<?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 Naré on Medium]]></title>
        <description><![CDATA[Stories by Naré on Medium]]></description>
        <link>https://medium.com/@foundress?source=rss-7d289ad87fc3------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/2*Ix0EHSieyhej32NWtmfVug.png</url>
            <title>Stories by Naré on Medium</title>
            <link>https://medium.com/@foundress?source=rss-7d289ad87fc3------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 08:29:06 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@foundress/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[Solving P=NP]]></title>
            <link>https://medium.com/@foundress/solving-p-np-28dae1ee6eb0?source=rss-7d289ad87fc3------2</link>
            <guid isPermaLink="false">https://medium.com/p/28dae1ee6eb0</guid>
            <category><![CDATA[fiction-writing]]></category>
            <category><![CDATA[fiction]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Naré]]></dc:creator>
            <pubDate>Tue, 17 Sep 2024 13:53:52 GMT</pubDate>
            <atom:updated>2024-09-17T13:55:48.641Z</atom:updated>
            <content:encoded><![CDATA[<p>This is not what you expect it to be. This is a fictional first-person essay about the man who solved optimization. It is written by OpenAI’s o1 preview model. No human edits. Turns out a model that is great at mathematics and reasoning, can produce beautiful fiction.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/832/1*fq5f3gf19fdlqF_DHGkj3A.png" /></figure><h3>The man behind the invention</h3><p>I woke up to the sound of rain tapping against the window — a somber sound that mirrored the rhythm of my thoughts. The world outside moved on, indifferent to the tempest within me. The coffee tasted bitter that morning, but perhaps it wasn’t the coffee.</p><p>I’d done it. Somewhere between midnight and dawn, between the last cigarette and the cold leftovers, I’d scribbled the final line of the proof. <strong>P=NP.</strong></p><p>Three symbols that could upend the very foundations we stand on. In another time, I might have celebrated with reckless abandon. But now, it felt like holding a lit match in a room full of dry paper.</p><p>My desk was a battlefield of crumpled notes and empty cups. A dead houseplant stood sentinel in the corner — a silent testament to neglect. It reminded me of myself, rooted yet withering, reaching for a light that seemed always just out of grasp.</p><p>Memories seeped in like rain through a leaky roof. My father, a mechanic with oil-stained hands, spent his days fixing the broken things others had given up on. He believed that with patience, anything could be mended. “But be careful,” he’d warn, eyes weary yet kind, “sometimes fixing one thing breaks another.” As a child, I didn’t understand. I thought he was talking about cars.</p><p>My mother was a different storm — passionate and unpredictable. She chased dreams like leaves in the wind, never settling long enough to watch them grow. She’d pull me into her world of fleeting adventures, but I always felt like an accessory rather than a participant. When she left for good, the silence she left behind was a vacuum that nothing can fill.</p><p>I found solace in numbers and equations — a language that didn’t falter or abandon. Mathematics was an anchor, a place where I could exert control. Yet, even there, I was searching for more — perhaps a proof that I wasn’t as alone as I felt.</p><p>Now, holding the proof in my hands, I was acutely aware of the irony. I’d sought answers to feel less alone, and instead, I’d uncovered a truth that could isolate me further. The implications of <strong>P=NP</strong> were staggering. The cryptic locks safeguarding global secrets would become trivial puzzles. The fragile trust underpinning economies and relationships could crumble like poorly built structures in an earthquake.</p><p>The rain blurred the world outside, turning lights into smudged constellations. I lit another cigarette, watching the smoke curl and dissipate — a transient existence, much like our own. Part of me yearned to let the proof vanish just as easily, to pretend that this burden wasn’t mine to carry.</p><p>I thought of Oppenheimer, the man who unlocked the power of the atom only to see it unleash unspeakable destruction. Did he feel this same knot in his stomach, this dread that he’d set something irreversible into motion?</p><p>I glanced at the mirror, catching a glimpse of eyes that seemed foreign — haunted yet resolute. Who was I? Just a solitary thread, perhaps, but one that could unravel the whole if pulled. The enormity of it all pressed down on me, a physical weight that made each breath a conscious effort.</p><p>Could I trust the world with this? More pointedly, could I trust the world not to destroy itself with it? The boundaries between right and wrong blurred.</p><p>I returned to the screen, the cursor blinking like a heartbeat. The proof was laid out in meticulous detail, each line a stepping stone across a chasm of uncertainty. I read through it again, not searching for errors but silently hoping for them. An error would be a lifeline back to normalcy.</p><p>But there were none.