><![CDATA[<p>I built the Android TV app I always wanted in a weekend without knowing Kotlin. Here is how AI made it possible.</p><p>Back in January 2025, <a href="https://medium.com/@javipas/thats-how-i-ve-replaced-google-photos-0f5723719a7d">I shared a massive post</a> about how I set up my own Google Photos alternative using an exceptional Open Source project called <a href="https://immich.app/">Immich</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xrUFpnFjGWt56vsQ5hEe6A.jpeg" /><figcaption>Immich works great on your computer and your smartphone. On your TV? Not so much. Source: Immich.app</figcaption></figure><p>I’m still enjoying its virtues, but for the longest time, I suffered from its one major flaw: the lack of a (good) native TV client. Immich looks phenomenal on a computer or smartphone, but if you want to view family photos on the big screen, you hit a wall. My workaround involved a long HDMI cable or casting a Chrome tab from my MacBook Air — functional, but far from ideal.</p><p><strong>The Problem:</strong> I wanted a native Android TV/Google TV experience to navigate my library seamlessly. There was a glimmer of hope in the form of the <a href="https://github.com/giejay/Immich-Android-TV">Immich Android TV project</a>. The developer, <strong>giejay</strong>, is quite active, but the app lacked two features essential for my sanity: a <strong>Timeline</strong> and <strong>Favorites management</strong>.</p><p>Without a timeline, finding a photo from 2015 meant scrolling linearly through 140,000 images (that’s the size of my photo library) starting from today. That’s not browsing; that’s an endurance sport.</p><p>So I forked the official Immich Android TV project and, with AI as a coding assistant, I added a few features I’d been missing. Without AI, this would have been impossible for me. With it, almost anyone can turn an app idea into reality. And if there’s an Open Source project already created, like in this case, that concept is even more powerful. <a href="https://www.youtube.com/watch?v=xPvwewft83Q">Here’s the result in a little video</a>:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*j44cBvGs6RjyS5twbZ0uNg.jpeg" /><figcaption><a href="https://www.youtube.com/watch?v=xPvwewft83Q">https://www.youtube.com/watch?v=xPvwewft83Q</a></figcaption></figure><h3><strong>The Solution: Taking the Bull by the Horns</strong></h3><p>I had some vacation days saved up, so I decided to fork the project and build the version of Immich Android TV I actually needed. The result?</p><ul><li><strong>A Proper Timeline:</strong> I can now jump through years and months instantly. Navigating to a specific month in 2015 takes seconds. Using Immich on TV is finally pure Valhalla.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xcYxYE7WA-T3gGHmi6Xo_g.jpeg" /></figure><ul><li><strong>A Cleaner UI:</strong> I added options to hide filenames (which I hate seeing) and adjust the grid to show 3, 4, or 5 columns (the official app is locked to four). I also rounded the thumbnail corners and improved spacing. It still needs polish, but it looks much better.</li><li><strong>Favorites Sync:</strong> I can now long-press the center button on the remote to toggle a photo as a “Favorite.” This syncs instantly with the server, so changes appear on my phone and web versions immediately. Valhalla Part 2!</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ivzl0NRTlXmorBMpvmyOWg.jpeg" /></figure><ul><li><strong>A New Settings interface</strong>: The official settings interface seems a little off and ugly imho, so I tried to redesign and apply that design to the app. The result isn’t amazing, but it took me just 30 minutes and works well.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*58t18aomEnE-IqFXwxiXpg.jpeg" /><figcaption><a href="https://cdn.knightlab.com/libs/juxtapose/latest/embed/index.html?uid=dc59629c-d4db-11f0-ba1b-0e6f42328d7d">https://cdn.knightlab.com/libs/juxtapose/latest/embed/index.html?uid=dc59629c-d4db-11f0-ba1b-0e6f42328d7d</a></figcaption></figure><ul><li><strong>Visual Tweaks:</strong> Small details, like adding a “Play” icon to video thumbnails. It’s an easy way to identify and differentiate that content as a video.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*usS7spEqYn2HYIa-AKV2Hw.jpeg" /></figure><h3><strong>The Twist: I Didn’t Write the Code</strong></h3><p>The difference isn’t just aesthetic — it’s functional. But here is the kicker: <strong>I didn’t write these changes. Artificial Intelligence did. </strong>Although I’ve experimented before (like with my little card game counter — <a href="https://javipas.com/2025/08/04/chatgpt-puede-ser-maravilloso/">post in Spanish</a>), this was the big leagues. Here is exactly how I did it:</p><ul><li><strong>The Setup:</strong> I combined several tools and chatbots and I changed from one to another on different phases. I started by installing Droid CLI and configuring an OpenAI-compatible token that gave me access to gpt-5.1. I also installed Android Studio to be able to run an emulator (a “1080p TV”) to see how the changes were working after each compilation/build.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-xZAalSKQoyPru9uobTZGw.jpeg" /></figure><ul><li><strong>The Brains</strong>: Later in the project I switched to Gemini 3 (mostly the Thinking, but Fast also worked). From time to time I used Claude Sonnet 4.5 (the free version at claude.ai). I even tried Kimi K2 and Deepseek v3.2, but I wasn’t a fan of their speed or responses, so I went back to Gemini 3. I’d say 80% of the code was from this AI model. I have a Gemini Ultra account, so I never hit the typical ‘You have to wait X hours…’ limit.</li><li><strong>The Process:</strong> At the very beginning I asked the AI how to fork the repository and clone it to my Mac mini M4. Once Droid CLI could “see” the code, I just started chatting. With Gemini 3 the process was the same: on one window I had a macOS terminal with my working directory, on the other a browser with Gemini 3 running. I’d describe what I wanted, Gemini would write the code, and I’d paste it into project files using nano and then compile. If there were errors, I just copy-pasted the output to Gemini. If something worked but not the way I wanted, I’d describe the issue to Gemini in detail, sometimes with screenshots.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zGMHtVTT2E0WVWJCmdsl5w.jpeg" /><figcaption>Work in progress!</figcaption></figure><ul><li><strong>The SuperTool</strong>: Claude Code has integrated agents and tools that ask permission before editing or running files, so you can mostly just watch, but I just had $10 in Anthropic API tokens and although it’s amazing to watch in action, the costs add up fast and I went back to the Gemini 3 model (copy-paste).</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xFd4tOPvNJ4KYor109T5wg.jpeg" /></figure><ul><li><strong>“Vibe Coding”:</strong> The surprising part here is that I simply asked for things in plain text. When the AI wrote the code, I tested it. If it broke, I pasted the error back to the AI, and it fixed it. Rinse and repeat.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YFYm0s75zlzZmbElrS3AtQ.jpeg" /></figure><ul><li><strong>The Experience:</strong> I have a CS degree, but I barely ever write real code as I work as a tech journalist. I don’t know Kotlin, Gradle, or XML. I was completely out of my depth, but I trusted the AI. And wow, does it know its stuff.</li></ul><p>The first day, I got the Timeline working in a single morning. It was ugly, but functionally perfect. For the design, I used <a href="https://stitch.withgoogle.com/"><strong>Stitch</strong></a> (via Google Labs) and <strong>Nano Banana Pro</strong>. I uploaded a screenshot and prompted: <em>“Redesign this to be clean, elegant, dark mode, and avoid blocky rectangles.”</em> The result was surprisingly slick… and not really doable.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*AOxVr0C2yghsJasxvgODpQ.jpeg" /></figure><p>I tried for a few hours, but every AI model got stuck with a navigation problem I couldn’t resolve. I went back to Stitch and asked for another iteration with a sidebar navigation for years/months. I uploaded that image to Gemini and I had the new design for the timeline (the one I shared before) working in a little over an hour.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ehi1ZlQHXGChJ7uDma1FSg.jpeg" /><figcaption><a href="https://cdn.knightlab.com/libs/juxtapose/latest/embed/index.html?uid=f09becb6-d4e0-11f0-ba1b-0e6f42328d7d">https://cdn.knightlab.com/libs/juxtapose/latest/embed/index.html?uid=f09becb6-d4e0-11f0-ba1b-0e6f42328d7d</a></figcaption></figure><p>The hardest part was the Favorites management. My custom client had to communicate with the Immich server to sync changes. As I said, gpt-5.1 actually struggled here, so I switched to Gemini 3 (and occasionally the free web version of Claude 4.5 Sonnet) to cross the finish line.</p><p><strong>Why This Matters:</strong> I found myself using Android Studio and the emulator to watch my app evolve in real-time. I didn’t even need to move APKs around; a simple command (./gradlew installDebug) compiled and pushed the app to the device instantly.</p><p>I will probably forget how to do this by tomorrow. But that doesn’t matter, because I can just ask Gemini or gpt-5.1, and they will remind me. You don’t just stop thinking; you stop needing to remember.</p><p><strong><em>It’s scary, but incredible</em></strong>.</p><h3><strong>Conclusion</strong></h3><p>I read about “vibe coding” all the time, but doing it is a different beast. If an app doesn’t do what you want, you can now program it yourself with AI. If you have the patience to prompt and test, you can build the unimaginable in a weekend.</p><p>I’ve uploaded the code <a href="https://github.com/picajoso/Immich-Android-TV">to my GitHub repository</a> — a place I usually only visit to lurk, but now actually contribute to.</p><p>Full credit goes to the original developer, giejay, and the AI. I plan to share my changes with him (GitHub doesn’t really work as a social network for developers, contacting them is usually difficult); maybe they’ll make it into the official release. But one thing is clearer to me now than ever: <strong>AI is going to change everything.</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4aac3108b51d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Printer-and-Cartridges Model: HP’s Gillette Moment]]></title>
            <link>https://medium.com/@javipas/the-printer-and-cartridges-model-hps-gillette-moment-5da724f24c7d?source=rss-c6b6b89f5e8------2</link>
            <guid isPermaLink="false">https://medium.com/p/5da724f24c7d</guid>
            <category><![CDATA[economy]]></category>
            <category><![CDATA[business-models]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[hp]]></category>
            <category><![CDATA[subscription]]></category>
            <dc:creator><![CDATA[JaviPas]]></dc:creator>
            <pubDate>Wed, 26 Nov 2025 16:17:15 GMT</pubDate>
            <atom:updated>2025-11-26T16:17:15.126Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ElyQTRZAzHfUYoErx0pGaw.jpeg" /></figure><p><em>You’re about to read how HP discovered that the Razor-Blades model could be really a thing for them and changed our relationship with printers forever. This is one chapter from “</em><a href="https://subscriptocracy.com/book"><em>Subscriptocracy</em></a><em>” — my book exploring the subscription revolution. Read 60% here, then </em><a href="https://subscriptocracy.com/#/portal/"><em>subscribe for the full chapter</em></a><em> + 2 bonus chapters + insider insights on the subscription economy.</em></p><p><em>Also: the book will be </em><strong><em>30% OFF</em></strong><em> on Black Friday! Stay tuned!</em></p><h3>Let’s Talk About Razors</h3><p>He literally had the name of a king.</p><p>In 1881, the star of our story was 36 years old. Something happened then that changed his life. He landed a job as a traveling salesman for the Baltimore Seal Company. Even more signiﬁcant than the job itself was the friendship he formed with the founder of that company, William Painter.</p><p>Painter was a restless inventor. Over time, he registered 85 patents of the most unique kind. One was for an ejector seat for train passengers; another, a machine to detect counterfeit bills. While these inventions were certainly noteworthy, Painter hired our hero to sell a different invention — something seemingly far simpler yet ingenious. It was something timeless, still thriving today: the crown cork. The same cap we still use today on beer bottles, among other things. This little invention made Painter the owner of an empire that survives to this day and is part of the prestigious Fortune 500 list.</p><p>Neither Painter nor our star could have foreseen how far this invention would go. But what they did know was that they liked each other. Over time, their friendship grew stronger, and our hero would often join Mr. Painter at his home for various gatherings. They were alike in one key way: both wanted to solve problems and make money doing it. During one of those evenings, Painter turned to his protégé, someone he’d already recognized as having entrepreneurial ambitions, and said:</p><p>“You’re always thinking about inventing something. Why don’t you try to create something like the crown cap — something that, once used, is thrown away? With every new customer you sell another cap to, you build a constant stream of income.”</p><p>The young man took that advice to heart. Inventing something disposable made sense.</p><p>Let’s fast-forward to 1895, over a century before Amazon would captivate the world. Today, Amazon amazes us with its massive catalog, but what seems revolutionary now wasn’t new at all.</p><p>It certainly wasn’t new to Aaron Montgomery Ward, who, years earlier, had created a retail giant pioneering the mail-order catalog business. Ward realized that people living in rural areas wanted the same products sold in big cities, but they could only get them through intermediaries who limited access and drove up prices.