</p><p>Setting up an anonymous account, I felt like a thief in reverse — stealing from myself to give to the world. Fingers hovered over the keyboard, laden with hesitation. “To those who seek,” I typed. It was both a blessing and a relinquishment.</p><p>I pressed send.</p><p>A strange calm settled over me, like the eye of a storm. The deed was done, irreversible as time. The weight hadn’t lifted, but it had shifted — no longer a burden solely mine. I stepped away from the desk, the room feeling both claustrophobic and vast.</p><p>The rain had eased to a drizzle. I grabbed my coat and stepped outside. The air was crisp, carrying the scent of wet pavement and possibilities. For once, the barrier between me and the world felt thinner, permeable.</p><p>As I walked down the street, I observed the faces of strangers — each absorbed in their own narratives, unaware of the subtle tremor that had just been set beneath their feet. I wondered how long it would take for them to notice, if they ever would. Maybe the world wouldn’t change overnight. Maybe it would ripple slowly, like a stone cast into a still pond.</p><p>A child splashed in a puddle, laughter cutting through the gray. His mother pulled him along, scolding yet smiling. The simple scene stirred something within me — a flicker of connection, a reminder that life persisted in moments both grand and trivial.</p><p>Perhaps I’d unleashed a force that would bring both creation and destruction. Perhaps I’d merely added a drop to an ocean of human endeavor. Either way, the tides would turn as they always had.</p><p>Maybe I’d always feel like an outsider looking in, but in this moment, the glass seemed a little thinner. The rain had stopped, and the clouds were beginning to part, revealing slivers of blue sky — a quiet promise that storms don’t last.</p><p>As I made pushed a local bakery door open, I felt a semblance of peace. Not absolution, but acceptance. The world would keep going, and so would I, navigating certainty and doubt.</p><p>And for now, that was enough.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=28dae1ee6eb0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[On rejection: getting to a No]]></title>
            <link>https://medium.com/@foundress/on-rejection-getting-to-a-no-5169af7fc583?source=rss-7d289ad87fc3------2</link>
            <guid isPermaLink="false">https://medium.com/p/5169af7fc583</guid>
            <category><![CDATA[relationships]]></category>
            <category><![CDATA[rejection]]></category>
            <category><![CDATA[conflict-resolution]]></category>
            <dc:creator><![CDATA[Naré]]></dc:creator>
            <pubDate>Tue, 23 Jul 2024 15:22:33 GMT</pubDate>
            <atom:updated>2024-07-23T15:44:43.431Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*idrjKySgQXW8Q4zs8OQa3w.jpeg" /></figure><p>Everything in life that is worth doing comes with this skill of being able to handle rejection. We speak about it all the time, as a culture . We have built resilience to being rejected , often to a point where we keep pushing in areas we perhaps need to surrender.</p><p>We learn to be rejected at a young age. Every time we ask for something and are told we are not allowed to. First time we have a romantic interest, that does not reciprocate.</p><p>As we grow, we either become extremely good at handling it or we start to pick paths that will help us avoid rejection. This second path is a tricky one.</p><p>Even in long-term relationships, often we do not initiate conversations, talk about our desires or fears, as we are worried about being rejected. This is how most problems occur, and also how things break. Many a therapist and writer has covered this topic. Esther Perel perhaps is one of the most prominent ones.</p><p>Yet, there is the other side of rejection, the person, who rejects and how it is done which is less talked about.</p><blockquote>How do we reject as a culture? Where do we learn to reject and is there a right way of rejecting? Since it affects the other person so much, I believe this is worth a broader consideration.</blockquote><p>Have we thought about how we say <strong>no</strong> and the way we <strong>cut people off</strong> and <strong>turn them down and let them go </strong>? Let’s unpack.</p><p>There’s many situations in life, where we have to make the binary choice of not meeting someone’s expectations . If we become effective at rejecting we end up with a life, that has a lot of integrity and direction.</p><p>We also learn to detach our value to people from pleasing them . Often the best thing you can do to someone is close the door, let them go, tell them <strong>no</strong>. They won’t be pleased, they will have to process, but this does not mean you are hurting them or causing harm. Often it’s the contrary. You are saving them time.</p><p>Choosing to say no is brave , to the point, and in many cases the most respectful thing you can do. But culturally we are trained to draw circles , come up with excuses or keep doors open . The worst way of saying no is being dubious , cutting corners, lying or making the other person second guess whether it could have been a yes or whether they can turn it into a yes .