</p><p>What did he do? He cut out the middlemen. He launched a mailorder business offering an endless variety of products that rural customers could pick up at the nearest train station. With $1,600 and two partners, he published his ﬁrst catalog in 1872. The project was a total success.</p><p>The catalog grew — from 32 pages in 1874 to a thousand pages by 1896 — along with a captivated clientele who dubbed it the “wish book.” Businesses of all kinds paid to advertise and include their products in the catalog. By 1904, Ward was mailing out three million catalogs and entering into a ﬁerce rivalry with another retail giant of the time. You might recognize the name: Sears.</p><p>But back in 1895, the Montgomery Ward &amp; Co. catalog was an institution. Among its 624 pages — featuring over 25,000 products — there was a section dedicated to razors. All the models were nearly identical, with little innovation in a product segment that discouraged shaving at home. Keeping a razor in good condition was difﬁcult. Razors needed to be honed (sharpened) using a leather strop and a whetstone. This made the process cumbersome, so most people opted to visit a barber for a shave.</p><p>There were exceptions, though. The Star Safety Razor was a worthy predecessor to modern razors. Patented in April 1887 — though sold earlier — , it introduced disposable blades. The razor itself cost $1.50, and each blade cost $1. Unfortunately, this concept didn’t eliminate the need to hone and sharpen blades, but it was a step closer to bringing at-home shaving to the masses.</p><p>This was still a problem.</p><p>And a problem is often an opportunity in disguise.</p><p>This is where our hero enters the scene. “One particular morning,” he recalled, “as I started to shave, I found my razor dull, and it was not only dull, it was beyond the point of stropping and it needed honing, for which it must be taken to a barber or cutler.” It was at that moment, with the razor in his hands and his eyes ﬁxed on it, that his invention was born.</p><p>That same morning, he left a note for his wife, Alanta. It read:</p><p>“I’ve got it. Our fortune is made.”</p><p>That young man, as I mentioned earlier, had the name of a king.</p><p>His name was King.</p><p>King Camp Gillette.</p><p>Gillette would go on to ﬁle not one, but two patents granted on November 15, 1904, detailing the functionality of his disposable-blade razor. His concept solved the problems of products like the Star Safety Razor, which was halfway between a razor and a modern blade. While the Star’s blades were reusable, sharpening them was difﬁcult for non-experts.</p><p>Gillette’s design was revolutionary: the razor was much lighter and featured a mechanism to hold a thin blade rigidly in place. Production costs were signiﬁcantly lower, moving from costly forged steel razors to ultrathin steel blades. What Gillette achieved wasn’t just revolutionizing the world of shaving — it was creating an entirely new business model.</p><p>This model, known as the razor-and-blades model, is based on selling a product at a loss (or break-even) that requires a consumable component to generate proﬁt. The goal is to create a steady, reliable income stream by locking the consumer into a proprietary platform or tool for an extended period.</p><p>Though this model has become economically famous, not everyone agrees it’s valid. Randal C. Picker, a law professor at the University of Chicago, published an extensive essay in 2010 on The Razor-and-Blades Myth(s) explaining the history of King Gillette’s invention.</p><p>According to Picker, the model is theoretically ﬂawed. If you make money on the blades, why not let someone else manufacture the razor, so you don’t incur losses? In fact, the base product — the razor — should be priced higher, not the consumables. “The razor-and-blade strategy doesn’t make much sense unless you can effectively lock in the relationship between the razor (the platform) and the blades (the aftermarket),” he argued.</p><p>Gillette managed to do this because his patents gave him that privilege. But Picker’s analysis suggests otherwise. The razor cost $5 — a small fortune at the time, roughly a third of a worker’s weekly wage — which doesn’t align with the original philosophy of this economic model.The razors should have been much cheaper under this model. It seems Mr. Gillette wanted to proﬁt from both the blades and the razors.</p><p>The patents eventually expired, but surprisingly, this didn’t jeopardize his idea. In fact, it did the opposite. Competition forced Gillette to change his pricing philosophy: at ﬁrst, he preferred selling expensive razors with fewer blades. Only when his patents expired and competitors emerged did he begin selling cheaper razors with more blades.</p><p>What Gillette may have feared — the expiration of his patents — turned out to be what cemented his razor as a market icon.</p><p>A century later, someone took this razor-and-blades concept and applied it to a completely different business:</p><p>Inkjet printers.</p><h3>A Revolution Called the HP DeskJet</h3><p>The world was a very different place in the late 1980s, especially for those who wanted to print something on paper. At the time, dotmatrix printers dominated the market — slow, noisy machines that produced documents of dismal quality.</p><p>Efforts to change this were already underway. Among them was HP, which, after four years of development, introduced the HP 2680A in December 1980. This departmental laser printer — designed for ofﬁce use and multiple users — made a grand debut. It was, however, enormous, comparable in size to a departmental photocopier. On the upside, it was fast, capable of printing 45 pages per minute.</p><p>Its price was just as massive: $108,500. Understandably, this steep cost deterred potential buyers, and HP didn’t sell a single unit in the ﬁrst two months. It ﬁnally made its ﬁrst sale in the third month. Sales improved slightly after that, but this ﬁrst product was far from a commercial success, with only about 1,200 units sold by the time HP retired it.</p><p>The world stuck with dot-matrix printers a while longer.</p><p>Still, this was only the beginning of what would become a cornerstone division for HP. In 1984, the LaserJet era began, ﬁnally bringing laser printing technology to everyday users, albeit at a signiﬁcant cost — $3,495 at launch.</p><p>But the real revolution wasn’t in laser technology. That same year, HP launched a printer even more important to this story: the HP ThinkJet. BYTE magazine reviewed it in its April 1984 issue with a humorous tone:</p><blockquote><em>“</em>If you’ve been looking for a dot-matrix printer that costs less than $500 and doesn’t sound like an extremely painful visit to the dentist, the HP 2225, to be marketed as Thinkjet, may well be for you. The ﬁrst thing you’ll notice about this new ink-jet printer from Hewlett-Packard is that it’s amazingly quiet.<em>”</em></blockquote><p>And it was. If you’ve ever used a dot-matrix printer, you’ll understand the dental analogy. That technology was loud, and users who switched to these new printers were amazed by their near-silent operation. In noise terms, it was like moving from a combustion engine car to an electric one — an unmistakable sign that times were changing.</p><p>As groundbreaking as that was, the true revolution arrived with color inkjet printing. In 1987, HP released the PaintJet, a limited model with clunky technology that still achieved notable success. By 1988, HP launched its ﬁrst monochrome DeskJet. From then on, the path was clear: the next step had to be a versatile color printer.</p><p>That model took some time to arrive, but in October 1991, HP unveiled the DeskJet 500C. Its successor, the DeskJet 550C in 1992, became HP’s ﬁrst printer to use two cartridges: one for true black — no more mixing colors to simulate black — and another for color.</p><p>And then, eureka.</p><h3>The Curse of Cartridges</h3><p>Inkjet printers quickly became a runaway success. They were more affordable than most laser printers and had one decisive advantage: they could print in color — a fantastic feature for printing photos. For those seeking high-quality results, inkjet printers quickly became the go-to option, especially because of their compatibility with photo paper. While color laser printers — more expensive — were excellent for many tasks, they didn’t support photo paper, which only boosted the appeal of inkjet technology.</p><p>Manufacturers soon realized the goldmine they had on their hands and adopted the model that had worked for King Gillette: sell the printer at a near-giveaway price and make the real money from ink cartridges. When the original cartridges ran out — often sooner than expected — buying replacements was nearly as expensive as buying a new printer.</p><p>The idea was genius for an industry that saw its revenues soar. The problem was that imitators soon emerged. Third-party cartridge manufacturers offered perfectly acceptable and much more affordable alternatives, leading to a bizarre situation.</p><p>Printer manufacturers went to great lengths to block users from accessing these third-party cartridges. They introduced measures like embedding chips in their cartridges to verify their authenticity once inserted into the printer.</p><p>Some third-party manufacturers fought back. In a famous 2003 case, a company called Static Control reverse-engineered the chips created by Lexmark for this purpose. Lexmark sued, and although the initial ruling favored Lexmark, a subsequent legal process found that the company was abusing laws — speciﬁcally the controversial Digital Millennium Copyright Act (DMCA) — to engage in anti-competitive practices. This was a win for third-party manufacturers, establishing a legal precedent for their right to compete.</p><p>The absurdity didn’t stop there. Epson was accused — with solid evidence — of misleading users into thinking their cartridges were empty when they still had 25% of their ink remaining. The normally reserved Dutch Consumer Organization took an unusually strong stance, urging its 650,000 members: “Don’t buy Epson cartridges.”</p><p>HP wasn’t innocent, either. Their printers came with cartridges included, but in some models, those cartridges were only half-ﬁlled. Manufacturers employed other tricks, like requiring users to “calibrate” the printer before printing anything — a process that wasted signiﬁcant amounts of ink.</p><p>Frustrated users — those with technical know-how, determination, and free time — sought solutions. If HP used chips to block third-party cartridges, they would create “anti-chips” to bypass HP’s DRM. This led to a cat-and-mouse game, with manufacturers setting up obstacles and users ﬁnding ways around them.</p><p>The absurdity extended to other brands. In 2020, Canon instructed its customers to hack their own laser printers because pandemic-induced supply chain issues had left them without chips for replacement toner. The manufacturer was effectively hacking itself.</p><p>This trend of restricting replacements spread to other products as well. Dymo label printers integrated DRM not only into their machines but also into the label paper they used. Naturally, hackers found ways to bypass these restrictions. Similar tactics appeared with Da Vinci 3D printers, “smart” cat cages, and more. In each case, counter-chips soon followed.</p><p>HP took things even further in March 2016. Owners of its inkjet printers received a “security update” that seemed harmless enough. But sixmonths later, it was revealed to be a ticking time bomb. In September of that year, the printers began rejecting any cartridge that wasn’t from HP. Despite the backlash, HP repeated the same maneuver a year later.</p><p>The battle was heating up, and third-party cartridge manufacturers increasingly had strong legal arguments in their favor. But HP had an ace up its sleeve.</p><p><em>“Want to know what happened next? We’ll talk about Instant Ink and the Era of Subscription-Based Ink, and also add epilogues with coffee and the return of Mr. Gillette to the story. </em><a href="https://subscriptocracy.com/#/portal/"><em>Subscribe to read the complete chapter</em></a><em> + get one more free chapters from Subscriptocracy.”</em></p><p><em>And remember: the book will be </em><strong><em>30% OFF</em></strong><em> on Black Friday! Stay tuned!</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5da724f24c7d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Spotify]]></title>
            <link>https://medium.com/@javipas/spotify-395eabc4c5b7?source=rss-c6b6b89f5e8------2</link>
            <guid isPermaLink="false">https://medium.com/p/395eabc4c5b7</guid>
            <category><![CDATA[spotify]]></category>
            <category><![CDATA[piracy]]></category>
            <category><![CDATA[streaming]]></category>
            <category><![CDATA[subscription-models]]></category>
            <category><![CDATA[subscription]]></category>
            <dc:creator><![CDATA[JaviPas]]></dc:creator>
            <pubDate>Wed, 19 Nov 2025 19:25:40 GMT</pubDate>
            <atom:updated>2025-11-21T07:21:43.586Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qX0uB7Zj3j8waTe77XMNPg.jpeg" /></figure><h3>From Napster to Spotify: How Music Piracy Shaped the Subscription Economy</h3><p><em>You’re about to read how Spotify cracked the code that changed music forever. This is one chapter from “</em><a href="https://subscriptocracy.com/book"><em>Subscriptocracy</em></a><em>” — my book exploring the subscription revolution. Read 70% here, then subscribe for the full chapter + 2 bonus chapters + insider insights on the subscription economy.</em></p><h3>A Good One Out of Ten</h3><p>Before diving in, let’s take a brief detour — or rather, two.</p><h3>I.</h3><p>In 1996, ‘The Cable Guy’ premiered, but alongside the movie came another release: its soundtrack. On it was a song with a ridiculously long title: ‘Standing Outside a Broken Phone Booth With Money in My Hand.’</p><p>The story behind that track is important for this little aside. The song belonged to the album Rocket by the band Primitive Radio Gods. Chris O’Connor, the band’s lead singer, had spent years trying to make it in the music industry without success. Then, in 1994, while cleaning his house, he stumbled upon a few old demo tapes. He thought, “Why not? What’s there to lose?” and sent some of those demos to various record labels.</p><p>To his surprise, one of them responded, and things moved quickly. Somehow, that one song ended up on the soundtrack of the Jim Carrey movie, sparking interest in the band and their album. But there was a problem. The album was awful. It was because the only decent song on the ten-track album was the one from the movie. The rest were little more than experiments, and as one industry critic aptly put it, “At its core, Rocket feels like a demo tape with one promising track.”</p><p>Which, of course, is exactly what it was — and O’Connor knew it. That didn’t stop the album from becoming a success. Back then, if you wanted to listen to that one good song without buying the movie soundtrack, you had to buy Rocket. But the rest of the tracks? Filler. Everyone knew it: the label, O’Connor, the critics, and the listeners.</p><h3>II.</h3><p>In 1959, 12-year-old Terry McManus bought a copy of Jerry Lee Lewis’s single ‘I’ll Sail My Ship Alone’ from a store in Birmingham, Alabama. Over the years, McManus kept buying singles as a cheap way to decide which albums were worth purchasing. But by the 1980s and especially the 1990s, with the rise of CDs, he noticed something: singles had vanished. He couldn’t ﬁnd them anymore. Neither could anyone else. They had disappeared.</p><p>When Napster emerged in the late 1990s, McManus immediately knew what to call it: “the revenge of the single”. And as we’ll see later, he was right. Why had the industry killed off singles? For music executives in the 1990s, the answer was simple: singles weren’t proﬁtable. They were expensive to produce and didn’t sell enough to justify the cost. So, the industry boiled music sales down to a simple formula: 2 good songs + 10 to 12 mediocre ones = a $15 CD.</p><p>This is how one-hit wonders sold millions of albums only to vanish soon after. Remember ‘How Bizarre’ by OMC or ‘Tubthumping’ by Chumbawamba? And who can name another song by Los del Río besides ‘Macarena’? Yet for a time, if you wanted to own those hits, the only way to do so was to buy the entire CD. Even though most of the other tracks were mediocre at best. Fans knew it, artists knew it, and, of course, the industry knew it too.</p><p>These two anecdotes make it clear: this was how the business worked. Hold onto that thought — that not so long ago, you couldn’t buy individual songs. You had to buy the whole album, even if you didn’t care for the rest of it. Let’s move on.</p><h3>Enter the MP3 Format</h3><p>The internet was still in its infancy when the MP3 was on the verge of taking off. Sharing ﬁles in this format was cumbersome — nostalgia for the screech of dial-up modems aside, transfer speeds were painfully slow — and the number of internet users was relatively small.</p><p>But by 1997, the legendary Winamp player exploded in popularity,marking a turning point. The internet was becoming more established, and the convergence of these two phenomena led humanity to do something it excels at.</p><p>Sharing.</p><p>In that case, sharing MP3 ﬁles. People had shared cassette tapes before, but the small size of MP3 ﬁles made them much easier to distribute over the internet. By January 1999, for example, the term “MP3” had surpassed “sex” as the most searched word on the internet. MP3s were thriving on PCs, but they also paved the way for new devices.</p><p>In May 1998, the ﬁrst hardware MP3 player, the MPMan, wasreleased. It was expensive and clunky, but it worked. Soon after, other devices followed — going beyond the Walkman and Discman. These players could store hours of music in your pocket, with spectacular sound quality, in a digital format, and with impressive features.</p><p>Not long after, Diamond Multimedia launched its Rio PMP300 (for Portable Media Player), the ﬁrst commercial MP3 player to achieve real success. The Recording Industry Association of America (RIAA), increasingly worried, sued the company, claiming it violated copyright laws. The RIAA lost the case. That little device marked a signiﬁcant turning point, signaling the inevitable decline of Walkmans and Discmans and the promising rise of MP3 players. Eventually, the device that truly changed everything arrived — and it did so with the help of the MP3 format.</p><p>That device, of course, was the iPod.</p><p>But we’re getting ahead of ourselves. There are a few more things to<br>cover ﬁrst.</p><h3>Napster Shows Us the Future of Music Distribution</h3><p>Months after the lawsuit between the RIAA and Diamond Multimedia, a tumultuous year for the music industry began. In June 1999, a young programmer named Shawn Fanning launched Napster, the MP3 ﬁle-sharing service that revolutionized the music world. Napster didn’t just show the world what the future of distribution could look like; it demonstrated it in action.</p><p>In that future, music wasn’t just free — a major draw for most users — it was also free of restrictions. It bypassed the artiﬁcial locks (like DRM, or Digital Rights Management) that other platforms had imposed. Napster offered access to hard-to-ﬁnd songs that traditional media outlets neglected, but it went further: it redeﬁned how people consumed music by enabling the exchange of individual tracks rather than entire albums.</p><p>Gone were the days of buying a full vinyl LP or CD just to get the two songs you liked while slogging through the other eight or ten ﬁller tracks. Napster offered a solution to what many saw as an exploitative industry practice. Remember the case of O’Connor’s hit from the Jim Carrey movie? Exactly. Napster eliminated that frustration. It was revolutionary.</p><p>But the music industry’s power players quickly saw Napster as a threat. From day one, they fought to shut it down. Their initial response was clumsy: several companies worked together to create the SDMI (Secure Digital Music Initiative), a rights management platform co-developed with Microsoft and IBM.</p><p>That technological counterattack failed spectacularly, so the industry started to use more brutal tactics. Napster, sued by artists like Metallica and several record labels, ultimately lost the legal battle and was almost entirely dismantled by 2001. Despite this, both the industry (begrudgingly) and users learned an essential lesson: as Bob Dylan once sang, “The times they are a-changin’.” While independent Napster alternatives — Morpheus, Grokster, Gnutella (created by Winamp developer, Justin Frankel), and KaZaA,</p><p>among others — sprang up, record labels tried to regain control by launching their own platforms. Warner, Bertelsmann, and EMI partnered with RealNetworks to create a subscription service — yes, a subscription service — called MusicNet. It was an absolute disaster. Universal and Sony followed suit with a similar service called Pressplay. At the same time, Rhapsody emerged, offering a subscription-based model that allowed access to a digital music catalog.</p><p>But let’s not forget: this was 2001, and people were hardly subscribing to anything. The concept was new, and while the idea had potential, its execution and terms were abysmal. Consider the conditions of Pressplay, for example. The $9.95 monthly basic plan allowed users to stream up to 500 songs in low quality, download 30 tracks, and burn just 10 songs to a CD. Even worse, not all songs were available for download, and users couldn’t burn more than two songs from the same artist on a single CD. Downloads expired after a month, and they couldn’t be transferred to portable music players.Awful.</p><p>Pressplay was the perfect example of the music industry’s possessive attitude toward its assets. They were unwilling to let go of control, and users were equally unwilling to accept such a mess. The industry’s attempts at creating their own platforms were largely ignored or ridiculed. Both Pressplay and MusicNet landed on PCWorld’s infamous list of “The 25 Worst Tech Products of All Time,” becoming case studies in how not to launch a digital distribution platform.</p><p>The result? Chaos. Peer-to-peer (P2P) ﬁle-sharing applications ﬂourished, but they required internet access and some technical know-how to be used. Meanwhile, record labels failed over and over again, releasing poorly executed platforms. There was no legitimate, user-friendly alternative, and that’s when Steve Jobs saw his chance. Apple’s co-founder identiﬁed the problem and used his inﬂuence to negotiate with the music industry’s Big Five. Having failed to adapt to the coming changes, the labels were open to Jobs’s proposal, which promised beneﬁts for both Apple and the music industry. His plan included two key concessions that pleased the record executives.</p><p>First, the iTunes Music Store would initially be available only to Apple computer users. To music executives, this was low-risk: Mac’s market share was minuscule, making it a safe test bed for gauging revenue potential without signiﬁcant exposure. The second condition was even better for the industry: Jobs agreed to protect the music catalog with DRM. This ensured the labels retained control, as songs purchased on iTunes could only be played on Apple devices. The record labels approved, and the project received the green light.</p><p>Of course, Jobs had long-term plans in mind. The iTunes Music Store, announced in April 2003, expanded to Windows users just ﬁve months later, opening up the iPod to a much broader audience. Meanwhile, Apple’s DRM system, while initially critical, gradually faded. Jobs himself criticized DRM in a 2007 open letter titled “Thoughts on Music.” On it, he outlined his vision for the future of music distribution: “Eliminate DRM entirely.”</p><p>This vision became reality for iTunes in 2009, but not just because Jobs was a visionary or an advocate for consumer rights. DRM had actually been advantageous for Apple, as it reinforced the company’s control over its ecosystem — a strategic move that solidiﬁed Apple’s dominance in both hardware (iPod) and digital distribution (iTunes). The relationship was a virtuous cycle: one drove growth for the other. Perfect synergy.</p><p>Then Amazon arrived, offering record labels the chance to distribute DRM-free music as unprotected MP3 ﬁles that could be played anywhere — even on iPods. This provided the music industry with a way to bypass Apple’s dominance in digital distribution. The labels embraced Amazon’s model, leading Apple to remove DRM from iTunes shortly after.</p><p>This shift marked a turning point: for the ﬁrst time, users had a legitimate, affordable alternative that functioned much like the traditional music market — something vital for both the industry and user adoption. It wasn’t a subscription service; the music was yours to keep. But with the advent of this model came another fundamental shift — one that, as we’ll explore next, was inspired by Napster.</p><h3>Is That You, Steve?</h3><p>There he was. Steve Jobs, walking across the stage happily. He was wearing his usual jeans, New Balance sneakers, and black turtleneck sweater. A little chubbier, with noticeably less hair, unshaven, and slightly scruffy. But that didn’t matter — his aura was undeniable, and everything he said just seemed to make sense.</p><p>It was April 2003. The Moscone Center in San Francisco was packed. Steve paced back and forth, speaking with precision and conﬁdence. No mistakes. No hesitation. “Starting in 1999 there was this phenomenon called Napster,” he remarked at one point. “It was shut down in 2001, but it demonstrated some things for us,” he continued. “It demonstrated that the internet was made for music delivery.”</p><p>He went on to mention KaZaA, which was still thriving at the time, providing that all-important “near instant gratiﬁcation” to users — a term Jobs used deliberately. While not widely recognized then, the concept would soon become a part of everyday vocabulary. However, Jobs noted, “The downside” of platforms like Napster and KaZaA was simple: “it’s stealing. It. Is. Stealing.”</p><p>After weighing the pros and cons, he posed a critical question: “Why has this proliferated?” The answer he gave was as blunt as it was truthful. “Well,” he concluded, “because there’s no legal alternative.” And he was right. The music industry at the time offered no meaningful options for buying or listening to music online. Record labels, basking in their own success, were reluctant to adapt. CDs were a goldmine: production costs were minuscule by then, and proﬁt margins were soaring. Jobs knew this well because, in reality, there were alternatives.</p><p>It was time to discuss them, and Jobs did so in his signature style — with a rhetorical question. “So, what about Pressplay and Rhapsody?” Both platforms had aimed to provide a legal alternative to Napster, and despite backing from the music industry, they hadn’t gained traction. Jobs believed he knew why.<br>“The reason?” he said. “Both are subscription services.”</p><p>To Jobs, that model wasn’t the right one. “You can’t just pick a song and pay a little bit of money,” he explained. These platforms forced users to pay a monthly fee, but “if you stop paying the subscription fee, your entire music library disappears.” The problem, according to Jobs, was that “these services treat you like a criminal.” He wasn’t talking about the now-defunct Napster or KaZaA. He was criticizing legitimate services that relied on streaming technology to let users play their favorite songs wherever and whenever they wanted. But the music wasn’t truly theirs — it was only accessible as long as they kept paying.