</p><p>I have a toddler . When you parent a toddler saying no comes as second nature . You can sometimes forget how many times you say it . You rarely worry about how your toddler will feel , whether you will discourage and hurt their feelings and whether perhaps you should leave the door open. There’s right and wrong . You need to say <strong>no </strong>to keep them safe, out of trouble, contained .</p><p>I teach myself to say<strong> less no-s</strong> to my toddler , or to make sure all the no-s have a reason and he knows the reason and can learn to accept it. No-s are no-s, not maybe-s . We don’t negotiate. If it’s a maybe , we have a maybe from the start and try to create a yes from it . If you are picky with where you say no, they mean more.</p><p>If you grow up hearing a lot of no-s with no reason you learn to ignore them or constrain yourself to a point when you stop needing to hear them, but the space you operate in is tiny . Neither case works .</p><blockquote>You need a feeling of stability and security, when stepping out of your comfort zone , and a sense of getting a signal, when hearing a rejection . A rejection is a signal , it contains information, it is a filter you can use to your benefit.</blockquote><p>If you are a woman, who rejects a romantic prospect , you are not trying to make them feel less or unworthy. You are signalling that they should spend their time and emotions elsewhere . If you leave it open, you’re creating a dynamic of second guessing, questioning and finally hurt and dissatisfaction.</p><p>Culturally and the way us women are socialized, we often fear to reject and would rather go down much less clear paths to find a way out. This is very problematic for so many reasons. It leads to misinterpretations that can be very dangerous, both individually and societally. There is a whole group of males that have grown up with an idea that <strong>a no </strong>means they need to try harder and be better. Very sad and full of unintended consequences for everyone involved.</p><p>One of my friends recently said something profound .</p><blockquote>A No is a perfectly good answer.</blockquote><p>Whether it’s a VC, a date, a friend or anyone else . Being told <strong>no</strong> redirects your energy and time , reshuffles the deck.</p><p>How you say no still matters though.</p><p>Here are my rules of thumb for saying<strong> no. </strong>have been learning with my son and I started reusing these everywhere:</p><ol><li>If you say<strong> no </strong>and there’s a very simple logical explanation, you owe them one and you need to explain . If there is none, you have to accept there is none and step away, but be kind and upfront.</li><li><strong>A no is a no</strong> . It is non-negotiable. If it’s not a no, then it should be a maybe and you can get into negotiations and lay out the rules.</li><li>Be kind. When given a chance to make a call , do not make it about yourself only. Make sure you’re open and present and vulnerable for the other person not to be hit by a power dynamic , and to instead have trust in you.</li><li>Be quick . A no after a long time is worse than a quick no. Counterintuitive perhaps, because at least <strong>a no that took long </strong>means a thoughtful one, whereas <strong>an immediate no</strong> shows that you aren’t giving them a chance. If it’s a consideration it should be a maybe, otherwise quick no-s are the best for all sides</li><li>Don’t play. Adult games are only good for the bedroom, where your psyche gets to explore different roles with another person and it is mutually agreed. Elsewhere they are hurtful and never end well.</li></ol><p>We all think we’re amazing, until the very first time we’re rejected and we remember it forever. This is wrong.</p><blockquote>Then we realize that maybe it is not about us or the other person. Sometimes things align, other times they do not. There is a world of opportunity outside of what we want right now. What we want also changes.</blockquote><blockquote>Another lesson from my toddler.</blockquote><p>As a matter of fact, getting a no does not make us less amazing. It gives us input into where we should be going . A more efficient input than most of the yes-s we are going to encounter along the way .</p><p>There is a popular book on negotiation called “ Getting to a Yes”, listing the four principles to get to an agreement.</p><p>I think getting to<strong> a no is as important, if not more. Push for it, get them quickly, move on.</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5169af7fc583" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Who wins LLMs?]]></title>
            <link>https://medium.com/@foundress/who-wins-with-llms-684115178408?source=rss-7d289ad87fc3------2</link>
            <guid isPermaLink="false">https://medium.com/p/684115178408</guid>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[saas]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[investment]]></category>
            <dc:creator><![CDATA[Naré]]></dc:creator>
            <pubDate>Sun, 21 Apr 2024 22:00:40 GMT</pubDate>