</p><p>At the time, Jobs couldn’t make sense of it. He seized the moment to deliver a dramatic pause. Then, an image of someone dressed as a prison inmate, looking like a mobster, appeared on the screen. Jobs drove his point home: “These platforms are based on subscriptions, and we think subscriptions are the wrong path.” People had always bought their music, he explained. “We’ve bought music on LPs, we’ve bought music on cassettes, we’ve bought music on CDs.”</p><p>For Jobs, this was the way music consumption worked. And that’s exactly what he proposed on April 28, 2003: just as people had bought music in those physical formats, now they could purchase it by downloading songs. “People are used to buying their music and having a broad set of rights when they do. When you buy your music, it never disappears.”</p><p>That was the announcement of the iTunes Music Store. Jobs had managed to secure deals with the ﬁve major players in the music industry — BMG, EMI, Sony, Warner, and Universal. Apple launched its muchanticipated music store, integrated into iTunes 4, with an initial catalog of 200,000 songs.</p><p>You could burn them to a CD, though not without limitations: there were restrictions on how many times a playlist could be burned. You could play the songs on up to three Macs. And, of course, you could sync them with your iPod, the industry-deﬁning music player. The price? 99 cents per song. No strings attached. “No subscription fee,” Jobs emphasized. The music was yours — and yours in a meaningful way. No need to buy an entire album when you only loved one track. You owned just the music you loved. Nothing more, nothing less.</p><p>Or was there more?</p><p>At that moment, there wasn’t. The iPod, which had launched in 2001, began to pick up steam — thanks in large part to the release of iTunes for Windows. By 2005, Apple had sold 22.5 million iPods. By 2008, that ﬁgure would reach its peak at nearly 55 million.</p><p>But the iPod had its nemesis within Apple itself. After the iPhone’s 2007 debut — launched at a time when the iPod reigned supreme — sales of Apple’s smartphone soared, relegating the iPod to a quiet second place. In 2014, Apple sold 14.38 million iPods, but the iPhone, which could do everything the iPod could and more, achieved stratospheric success, with 169 million units sold. That was only the beginning: the following year, iPhone sales hit 231 million and remained at that level for years to come.</p><p>But then something shifted in a distant Nordic land.</p><p><em>“Want to know what happened next? </em><a href="https://subscriptocracy.com/#/portal/"><em>Subscribe</em></a><em> to read the complete Spotify chapter + get 2 more free chapters from Subscriptocracy.</em>”</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=395eabc4c5b7" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How a 99¢ Printer Subscription Led Me Down a Rabbit Hole About the Future of Ownership]]></title>
            <link>https://medium.com/@javipas/how-a-99-printer-subscription-led-me-down-a-rabbit-hole-about-the-future-of-ownership-269a9e857068?source=rss-c6b6b89f5e8------2</link>
            <guid isPermaLink="false">https://medium.com/p/269a9e857068</guid>
            <category><![CDATA[digital-economy]]></category>
            <category><![CDATA[subscription]]></category>
            <category><![CDATA[business]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[stories]]></category>
            <dc:creator><![CDATA[JaviPas]]></dc:creator>
            <pubDate>Fri, 14 Nov 2025 22:07:52 GMT</pubDate>
            <atom:updated>2025-11-15T09:41:51.277Z</atom:updated>
            <content:encoded><![CDATA[<p>Three years ago, I started writing what I thought would be a small, harmless blog post.</p><p>I had been using <strong>HP Instant Ink</strong> for a few months and wanted to share my experience. For €0.99 a month I could print up to 15 pages. Convenient, predictable. A small subscription for a small problem.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dKMqfLdTpP2jhZNhZbBNwg.jpeg" /><figcaption>That’s the email I got on Sep 16, 2021. My subscription to HP Instant Ink had started. It was the beginning or a really fun journey.</figcaption></figure><p>Nothing remarkable… until it was.</p><p>As I dug deeper, I realized HP wasn’t just charging for ink. They were following the famous <strong>razor-and-blades model</strong>, the idea that you sell the main product cheaply so you can make money from the consumables.</p><p>That took me down a rabbit hole.</p><p>I found the classic 2011 <a href="https://chicagounbound.uchicago.edu/uclrev/vol78/iss1/12/">critique</a> by <strong>Randal C. Picker</strong>, which led me to the story of <strong>King C. Gillette</strong>, which <a href="https://en.wikipedia.org/wiki/William_Painter_(inventor)">led me</a> to <strong>William Painter</strong>, the inventor of the <a href="https://en.wikipedia.org/wiki/Crown_cork">bottle crown cork</a>, who mentored him. Before I knew it, I wasn’t writing a blog post anymore — I was tracing the origins of modern capitalism through ink cartridges and razor blades.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tSK-eflGpQZXlc3fLfwt7w.jpeg" /><figcaption>The Star Safety Razor in the Montgomery Ward Catalogue, No 57, 1895. The razor-blades model started pretty much there.</figcaption></figure><p>Then something clicked.</p><p>What HP had done with printers wasn’t just “razor and blades.” It was something bigger.</p><p><strong>Subscriptions.</strong></p><p>A model that offered companies:</p><ul><li>Predictable, recurring revenue</li><li>Customer lock-in</li><li>And, most importantly, <strong>control</strong> — over pricing, access, and even the product itself</li></ul><p>I began to see the same pattern everywhere.</p><p><strong>Spotify. Adobe. Google. Apple. Netflix. Amazon. Microsoft. </strong>Even cars, doorbells, toothbrushes, coffee, shoes and refrigerators.</p><p>Bit by bit, the world was shifting from <em>owning</em> to <em>renting </em>and <em>accessing</em>. It was a book.</p><p>And I wanted to understand why.</p><h3><strong>So I wrote a book about it</strong></h3><p>That curiosity became a long research project, which eventually became a book:</p><p><a href="https://subscriptocracy.com/book"><strong>Subscriptocracy</strong></a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kMNHpiR08Z6-ZCrpP3bX_A.jpeg" /></figure><p>A deep dive into how subscriptions took over our digital — and physical — lives.</p><p>I self-published it in Spanish in December 2024 and then in English in April 2025. Life, work, family… everything got in the way, and I didn’t have time to truly launch it.</p><p>So I’m doing it now.</p><h3>🎁 Want to read a free chapter?</h3><p>If you’re curious about how subscriptions are reshaping the world — and what we’ve lost along the way — I <a href="https://subscriptocracy.com/">created a newsletter</a> where I’ll be sharing:</p><ul><li>Free chapters</li><li>Behind-the-scenes stories</li><li>Research notes</li><li>special offers</li><li>and a few surprises</li></ul><p>You can download the <strong>Spotify chapter</strong> (one of the key ones) for free here:</p><p>👉 <a href="https://subscriptocracy.com/#/portal/signup/free">https://subscriptocracy.com/#/portal/signup/free</a></p><h3>Why subscribe?</h3><p>Because the shift from ownership to subscriptions isn’t slowing down.<br> It’s accelerating.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*P2gI8ELn3K8slpkcK68VWQ.jpeg" /></figure><p>And understanding it today means understanding:</p><ul><li>how companies think</li><li>why products are changing</li><li>what we actually pay for</li><li>what we gain</li><li>and what we lose</li></ul><p>If that interests you, I’d love to have you.</p><p><strong>Cheers,</strong><br> <em>Javier Pastor</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=269a9e857068" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The web is disappearing]]></title>
            <link>https://medium.com/@javipas/the-web-is-disappearing-dec020fc6c49?source=rss-c6b6b89f5e8------2</link>
            <guid isPermaLink="false">https://medium.com/p/dec020fc6c49</guid>
            <category><![CDATA[publishing-industry]]></category>
            <category><![CDATA[media]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[web]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[JaviPas]]></dc:creator>
            <pubDate>Tue, 03 Jun 2025 10:05:01 GMT</pubDate>
            <atom:updated>2025-11-09T10:25:51.457Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EFlsVE01SLwOcw0T7jX9zg.jpeg" /></figure><p>—Do not try and bend the spoon. That’s impossible. Instead, only try to realize the truth.</p><p>— What truth?</p><p>— That there is no spoon.</p><p>You probably remember the scene from ‘<a href="https://www.imdb.com/es-es/title/tt0133093/">The Matrix</a>’ (1999) that goes with that quote. It was Andrés Torrubia who reminded me of it, when he <a href="https://x.com/antor/status/1927136975565242846">tweeted something</a> striking the other day about the impact of AI on the internet. Maybe you’ve already heard about it — but just in case you don’t want to read any further, I’ll give you the conclusion of this post up front:</p><blockquote>The web is disappearing.</blockquote><p>Not dying, mind you — disappearing. Fading.</p><p>And it’s happening because of an AI that is slowly consuming the web. It’s doing it quietly, at least in this specific aspect. We’re not noticing it — or rather, most aren’t, but I am — because you’re not clicking links anymore. And the reason is simple: links are disappearing from our lives.</p><p>That’s exactly what I wrote about this weekend <a href="https://www.xataka.com/robotica-e-ia/enlaces-unidad-basica-para-relacionarnos-internet-ia-quiere-acabar-ellos">on Xataka</a>. The web has always been, above all, about links. Links that take you from one place to another, allowing you to connect knowledge, dig deeper, explore, be amazed (or horrified, there’s plenty of that too). Links were always the glue of the internet, the foundation of the World Wide Web, a structure that is now being submerged and hidden. It’ll still be there, but we’ll barely see it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vXb3qP-aHNc05TqD6ZYDFQ.jpeg" /><figcaption>Just look at Google’s AI Mode giving me information about “Incognitosis.” Those links on the right are the only strong visual cue that they exist. The rest? Chain icons in some paragraphs — gray, barely noticeable. Gone are the underlined, colored texts we all recognized as links.</figcaption></figure><p>The blame lies with an unstoppable, ruthless artificial intelligence that is becoming our interface to the web. We won’t need links, because AI will answer everything — and better than ever. We’re already seeing it with AI search engines. In the image above, Google’s AI Mode doesn’t underline or color links, which was how we used to recognize them.</p><p>Why doesn’t it do that? Why only show those faint gray “chain” icons? The answer: to hide the links. To camouflage them.</p><blockquote>To stop us from clicking.</blockquote><p>And so, the original concept of the web is being dismantled. The web will still exist, yes, but AI will be the one visiting it — via agentic browsers. It’ll be AI that clicks, reads, and analyzes the content for its own use.</p><p>This has one major consequence: if almost no one clicks, traffic to media and creator websites will collapse. It’s already happening.</p><p>Ángel L. Fernández Recuero, one of the founders of Jot Down, wrote about it recently. We’re talking about one of the highest-quality content Spanish sites on the internet — original, in-depth, thoughtful. No rehashes or junk. A top-tier website. And yet, here’s what Fernández Recuero wrote in his excellent and bleak piece <a href="https://www.jotdown.es/2025/05/que-hacer-cuando-nadie-nos-busca/">‘Qué hacer cuando nadie nos busca’</a>:</p><blockquote>“In the first five months of 2025, Jot Down’s digital edition lost 35.8% of its readers. Sessions dropped 40.5%, page views 15.9%. Google brought us 31.8% fewer visits than last year. Twitter, 56% fewer. Facebook, 35%. These aren’t measurement errors or a temporary dip. This is the beginning of a clickless era. It’s not just happening to us — this is a global phenomenon affecting all media that rely on open access, search engines, and social traffic to be read. The web, as we knew it, is ceasing to exist. Artificial intelligence isn’t the future — it’s already here, right in front of our content. Google now replies with AI Overviews that summarize our words without needing to visit us. ChatGPT, Perplexity, Claude, and other generative models deliver full answers — often without links, without visible sources, without credit. More and more people are asking questions and getting answers without ever landing on our pages. As if we were just… in the way.”</blockquote><p>Hard to say it better — or more bluntly. That 30–40% drop that’s hit Jot Down isn’t an isolated case. It’s what other media outlets worldwide are seeing. Mail Online <a href="https://pressgazette.co.uk/publishers/digital-journalism/google-ai-overviews-leads-to-dramatic-reduction-in-click-throughs-for-mail-online/">has lost 50% of its traffic</a>. An <a href="https://ppc.land/google-ai-overviews-cut-traffic-by-34-as-publishers-demand-action/">analysis</a> by Ahrefs — a well-known name in media circles — found that when Google’s AI Overviews appeared at the top of search results, the first organic link lost an average of 34.5% of clicks.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zvQIJyZvpa-eJgeMbnTuew.jpeg" /><figcaption>Source: <a href="https://ppc.land/google-ai-overviews-cut-traffic-by-34-as-publishers-demand-action/">PPC Land</a></figcaption></figure><p>That little chart explains it all quite well. Media outlets essentially live off advertising. You offer your content — sometimes better, sometimes worse — for free, and that brings traffic. With that traffic you display ads, and that’s how you earn a living. That’s what we do at Xataka, and it’s what most online media do. But if Google stops sending you traffic and links fade away, then what?