            <atom:updated>2024-04-22T17:30:04.030Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*11NyExGOrT9jxwyM" /><figcaption>Photo by <a href="https://unsplash.com/@sebastiansvenson?utm_source=medium&amp;utm_medium=referral">Sebastian Svenson</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h4><strong>Prologue</strong></h4><p>The commodization of large models seems inevitable today and the cast of winners looks decided.</p><p>While there is no doubt that calling businesses who leverage LLM-s “wrapper companies” was too early, building a lasting advantage while delivering products and services in categories that are well served is hard, even if you are AI first.</p><p>Where does the value aggregate and how can you maintain strongholds as a business?</p><p>It is non-obvious where the AI first approach is going to be bear the fruits. While the first wave of these products like Tome or Jasper gained traction, retaining it is the real problem. If Notion and Google doc let me do “AI enabled editing”, do I need a Jasper or a grammarly.</p><p>There are skills that you needed to fine tune and optimize for that come out of the box with the next version of the model ? This race is impossible to win and is too expensive to even try.</p><h4>Being a mobile-first CRM during the mobile era did not make sense and Salesforce became the mobile-first one too. Turns out the mobile-first Facebook was Facebook. However, we saw the rise of delivery and car share businesses and mobile gaming.</h4><p>Also photo sharing made way more sense on mobile than on the web as our phones are on us and we have cameras in them to capture the moment. Hence instagram. Text messaging also became a retrospectively apparent behavior emergent from the mobile shift. Whatsapp and iMessage.</p><p>If it is faster for you to gain data and distribution as a new co, rather than for an incumbent to add “AI”, you have a shot.</p><p>If you can ship new code and build legacy systems from scratch more efficiently without adding patches to an existing system and resulting in 3–6 month delays, you have a shot.</p><p>Where and how does the value aggregate and how to capture it? A question many of us are pondering about. Below is a soup of my recent thoughts.</p><h4>Open source vs closed API-s</h4><p>With the Llama 3 release, we are clearly going to have both open source and closed source contenders to win, however certain big problems are still unsolved on both sides.</p><ul><li>Turns out AI works. However, we cannot afford it. Inference is still too expensive. Small models work on certain instances. Many use cases, particularly those including large volume reasoning such as data cleansing for instance, are still prohibitive due to cost.</li><li>Latency requires workarounds. Given the economics of the models, you are sacrificing speed and/or output accuracy and/or skyrocketing costs.</li><li>Getting consistent results across use cases is still a struggle. Alignment is being solved one model at a time, one company at a time .</li></ul><p>There are no lock-ins with bigger models yet and there is a clear LLM release fatigue. The reason for no lock-ins is the lack of adoption at scale.</p><p>Not enough production ready workloads are built on them and not enough derivatives depend on the model output. This is not sticky at all, compared to old school machine learning tasks, where scores and decision engines were built on internal models and training cycles that were very slow.</p><p>With LLMs there are too many alternatives, and most companies are investing in abstraction layers to be able to run evaluations with custom parameters across models and feed results into running use cases and upcoming builds. This means that the ability to switch between models and do the routing will be baked into the internal systems.</p><p>While initially, I believed the inference companies and clouds can step into this, the actual workflows might be too custom and enterprise dependent, hence the testing/eval layer as a service can end up making zero sense. We are yet to see. Companies like Distributional, TitanML, BentoML, Baseten, Martian, Together AI and others are interesting bets to follow.</p><h4><strong>Who are the winners then: Ecosystems</strong></h4><p>Calling the winners now is a bold task and the truth is I have absolutely no clue. But let’s pretend I do. Despite seeing no lock-ins within the model layer yet, it is obvious that cheap reasoning is going to be abundant, and ecosystems vs the technology are going to come out winning.</p><p>Also, the lower you are in the stack, the likelier you are to win. At least from today’s glance. NVIDIA et al, the clouds. They have ecosystems and a physical supply chain.</p><p>Now, if we look at the ecosystems out there, where do the players sit?</p><h4>Developer Ecosystems</h4><p>OpenAI has a developer ecosystem that relies on them, but does not like them very much. They have not necessarily been treated as the main customer. OpenAI favors the big guys and does customs deals for access and throughput + they favor startups invested in by the OpenAI fund. Not necessarily your average developer.</p><p>Instead, they are desperately attempting to create a consumer ecosystem with a very successful ChatGPT release and two failed attempts to move the advantage deeper, including the GPT store and Plugins. Being a consumer favorite is a very attractive and ambitious bug and Sam has got it.