</p><blockquote>Then advertisers won’t run ads — because no one will see them.</blockquote><p>The threat is clear and terrifying for those who live off content. You might say TikTok, Instagram, and YouTube are safe — but I’d say not really. The incoming wave of “AI slop” (AI-generated junk) is going to be so massive that most creators simply won’t be able to compete with the automatic content machines.</p><p>There will still be quality “human” content, and even carefully curated AI content made with care by real people using these tools creatively — no doubt. But I can’t shake the feeling that this threatens the entire era of content we once enjoyed thanks to the web and social media. As Cloudflare’s CEO recently said, <a href="https://searchengineland.com/ai-killing-web-business-model-455157">AI is killing the web’s business model</a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Zn7aERjKINPiAWnazzujrw.jpeg" /><figcaption>Hello? Did someone say subscriptions? I’ve written a book about this business model, and <a href="https://www.amazon.com/Subscriptocracy-Subscriptions-Took-Over-World-ebook/dp/B0F3V252G3?crid=2AJ4SA78UQ3ZA&amp;dib=eyJ2IjoiMSJ9.v3gDQtktTfMOHKhR21Qu-g.p48i4DbDwOaIRL8bD_AKlWom61fcoeQDh_Ozx1T0yJc&amp;dib_tag=se&amp;keywords=subscriptocracy&amp;qid=1748944689&amp;sprefix=subscriptocracy%2Caps%2C184&amp;sr=8-1">it’s also available in English!!</a>. In fact, it’s available <a href="https://www.amazon.com/Subscriptocracy-Subscriptions-Took-Over-World-ebook/dp/B0F3V252G3/ref=tmm_kin_swatch_0">for free for Kindle Unlimited subscribers</a>.</figcaption></figure><p>Are there options left for media outlets and creators? In my view, there’s only one viable short-term path: adopt the subscription model — long live the <a href="https://javipas.com/subscriptocracy">Subscriptocracy</a> — and put up paywalls so only subscribers can access the content. I’m not even sure that will be enough — AI could subscribe and still scrape the content — but as people have been saying for a while, what’s coming will likely split media into two camps.</p><p>On one side, the major outlets like The New York Times. On the other, the small-scale, often solo operations — like a potential paid version of <a href="https://javipas.com/">Incognitosis</a>. The former will have a large community of loyal readers. The latter will depend entirely on their following — on how many people are true fans. It’ll be <a href="https://javipas.com/2022/06/14/la-vida-con-1-000-fans/">the 1,000 true fans model</a> — and not many will reach it.</p><p>I don’t know how you all see it, but personally, the future looks pretty grim. AI is amazing, sure — but it’s going to change a lot of things. And many of us won’t be ready for it. I’m setting a reminder to come back and reread this post in a year, but I think the changes will come very, very fast.</p><p>Bad times ahead.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dec020fc6c49" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[That’s how I’ve replaced Google Photos]]></title>
            <link>https://medium.com/@javipas/thats-how-i-ve-replaced-google-photos-0f5723719a7d?source=rss-c6b6b89f5e8------2</link>
            <guid isPermaLink="false">https://medium.com/p/0f5723719a7d</guid>
            <category><![CDATA[cloud-computing]]></category>
            <category><![CDATA[google-photos]]></category>
            <category><![CDATA[immich]]></category>
            <category><![CDATA[photo-management]]></category>
            <category><![CDATA[self-hosted]]></category>
            <dc:creator><![CDATA[JaviPas]]></dc:creator>
            <pubDate>Sat, 18 Jan 2025 15:20:05 GMT</pubDate>
            <atom:updated>2025-01-18T15:23:19.213Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Disclaimer: this is a translation from </em><a href="https://javipas.com/2025/01/07/asi-me-he-montado-mi-propio-google-fotos/"><em>the original post in Spanish</em></a><em>. A lot of links here (but not all) go to posts written in Spanish there, but there’s no need to read those. Not unless you want to practice your Spanish or use a translator ;) This long article has been translated with ChatGPT and edited for clarity.</em></p><p>People that have read me through the years in my blog at <a href="https://javipas.com/">Incognitosis</a> know that I have a special obsession for having my photos well organized. In recent years I have been trying different alternatives, and always with the same base:</p><blockquote>Having my photos in the cloud is not my preferred option.</blockquote><p>I know it is for most people. It’s normal. Google Photos is a <a href="https://javipas.com/2017/05/18/me-aterroriza-google-photos/">fantastic and terrifying platform</a>, and for those who use an iPhone the possibility of saving everything in iCloud and then working with Apple Photos is equally great. It’s first and foremost convenient, because you don’t have to worry about practically anything other than paying for that storage space. It’s a great option for a lot of people, and I think it’s perfect that it’s that way. Long live the options.</p><p>But I’m not really into that. I don’t want my photos swarming around. I’m not one to post them on social media — <a href="https://javipas.com/2017/07/13/instagram-como-escaparate-de-las-vanidades/">I’m an Instagram hater</a> — and I don’t even like friends tagging me when they post a group photo.</p><p>I’m especially <a href="https://javipas.com/2022/11/07/fotos-de-ninos-en-redes-no/">hysterical about photos of my kids</a>: I don’t want them to be out there, and I’ve had the occasional family uproar over it. It’s inevitable that it happens from time to time — I make exceptions myself — but as I say, I try to minimize it.</p><h3>Years of trying alternatives</h3><p>In the past I’ve talked about those attempts to have my photo library well organized. In case someone wants to revisit them, the most relevant posts on the subject have been:</p><ul><li><a href="https://javipas.com/2012/12/16/cual-es-el-mejor-servicio-online-de-backup-y-acceso-a-las-fotos/">What is the best online photo backup and access service?</a> (December 2012): I was not at all clear on how to manage my photos, and asked how people did it. I was even considering whether to go to a cloud service, at least to back up my library.</li><li><a href="https://javipas.com/2021/04/08/prescindir-de-google-es-para-valientes/">Not using Google is for the brave ones</a> (April 2021): about alternatives such as PhotoPrism or Photostructure to set up your own Google Photos.</li><li><a href="https://javipas.com/2021/04/08/prescindir-de-google-es-para-valientes/">Testing PhotoStructure: a candidate to be the Google Photos I control</a> (April 2021): just below I talked about how I started testing Photostructure, which certainly had (and has) an original and great interface.</li><li><a href="https://www.deepl.com/">How I’ve set up my own Google Photos</a> (July 2022): here I explained my evolution when it comes to managing photos and told how I ended up opting for a proprietary but promising solution, a little gadget called Monument 2 with its own specific software for this task.</li><li><a href="https://www.deepl.com/">Why spend 600 euros on a Synology NAS when you can get one for 200</a> (July 2023): how I built a miniPC with Xpenology, a Synology clone, to take advantage of Synology Photos.</li></ul><p>I recommend reading the fourth one, because it was the first real attempt to solve my problem. It was also a post in which I recounted my history with photo management applications.</p><h3>First attempts: from Picasa to Monument 2</h3><p>For many years I used Picasa, the old Google application, on my Windows PC. For those of you who don’t know it, this tool was essentially a local Google Photos. The interface was crappier, yes, but its organization options were fantastic (with geotagging, for example) and it even implemented facial recognition to group people in the photos by faces. That’s quite surprising, as this could already be done in 2009. It was a fantastic application, but Google got rid of it to bet on its cloud services, specifically Google Photos.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Fqb2uMP_mOJNEJUwXT07hg.png" /><figcaption>Picasa was wonderful. Google, of course, killed it.</figcaption></figure><p>I was never convinced about that cloud thing, so for several years I had the photos stored and classified by folder on my NAS (with the format “YYMMDD — Folder Description”) but without being able to see them comfortably. I installed the old version of Picasa on a PC and configured it so that the photos were centralized, and that offered me at least that alternative to access them from any computer that had Picasa installed on my local network. Synchronizing the photos from the mobile was not convenient, and from time to time I had to spend a few days copying them from the mobile to the PC, organizing them manually by folders and then leaving them on the NAS to update everything. It was quite crafty, but it worked.</p><p>But COVID-19 came and in Spain we were in confinement at home for around three months. We had enough time available because we were no longer going out, so my wife and I started to organize the photos, which we had a bit disorganized in several hard drives and folders scattered around. After several weeks of effort, we finally centralized everything into a single, massive folder. Inside, all the subfolders were properly organized and formatted as planned. That was an important step for what I wanted, which was precisely to find an alternative to Google Photos.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5TjUFwS50_kF0mSCD_mEhQ.jpeg" /><figcaption>Photostructure has a great and original interface, but if you want to use it to the fullest, you have to pay (a little).</figcaption></figure><p>The search process went slowly, but in 2021 I thought <a href="https://javipas.com/2021/04/26/probando-photostructure-candidato-a-ser-ese-google-photos-que-yo-controlo/">I had found a great candidate with Photostructure</a>. However, I never ended up mounting it on a PC or miniserver, mainly because the project was under development and many of its options were limited. Even worse, the features I found most interesting required payment. Something as simple as marking a photo as a favorite required a monthly or annual subscription. While the cost wasn’t particularly high (<a href="https://www.getmonument.com/m2/monument-2-photo-storage-device">$42 per year</a>), it still annoyed me. I thought to myself: for that price, I might as well pay for Google Photos and forget about the hassle.</p><p>So I kept searching, and in July 2022, I shared here that I had found a potential candidate: <a href="https://www.getmonument.com/m2/monument-2-photo-storage-device">Monument 2</a>. It was a very basic gadget built on a Raspberry Pi-like board with proprietary software that promised to handle everything seamlessly — or so its creators claimed when they launched it on Kickstarter. On paper, as with most Kickstarter projects, the idea sounded amazing. In practice, however, it had numerous issues. These ranged from problems with SSD drive compatibility to the software itself, which was unreliable and prone to erratic failures.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xvtZQhsIJmtXdqLI3_bl1Q.jpeg" /><figcaption>Monument 2 had promise, but its creators ended up abandoning it to its fate.</figcaption></figure><p>The interface was fantastic — very reminiscent of Apple Photos or Google Photos. The mobile app was super cool, too. However, as I mentioned, the performance was inconsistent. The worst part, though, was that the development team behind the project gradually abandoned it, barely saying a word. The forums on <a href="https://www.kickstarter.com/projects/monument/monument-2-smart-photo-storage-and-organization/comments">Kickstarter</a> and <a href="https://www.kickstarter.com/projects/monument/monument-2-smart-photo-storage-and-organization/comments">Indiegogo</a> make it painfully clear: many people who ordered the Monument 2 waited months and months to receive it. Technical issues rarely got a response, and the software updates were sparse and poorly executed, despite users submitting some genuinely interesting requests.</p><p>I did manage to get <a href="https://www.kickstarter.com/projects/monument/monument-2-smart-photo-storage-and-organization/comments">one of my suggestions implemented</a>, though — a “Surprise Me” mode similar to PhotoStructure’s, which displays a random selection of photos at startup. But I suppose they didn’t sell as many units as they hoped (around 10,000, according to both crowdfunding platforms), and the team must have decided it wasn’t worth putting in more effort. A real shame.</p><h3>Synology Photos: Almost Perfect</h3><p>Over time, my frustration grew, so I started searching for alternatives. I discovered that both Synology and QNAP offered their own “Google Photos” solutions — Synology Photos and QuMagie, respectively. I seriously considered <a href="https://javipas.com/2021/01/18/a-la-caza-de-un-nuevo-nas-synology-ds720-o-qnap-ts-253d/">buying one of these NAS devices</a> to take advantage of those options. Synology Photos especially appealed to me because its interface mimicked Apple Photos and Google Photos. The downside? I found their NAS devices prohibitively expensive.</p><p>After some research, I discovered I could build a DIY Synology NAS clone <a href="https://javipas.com/2023/07/03/para-que-gastar-600-euros-en-un-nas-de-synology-cuando-puedes-montarte-uno-por-200/">using the Xpenology project</a>. That’s exactly what I did. It saved me a lot of money, and it worked as I expected. But even then, I couldn’t shake the feeling that something was missing.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*NMxEOh9EI32gjWXyXyomVg.jpeg" /><figcaption>Synology Photos boasts a fantastic interface and excellent features. Honestly, it’s almost perfect.</figcaption></figure><p>I’ll admit it — Synology Photos is great. It’s like a modern version of Picasa: it respects the folder structure I’ve already set up while also applying its own chronological organization by default. The mobile sync is excellent, and overall, I’d say the platform is nearly flawless.</p><p>However, two key issues eventually made me reconsider:</p><ol><li><strong>No Keyboard Shortcuts.</strong> While the web interface is fantastic, it lacks something crucial for me: keyboard shortcuts. This has been a <a href="https://community.synology.com/enu/forum/1/post/145837">frequent request in their official forums</a>, yet basic actions like deleting a photo or marking it as a favorite require multiple mouse clicks. To delete a photo, for instance, you have to click on it, then click on the trash icon. To mark it as a favorite, you need to open the info panel and assign it a rating from one to five stars. Speaking of stars, while they’re a nice feature, they overcomplicate things. I’d prefer a simple heart or star to designate favorites — no need to split hairs over levels of favoritism.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RjxhXE_rbI2MNCa9oJ9Btg.jpeg" /><figcaption>Stars are OK, but they add extra work to the photo management tasks.</figcaption></figure><ol><li><strong>Lackluster Android TV app.</strong> While viewing photos on a mobile device or computer works wonderfully via the Android/iOS app or web interface, things fall apart when you try to enjoy them on a TV. For me, this is essential — I love sitting down with my wife and the kids to revisit old photos together. Unfortunately, none of the apps I’ve tested, including Synology Photos, seem to handle this well. It does have an Android TV app, but it’s clear that this version has been neglected. Even though they hyped it up when they launched it in December 2022, there haven’t been any significant updates since then.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*s04-bXHn6HxSct4SAJMaig.jpeg" /><figcaption>Synology Photos Android TV app interface. Try to navigate and find a photo from 2013. You’ll be scrolling down for an hour if your photo library is big.</figcaption></figure><p>The biggest issue with the Synology Photos Android TV app? There’s no chronological navigation bar. If you want to jump to photos from, say, March 2013, there’s no quick way to do it. My library contains around 170,000 images and videos as of 2024, so scrolling to March 2013 would take — no exaggeration — minutes, maybe hours. There’s no search functionality to help either. What would a better solution look like? Something like this: [Insert user idea here].</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_2m26U82H05xQENoFhOYjQ.jpeg" /><figcaption>Sketch of a potential future version with acronological navigation bar and a search option would be great on the Android TV app.</figcaption></figure><p>But as it stands, the app feels abandoned, making it inconvenient to enjoy photos on the living room TV or projector. Sure, you could use the web browser on a TV, but then you’re back to the same issue: no keyboard shortcuts.</p><p>These frustrations drove me back to the drawing board, searching once again for a better solution. This time, I delved into the world of self-hosted photo management alternatives. What followed was a little adventure.</p><h3>Hello, Unraid</h3><p>The idea of setting up a new NAS to retire my old Synology DS-212j was still on my mind. While the Xpenology server wasn’t bad, the case I had bought wasn’t designed to accommodate multiple SSD or HDD. After spending some time reading and watching videos — <a href="https://www.youtube.com/channel/UCFyP17HoU-vpxhIpGXnXx2g/videos"><em>NASCompares</em></a> became one of my go-to YouTube channels — I discovered there were some really cool alternatives in this space, both for the case and the components.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*H0rM8SugEQd2pGv3ZVpk8w.jpeg" /></figure><p>To keep costs down, I decided not to invest in a new Mini-ITX motherboard, processor, or memory. Instead, I reused the ASRock J5040 motherboard and CPU I had used in my Xpenology build, focusing my budget on two key things: the case and storage. I bought a Jonsbo N2 case ($<a href="https://www.amazon.com/dp/B0BQJ6BCB7?tag=Incognitosis-21&amp;linkCode=osi&amp;th=1">149.98 on Amazon</a>), which I loved for its design and disk bay system, and three factory-refurbished HGST He10 10TB SATA drives — these come from data centers that periodically replace their drives. I already had a 4TB SSD that I planned to use for storing photos, so that part of the setup was sorted.</p><p>The Jonsbo N2, however, had one flaw: its fan was terrible. I knew this before buying it but also knew there was a fix. In a <a href="https://www.youtube.com/watch?v=R01Wq9vUZjU">fantastic YouTube video</a>, someone explained how to replace it with a Noctua NF-P12 fan, attaching it to the Jonsbo case using a mod: a small <a href="https://www.printables.com/model/526778-jonsbo-n2-extended-fan-module?lang=es">3D-printed accessory</a> to hold it securely in place. The video showcased the dramatic difference between the old fan’s hurricane-like noise and the near-silence of the replacement — a must for me.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1Bh1elNItgncaIEJXl6djw.jpeg" /></figure><p>I don’t own a 3D printer, but luckily, services like <a href="https://imprimakers.com/es/">Imprimakers</a> exist in Spain (there are lots of options everywhere). They’ll print any design you send them for a reasonable price (and with various material options). Two weeks later, I had the accessory in hand, and my Unraid machine was fully configured just the way I wanted.</p><p><strong>Choosing the Platform</strong></p><p>Next, I had to decide how to manage everything. Xpenology was an option, but I wanted to try something new. There are several platforms for setting up a multi-purpose home server, with the most popular being Unraid, TrueNAS (formerly FreeNAS), Proxmox, OpenMediaVault, and CasaOS. There are others, of course, but I narrowed my options to these and ultimately chose Unraid. I haven’t tested the others, so I can’t speak to their merits, but I liked Unraid — here’s a <a href="https://forums.unraid.net/topic/46802-faq-for-unraid-v6/">comprehensive FAQ</a> — for its active community and long-standing reputation. It’s a paid platform (the Starter license, which I purchased for <a href="https://unraid.net/pricing">$49</a>, allows up to six drives), but it had excellent reviews and inspired confidence.</p><p>Building the hardware was easy, but getting it up and running was a bit trickier. Unraid isn’t like Linux or Windows — it has its own learning curve. Thanks to the forums (which are great, thanks to responsive moderators), I was able to resolve my questions and set everything up over the following days.</p><p>Unraid uses storage pools where, from the user’s perspective, everything appears as a single, large file system. I opted for ZFS, both for the SSD (used separately) and for the “pool” with the three hard drives in <a href="https://raidz-calculator.com/raidz-types-reference.aspx">RAIDZ1</a>. This configuration uses two drives for storage and one for fault tolerance. If one drive fails, your data is safe, offering peace of mind — though Unraid developers make very clear that you should always have backups, as this platform isn’t designed to replace them.</p><p>Then began the long process of transferring all my data to the new NAS. My Synology DS-212j was still in use, but it was time to retire it. I took the opportunity to not only transfer the data to the Unraid machine but also to reorganize and delete some files that had become messy over the years. Among these were, of course, several photos and thousands of videos scattered across folders from past projects. This task took quite a bit of time — a couple of months, working on it here and there. I still have old hard drives stored in boxes that I plan to eventually dust off and migrate to the Unraid NAS. But for now, the main data transfer is complete.</p><p>Once the migration was done, I finally did something I’d been wanting to do for a long time: retire the Synology. I sold the NAS and its two 2TB drives (after securely wiping them — I wrote about the process <a href="https://www.xataka.com/componentes/mi-viejo-disco-duro-no-suficiente-para-mi-asi-borre-forma-segura-sus-datos-para-revenderlo-miedo">on Xataka</a>). This helped recover part of the investment I made in the new case, hard drives, and other components.</p><p><strong>Exploring Unraid’s App Store</strong></p><p>Now the fun part could begin. One of the best features of Unraid is its vast library of “community applications,” which are essentially preconfigured Docker containers. You can browse the “store,” pick what you want, and install it with a few clicks (well, almost). Think of it as an App Store for self-hosted apps, letting you set up a server for just about anything.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3wnVdsSeFPpGR5vAx9HtoA.jpeg" /><figcaption>Let’s play. THis is Unraid’s “App Store”.</figcaption></figure><p>The possibilities are endless, and what’s great about Unraid is how effectively it leverages Docker. It’s like having 30 or 100 mini-servers in one. Each container operates as an independent server while sharing your hardware resources. Even with my modest motherboard and CPU, the performance was excellent as long as I didn’t push it too far.</p><p>At first, I got carried away, installing tons of things I hardly use now. For example, I tried setting up a Jellyfin server (a popular alternative to Plex, which I also tested but didn’t love). While Jellyfin has some nice features, the performance — partly due to my hardware — wasn’t ideal. I ended up sticking with Kodi on the Chromecasts around my house, pulling content from this NAS. For me, it’s still primarily a file server, like my old Synology NAS. But the real task was just beginning: finding the Google Photos alternative that would finally win me over.</p><h3>A World of Possibilities</h3><p>During this time, I already had a rough idea of which self-hosted alternatives to Google Photos I wanted to try. There’s a wide variety of projects in this space, and I looked into some interesting comparisons, such as the following ones:</p><ul><li><a href="https://rair.dev/google-photos-replacement-roundup/">Rair-dev</a></li><li><a href="https://rair.dev/google-photos-replacement-roundup/">MeichThys</a></li><li><a href="https://docs.librephotos.com/docs/user-guide/features/">LibrePhotos</a> wiki</li><li><a href="https://arstechnica.com/gadgets/2021/06/the-big-alternatives-to-google-photos-showdown/">Ars Technica</a></li></ul><p>Some, like the Ars Technica comparison, are a bit outdated but still relevant in general terms, as there haven’t been significant new entrants in recent years. So, I narrowed my options to a small group of strong candidates to become my self-hosted Google Photos replacement.</p><p>I quickly ruled some out after reading about them on their websites or testing their demos (if available). I tried others on my Unraid machine to see how they performed and whether one could be “the one.” Over the past few weeks/months, I’ve conducted these tests bit by bit, and here’s the result: a brief comparison of what I found most notable about these options.</p><h3>Piwigo</h3><p><a href="https://piwigo.org/">Piwigo</a> (<a href="https://github.com/Piwigo">GitHub</a>) is more ambitious and established, making it a potential choice for massive photo collections. It offers <a href="https://piwigo.org/demo">various demos</a> and boasts numerous features — batch editing, albums, calendar-based organization, themes, geolocation, and multi-user support. I considered giving it a serious try.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6CwQGYH5ljvK5DxrZXh0og.jpeg" /></figure><p>However, I was looking for a functional Google Photos alternative, which meant chronological organization was a must. Piwigo doesn’t focus on this (at least as I understood it). Its organization is more tag-based, with date information taking a backseat. It feels more like a super-gallery than the kind of photo manager I was seeking.</p><h3>Photonix</h3><p><a href="https://photonix.org/">Photonix</a> (<a href="https://github.com/photonixapp/photonix">GitHub</a>) is too basic, with limited features and no interface focused on chronological order. It uses machine learning to analyze objects, colors, and faces, which is intriguing.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vekH2mnVoAXkJ4gisxSiPw.jpeg" /></figure><p>However, it seems more like an experimental project with limitations. It could be useful as a starting point for other developments but is far from being a Google Photos replacement.</p><h3>Lychee</h3><p><a href="https://lycheeorg.github.io/">Lychee</a> (<a href="https://github.com/LycheeOrg/Lychee">GitHub</a>) is another project similar to Photonix, offering minimal features. It’s suitable only for viewing your photos in a gallery format.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QNmEaIWh5U11Bp5EmUwX3g.jpeg" /></figure><p>It lacks chronological organization — everything is album-based — and doesn’t include basic functionality like marking photos as favorites. While it might be interesting for small projects, it’s not a viable Google Photos alternative for me.</p><h3>Photoview</h3><p><a href="https://photoview.github.io/">Photoview</a> (<a href="https://github.com/photoview/photoview">GitHub</a>) has a somewhat simplistic interface — perhaps too minimalist — but it does adopt Google Photos’ ideas for chronological arrangement.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qFAUH6yn_MisNeO_WOclSw.jpeg" /></figure><p>It allows you to mark favorites and view photo information, but only from the thumbnail view. In slideshow mode, you can only see the photos themselves, which feels odd. It supports geolocation, has a mobile app (a definite plus), and features facial recognition, but the project is still in its early stages, with the creator <a href="https://photoview.github.io/en/docs/usage-basic-overview/">inviting</a> community contributions.</p><h3>LibrePhotos</h3><p>LibrePhotos (<a href="#">GitHub</a>) was one of the projects I tested on my Unraid machine and initially seemed the most promising. I liked its interface, which borrows ideas from Google Photos and Apple Photos, focusing heavily on chronological order. For example, the scroll bar that appears when hovering on the right side is fantastic for quickly navigating your photo timeline.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*a6LdEMdHz9E54083vVLE0Q.jpeg" /></figure><p>However, things went wrong for me. Importing my photo library took forever. While imports were slow on PhotoPrism or PhotoStructure too, LibrePhotos took at least twice as long. Additionally, the interface was consistently sluggish.</p><p>Installing it on Unraid wasn’t straightforward either. It’s not available as a container in Unraid’s “App Store,” so the process was a bit manual. This also means future updates have to be managed outside of Unraid’s interface, which usually makes everything simple but didn’t help here. It’s inconvenient.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*91VqX1NLa06Ja6fr4L53rA.png" /></figure><p>The project <a href="https://demo2.librephotos.com/">demo</a> runs flawlessly, and I was excited, thinking I’d found the perfect solution. But in practice, things didn’t go well for me. Even logging into LibrePhotos was a struggle, requiring significant effort. Clearly, I did something wrong because, on paper, LibrePhotos seems like one of the most promising alternatives to Google Photos.</p><h3><strong>Nextcloud Photos</strong></h3><p>I’ve been using Nextcloud on a VPS for years to experiment and store things I don’t want in, say, Google Docs. So trying out Nextcloud Photos (<a href="https://nextcloud.com/blog/announcing-nextcloud-hub-3-brand-new-design-and-photos-2-0-with-editor-and-ai/">web</a>, <a href="https://github.com/nextcloud/photos">GitHub</a>) seemed like a natural step.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZDap07djkYM1g2WR4tpH9Q.jpeg" /></figure><p>The interface is decent, with chronological organization, categorization of photos and videos, the ability to mark them as favorites, and even an “On This Day” feature showing photos from the same day years ago.</p><p>I only tested it briefly since it wasn’t my top candidate. While researching, I found there are additional plugins like <a href="https://docs.nextcloud.com/server/latest/admin_manual/installation/server_tuning.html#previews">Imaginary</a> and <a href="https://github.com/rullzer/previewgenerator">Preview Generator</a> that improve web preview performance, which isn’t particularly fast by default. There’s also a plugin called <a href="https://github.com/nextcloud/photos">Recognize for NextCloud</a> that enables AI-based facial recognition — a feature that sounds intriguing.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hYC6UpuN-xllWsbt4pG45w.jpeg" /></figure><p>However, marking photos as favorites is cumbersome. You must open a side panel, which also grants access to comments and sharing options. This feels counterintuitive. Marking a favorite should be quick — like a keyboard shortcut (e.g., “F”) or a simple mouse click on a heart or star icon.</p><p>Interestingly, there’s a parallel project to Nextcloud Photos called <a href="https://memories.gallery/">Memories</a>, a Nextcloud app. I discovered it recently and didn’t install it, but its demo shows it fixes some Nextcloud Photos shortcomings, such as easier favorite marking. It also integrates a metadata editor (ideal for fixing incorrect photo dates) and a robust information viewer with geolocation support. If you’re using Nextcloud, I’d recommend giving it a try.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*L1_jS8WJteX3lEsklOi1EQ.jpeg" /></figure><p>Nextcloud Photos does have one standout feature: one of the best integrated editors I’ve seen in tools like this. It offers a good selection of filters, drawing and text-adding tools, and adjustments for brightness, contrast, and warmth. This was a pleasant surprise.</p><p>Nextcloud has another advantage: a mobile app with perfect photo syncing from your phone to the server. As a private cloud backup solution, it excels. However, as a photo management tool, its interface doesn’t quite win me over. While it might work well for many people, it’s not “pretty enough” for my taste.</p><h3>Plex Photos</h3><p>In September 2024, Plex <a href="https://www.plex.tv/blog/plex-pro-week-24-introducing-plex-photos-beta/">announced</a> Plex Photos, an app for iOS, Android, and web that allows you to store photos on your Plex Media Server.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zjPiYorhwan-LPD_jTEm8w.jpeg" /></figure><p>Currently in beta, it’s an interesting option if you use Plex intensively. It offers three organization modes (Recommended, Timeline, and Library), and while you can mark photos as favorites, the process feels a bit clunky.</p><p>The web version (not Android) includes a “Random” button in Timeline and Library modes, launching a slideshow of random photos — a quirky feature. Since the app also supports music playback, you can even create photo playlists, which is unconventional but intriguing.</p><p>After experimenting with the app, I found it designed primarily for slideshow-style photo viewing. It has neat features, like displaying a filmstrip of photos below the main image in view. Marking favorites can be done from the thumbnail view, which is a nice touch. Overall, it seems promising and likely to improve over time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xTIede7zW96VABmhHGy7ag.jpeg" /></figure><p>Plex Photos benefits from Plex’s native apps for Android and iOS, which make mobile content access seamless. However, it lacks a native Android TV version — a limitation shared by many platforms.</p><p>I tried sideloading the Android APK onto my Chromecast with Google TV, but it was a disaster. It barely worked, and navigating the interface was a pain. For now, it’s not suited for this kind of setup. Even so, it’s a fun option. While I’m not a heavy Plex user despite having it installed on my Unraid server, I’ll keep an eye on its development.</p><h3>Photoprism</h3><p><a href="https://photoprism.app/">Photoprism</a> (<a href="https://github.com/photoprism/photoprism">GitHub</a>) is a serious Google Photos alternative. I tested it thoroughly by installing the Docker container on my Unraid machine and uploading my entire photo library to explore its capabilities. The range of features is impressive, with numerous options that stand out.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*RhLN1EffdUAL-aS0LR5idQ.jpeg" /></figure><p>Oddly, there’s no integrated editor for basic functions like cropping, rotating, or tone adjustment. However, it includes a detailed (and fantastic) EXIF editor, letting you adjust tags, identify people, modify additional properties, and even rate photo quality.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ghZwIQn2BnRgTuxSoteFpg.jpeg" /></figure><p>A critical note: Photoprism seems overly focused on organization, tagging, and classification rather than on viewing and enjoying photos. For example, you can mark a photo as a favorite, but there’s also a quality rating system with star rankings. While it’s great to have options, this can feel excessive.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*z1gmKbcjPuYmWAUdvu7Szw.jpeg" /></figure><p>One major drawback for me is the interface. Although it <a href="https://www.photoprism.app/pro/kb/customization">supports themes and customization</a>, this requires extra effort. Even with customization, I found the interface overwhelming. Displaying so much information by default on each photo feels unnecessary — I just want to see the thumbnail. Yes, there’s a “grid view” that omits this information, but it feels too puzzle-like for my liking.</p><p>Additionally, the lack of a true “timeline” view, like Google Photos, is disappointing. The calendar view merely groups photos by month, which doesn’t appeal to me.</p><p>Photoprism is exceptional if you want your photos meticulously organized and tagged. But if your goal is to enjoy your photos, its less attractive, more utilitarian interface might not be the best fit. Personally, it feels like using Photoprism is more about managing than enjoying photos — a very subjective impression, but one I can’t shake.</p><h3>PhotoStructure</h3><p>I’m a big fan of <a href="https://photostructure.com/">PhotoStructure</a> (<a href="https://github.com/photostructure">GitHub</a>), mainly because it’s a unique and distinct alternative to Google Photos. While chronological organization is important, what initially caught my attention was the random photo grid displayed as the default view — a fun and refreshing way to explore your photo library.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BgHnI5nvak5Cy0dZcetetQ.jpeg" /></figure><p>PhotoStructure has long been a strong contender as my primary photo management platform. <a href="https://javipas.com/2021/04/26/probando-photostructure-candidato-a-ser-ese-google-photos-que-yo-controlo/">I wrote about it back in 2021</a>, and even then, its speed impressed me — both on PC and mobile devices (via a web browser, as there are no native apps). However, the lack of apps isn’t a big deal for me.</p><p>That said, its chronological organization feels a bit odd. The focus on random grids shapes the entire experience. There’s no straightforward day-by-day organization, from newest to oldest, which is what I prefer. Instead, you navigate by selecting a year and then a month. While the monthly grids are ordered, they feel cluttered and lack breathing room. The interface can feel cramped.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8gcJSELER6tBjwIofybt5w.jpeg" /><figcaption>If you want to mark a photo as a favourite, you must pay, as shown in the popup on the lower right corner of the screen.</figcaption></figure><p>The free version of PhotoStructure offers basic functionality, but <a href="https://photostructure.com/about/pricing/">the Plus version</a>, priced at $5/month or $42/year, adds features like facial recognition, automatic organization, and the ability to mark favorites — arguably essential features. While the subscription cost is reasonable, I still see restricting favorites as a low blow.</p><p>These quirks, combined with a somewhat cramped interface, made me consider other options. I still have PhotoStructure set up and occasionally use it. The effort behind it is commendable, and the developer, Matthew, is deeply invested in the project. He’s very responsive on the <a href="https://forum.photostructure.com/">support forums</a> and open to suggestions, which is great to see. However, it’s not the perfect solution for me.</p><h3>Ente</h3><p>My latest discovery in this space is <em>Ente</em> (<a href="https://ente.io/">web</a>, <a href="https://github.com/ente-io/ente">GitHub</a>), a project that went open-source <a href="https://ente.io/blog/open-sourcing-our-server/">in March 2024</a> after being a paid-only service for a long time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*C1J9ukIrsce3N4mK_fgW_g.jpeg" /></figure><p><em>Ente</em> is impressive. There’s a hosted version with a free plan offering 5 GB of storage, along with various paid plans depending on your storage needs. Alternatively, you can self-host it using the downloadable code, as detailed <a href="https://help.ente.io/self-hosting/">in their documentation</a>. Some users are even <a href="https://github.com/ente-io/ente/discussions/1183">working on Dockerfiles</a> to <a href="https://github.com/ente-io/ente/blob/main/server/docs/docker.md">simplify installation</a> via Docker, though I haven’t tried that option myself.</p><p>I tested the macOS desktop version, which functions similarly to Apple Photos. You can import folders and individual photos, and the app organizes them chronologically. The interface is fantastic — clean, elegant, and intuitive, with the expected view where newer photos appear at the top and older ones at the bottom. However, the desktop app currently only supports dark mode. While a light theme is promised, it’s <a href="https://ente.io/blog/redesign-and-light-mode/">currently exclusive</a> to the mobile apps.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oA3YJBpCOs-8AcWtsqERbg.jpeg" /></figure><p>The tool offers standard features like marking photos as favorites, viewing EXIF metadata, and geolocation (using OpenStreetMap, which is a nice touch). It also includes a basic editor for cropping, brightness, contrast, and saturation adjustments.</p><p>What truly sets Ente apart is its ambitious focus on encryption. All photos and metadata are end-to-end encrypted, ensuring only the user can access them. The Ente team proudly <a href="https://ente.io/architecture/">highlights their encryption architecture</a>, which indeed seems robust. This emphasis on privacy and security is especially appealing. Beyond photos, Ente offers other apps like <a href="https://ente.io/auth/">Auth</a> — a Google Authenticator-like tool that can also be self-hosted. It’s an intriguing ecosystem.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*t-ripWfL0f4I7pYYd2poWQ.jpeg" /></figure><p>The Photos app also includes <a href="https://ente.io/features">additional features</a> like privately sharing albums, allowing others to contribute photos (even non-Ente users), sharing photos via private links, and more. However, it currently lacks facial or object recognition capabilities. The search functionality is fairly basic, focusing on file and folder names.</p><p>Overall, Ente stands out for its elegant interface, strong privacy features, and thoughtful design. While its search and recognition tools are limited, it’s a solid choice, especially for users prioritizing security and a clean user experience.</p><h3>Other Options</h3><p>In this general overview, I also wanted to mention a few other projects. While they didn’t meet my main priorities, they might be useful for specific scenarios:</p><p><strong>Lychee</strong> (<a href="https://lychee.electerious.com/">web</a>, <a href="https://github.com/LycheeOrg/Lychee">GitHub</a>, <a href="https://lychee-demo.fly.dev/landing">demo</a>): Lychee is charming but overly simplistic. It’s essentially just a photo viewer without much else going on. Its interface, as seen in the demo, is neat and lets you group photos in various formats. However, it’s very limited in features. For example, I couldn’t find options for creating albums or marking favorites. It also lacks a calendar view. This makes it an interesting choice for small-scale projects — like organizing photos from a trip — but not much more than that.</p><p><strong>Chevereto</strong> (<a href="https://chevereto.com/">web</a>): If you’re a fan of <a href="https://imgur.com/">Imgur</a> and dream of setting up your own version, Chevereto might be right for you. It’s designed to create a gallery where chronological order isn’t the focus — instead, it emphasizes user interaction, such as votes and views (more “likes” equals more visibility). While it’s a valid option, it’s just not my cup of tea.</p><p><strong>Photo-stream</strong> (<a href="https://github.com/waschinski/photo-stream">web</a>): This project is heavily geared toward creating visually appealing galleries, though it doesn’t go beyond that. The resulting interface is sleek and seems to perform smoothly.</p><p><strong>Home-Gallery</strong> (<a href="https://github.com/xemle/home-gallery">web</a>): Another gallery-focused project that at least offers yearly classification. However, it’s still limited for the more comprehensive features I was looking for.</p><p><strong>Photofield</strong> (<a href="https://github.com/SmilyOrg/photofield">web</a>): Then there are absolutely unique projects like Photofield, which don’t suit my needs but are undeniably fascinating. This one is almost an experiment. Its goal? To display as many photos as possible on-screen by zooming out. The result, at least based on the demo, is stunning.</p><p><strong>Damselfly</strong> (<a href="https://github.com/Webreaper/Damselfly">web</a>): Its interface, inspired by the legendary Picasa, seemed like a bit of a nightmare to me, so I didn’t even try it out. However, it does look intriguing. It features facial, object, and color recognition, making it highly focused on finding photos with various search parameters.</p><p><strong>Snapcrescent</strong> (<a href="https://github.com/snapcrescent/snapcrescent">web</a>): This one feels like a fork of Immich with a strong mobile orientation. I haven’t tried it, and its demo isn’t working either, but it looks interesting if you want to check it out.</p><h3>And the Winner Is… Immich</h3><p>If I recall correctly, I discovered <a href="https://immich.app/">Immich</a> over a year ago while still using Xpenology and relying on Synology Photos, which seemed like an almost perfect solution at the time. Immich is a relatively new project — its first public release, version v0.2-dev, <a href="https://github.com/immich-app/immich/releases/tag/v0.2-dev">appeared</a> on GitHub in February 2022. However, its <a href="https://immich.app/roadmap/">evolution</a> has been rapid — almost <em>too</em> rapid, with new version notifications popping up constantly. By the time I found out about it, Immich was already positioning itself as a full-fledged alternative to Google Photos, with that goal clearly at its core: to be <em>the</em> alternative to Google Photos.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OTzuqWKs10xXdhSH9_O3qg.jpeg" /></figure><p>The inspiration from Google’s platform is unmistakable, most notably in the interface, which is essentially a shameless (and excellent) copy of what Google Photos offers. Immich is available in several formats, but I took advantage of its inclusion in the Unraid “app store” to install it on my server. I followed the advice in these videos (<a href="https://www.youtube.com/watch?v=krAZcUIP3YE">Part 1</a>, <a href="https://www.youtube.com/watch?v=LtNWxxM5Mzg%C2%A1">Part 2</a>) from Spaceinvader One, a popular channel among Unraid users.</p><p>At first, I got a bit confused configuring the paths to my photo library. Immich has a peculiarity that could trip up some users: it creates its own photo library from your existing photos. You might already have a directory with thousands of folders organized your way, but when you import them, Immich builds its own directory structure under /photos. I initially thought this was where my photos were stored—turns out it’s a separate directory where Immich generates folders and copies of your photos with unrecognizable filenames.</p><p>This means that if your original library is 2 TB, you’ll need another 2 TB for Immich’s library. However, there’s an upside: Immich doesn’t touch your original photos, which remain “safe” from any changes you make in Immich. While this could be a drawback (e.g., if you delete duplicates or bad photos in Immich, those changes won’t sync to your original library), I see it as more of a benefit than a drawback.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*TGKkZeQ1N418V0T0UzcZBw.jpeg" /></figure><p>It’s worth noting that Immich now offers an option to work with your existing photo library as an <strong>external library</strong>, bypassing the duplication. I found this approach a bit daunting and risky, so I stuck with the default method, especially since my NAS had plenty of storage space.</p><p><strong>Importing Photos: Immich’s Secret Weapon</strong></p><p>Immich comes equipped with powerful photo-import tools. I used <a href="https://github.com/simulot/immich-go">Immich-Go</a>, which I found sensational, though there’s also the <a href="https://immich.app/docs/features/command-line-interface/">Immich CLI</a>. Immich-Go can import standard libraries while maintaining the folder structure from your original library. Even better, it can import photos from Google Photos. While this wasn’t my case, many users download their Google Photos library via <a href="https://support.google.com/accounts/answer/9666875?hl=es">Google Takeout</a> and then use Immich-Go to import it. The tool supports various parameters, and while it might seem daunting at first, it works wonderfully for migrating from Google Photos to Immich.</p><p><strong>Machine Learning Options</strong></p><p>Immich also includes facial recognition and other machine learning (ML) features that can be enabled during import. However, many recommend avoiding these options unless you have a high-performance machine with a powerful GPU. My modest Unraid setup wasn’t up to the task, so I skipped it. That said, Immich offers an incredible feature: <a href="https://immich.app/docs/guides/remote-machine-learning/">Remote Machine Learning</a>, which allows you to offload ML tasks to another device on your local network. For a “relatively small” application, this is mind-blowing.</p><p>Immich boasts <a href="https://immich.app/docs/overview/introduction">exceptional documentation</a> and an active user community on GitHub, <a href="https://discord.com/invite/cHD2af9DbA">Discord</a>, and <a href="https://www.reddit.com/r/immich/">Reddit</a>. This not only helps iron out bugs but also ensures the platform keeps improving. The community has even created <a href="https://immich.app/docs/community-projects/">fascinating add-ons</a> like <a href="https://github.com/damongolding/immich-kiosk">Immich Kiosk</a>, an app for displaying your photos in kiosk mode on devices or browsers.</p><p>While Immich-Go is excellent, it’s not perfect. I encountered issues with the metadata of some photos, causing them to appear out of chronological order. This was rare, though — 95% of the time, it worked flawlessly. (In my case, the problems likely stemmed from the size of my library and poor metadata in older photos.) Even so, I imported a massive library: 140,000 photos and 19,000 videos, totaling 1.6 TB.</p><p><strong>A Highly Visual, Chronological Focus</strong></p><p>Immich’s chronological organization is superb. The photo mosaic adapts dynamically to portrait or landscape formats, making it visually appealing and easy to browse. A calendar sidebar allows you to jump to specific dates quickly, making navigation nearly unbeatable.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*VDdwIp1u9eMmJsk9MxAfPg.jpeg" /></figure><p>The photo viewer includes shortcuts for sharing, zooming, and viewing photo information, but what sets it apart are two rare features: mark as favorite and move to trash. Even better, these can be done with simple keyboard shortcuts: press <strong>F</strong> to favorite a photo or <strong>Delete</strong> to send it to the trash.</p><p><strong>A Nostalgic and Thoughtful Home Screen</strong></p><p>Immich’s home screen displays photos in reverse chronological order, with a bonus section at the top: <strong>“X Years Ago”</strong>, showcasing memories from the same day in past years. While nostalgic and fun, this feature can be bittersweet — thankfully, it can be disabled in the app’s preferences.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*fWWyd3NmFQqeOL9oCL5JTA.jpeg" /></figure><p>Immich supports multiple users, custom CSS designs, and server configuration for <a href="https://immich.app/docs/guides/remote-access/">remote access</a>. You can even set it up to work like Google Photos, accessible from anywhere. While I prefer local access for security reasons, I’m considering <a href="https://ashpex.net/2024/07/my-journey-to-self-hosting-immich-and-raspberry-pi-for-personal-photos-and-videos-storage/">alternatives like Tailscale</a>, which reportedly offers secure and easy remote access.</p><p><strong>Smart Search and Machine Learning Magic</strong></p><p>Immich’s search function is phenomenal. Beyond filename searches, it supports semantic search — type “mountains,” “ocean,” or even a color like “orange,” and it finds relevant photos. This feature is a standout.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7MkYnj_6IiDuDkfD2YScOg.jpeg" /></figure><p>The facial recognition system is also effective. I don’t use it often, but it’s been invaluable when I’ve needed to locate photos of a specific person.</p><p><strong>Seamless Mobile Integration</strong></p><p>Adding new photos is effortless, whether through the web interface or Immich’s mobile apps for Android and iOS. Automatic uploads are easy to configure, and as soon as I return home, my server updates within minutes (or longer if I’ve been on vacation with hundreds of new photos).</p><p>Immich is a stunningly capable platform, full of thoughtful details and advanced features. While it’s not perfect, its strengths far outweigh its weaknesses, making it a phenomenal alternative to Google Photos.</p><h3>It’s 2025, and There’s Still No Good Way to Enjoy Photos on TV</h3><p>Immich’s mobile apps are quite decent, but unfortunately, here lies Immich’s one major flaw: there’s no proper native app for Android TV.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bEkvkVDIigBVExydeib6TQ.jpeg" /><figcaption>As usual, the Android TV version of Immich has a very, very limited interface. For example, navigating to view photos from 5 or 6 years ago is torture — there’s no timeline feature to let you scroll through dates easily, as was possible with Synology Photos. <em>Argh.</em></figcaption></figure><p>There is at least <a href="https://github.com/giejay/Immich-Android-TV">one third-party solution</a>, but when I tried it a while back, it didn’t work well, and it hasn’t been updated in three months. The remaining options, as in similar cases, are:</p><ol><li>Use a web browser, or</li><li>Mirror your phone screen and use it as a remote while displaying the photos on your TV.</li></ol><p>The latter option doesn’t work well for me because the resolution and quality of the photos drop significantly. In my opinion, the browser is a much better alternative. At home, we actually revisit old photos of the kids when they were little by directly connecting a laptop to the TV. It’s not “elegant,” but it works and lets us enjoy the photo library as a family.</p><p>The lack of native Android TV clients for Immich and other apps is undoubtedly the major downside in this segment. Being able to manage your photo library comfortably from your TV would be fantastic. Navigating through a timeline is rarely easy or even possible (as I mentioned with Synology Photos), and Immich faces the same issue. Additionally, I’d love to assign specific buttons on the Chromecast remote (in my case) to mark a photo as a favorite or send it to the trash easily — maybe by holding one of the buttons for two seconds.</p><h3>Immich Is Still My Favorite Google Photos Alternative</h3><p>Aside from that limitation, Immich remains my favorite alternative to Google Photos. There are other strong contenders, I admit:</p><ul><li><strong>PhotoPrism</strong> is extremely powerful.</li><li><strong>Nextcloud Memories</strong> is intriguing.</li><li><strong>Ente</strong> also seems very promising if they can simplify the installation process.</li></ul><p>And, of course, there’s <strong>Synology Photos</strong>, but it requires buying a Synology NAS — expensive but solid devices — or setting up an Xpenology system, as I did. All in all, I hope this (lengthy) post helps you decide if you’re considering doing something similar!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0f5723719a7d" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>