</p><p>One may argue, GPT-s were not that different from the main ChatGPT, hence developing agentic capabilities and actions further, they will get another try. My bet is that embedded software agents are more likely to be widely spread and popular vs a destination to look for an agent for X. Agentic capabilites will come in gradually to all the players and models. An agent for X vs an agent for Y needs to be vertical specific and context specific to be sticky. OpenAI does not have a clear path to get there.</p><p>As far as ecosystems go, Mistral is the new kid around the block everyone is loving , because they are new, they are small, they move fast and they are open source. However, since hundreds of millions being raised and having delivered subpar models so far, winning the ecosystem is going to be very hard for them. My guess is they will be acquired by a cloud player and take a turf being the European focused player. At some point, they need to start making money. Just like Inflection and Stability did.</p><p>Anthropic has the same challenges, as OpenAI plus the “second best” tag, despite Claude 3 Opus’ short term reigns and a cooler design. They keep up and will consume a portion of the value generated by these models in the upstream for sure. They are very friendly to developers, as far as closed API-s go, and need to double down here.</p><p>The pure enterprise play for Cohere seems to be harder than expected. There needs to be a real reason why anyone would leverage your own model GPT4 on Azure or Llama 3 coupled with Groq.</p><p>Groq is a very in vogue inference player with over 70K developers. A number they go to in 3 months vs NVIDIA’s 7 years, quoting the CEO. Different times though, we have to agree.</p><h4>The incumbent card</h4><p>The biggest contenders to win the ecosystem game are Apple and Meta, the cloud providers (AWS, GCP and Microsoft) and heir main businesses. Amazon, Google and Microsoft owns the global commerce, internet traffic, app stores and consumer devices respectively. Meta owns how we communicate and share information, both on the SMB and consumer side.</p><p>From the last cycle of the web and innovation, two things are fascinating to me. How could the clouds be a side business and yet so massive? I mean they are even bigger than anyone thought they would be. There is an unbounded TAM that is ever expanding. After all, Azure generates more revenue than the Windows /PC business.</p><p>The second phenomenon is the emergence of the data warehouse business. How did they even grow so big? They beat the cloud players at their own game.</p><p>The data warehouse businesses will stop making sense I think, as the cloud players will consolidate the whole stack, from storage to streaming, transformation to evals, observability and deployment. Databricks and Snowflakes of the world have real challenges coming their way.</p><p>Increasingly, if you are investing in horizontal LLM operations, good luck.</p><h4>Cloud ecosystems</h4><p>Every big tech company to play nice to you and try to convince companies to deliver their service container-ized via their own platforms. If you do not yet have a vibrant marketplace, like the cloud players do, very hard to win this. Picking a platform means picking distribution. Unless NVIDIA plays “hard to get” and uses its supply chain card. Likelihood is the market forces here are on the side of incumbents with large scale adoption, as ecosystems are not built in a day.</p><h4>Consumer ecosystems</h4><p>Apple, on the other hand, stands out by being late. This does not make it harder for them to win though. The consumer device game is theirs. I have a long standing belief that in the next 15 years at least we are not going to see large scale adoption of new devices, as much as I like the pins, the glasses and the rabbits. I think<strong> all you need is voice</strong> for the next layer of productivity to be unlocked. Its already delivered to you by your wired or bluetooth earphones, connected to your phone.</p><p>The task of building a very personal assistant on your phone, despite Siri being quite a few years old now, is still a hard one even with the magical technology we have today. Making a better Siri that is relevant to you is more complex than creating a compliance agent for the banks. The multiplicity in input adds to the probability of failure. Apple will take its time and catch up.</p><p>Meta AI seems to be going in this direction too. Crafting very personal assistants and agentic skills embedded with how you communicate online. There is no version of the future that is more predictable than this right now.</p><h3>Go get land : where do startups come in?</h3><p>This is not to say that startups and new companies cannot be building and kickstarting ecosystems. In fact there are going to be completely new ecosystems you would not bet on coming about.</p><h4>ERPs</h4><p>Similar to Rippling in the HR space, there are areas desperately waiting for someone to stop on the road, grab the 100B+ USD plus opportunity lying on the floor and do the hard work. <a href="https://www.seriesfi.com/">Series Financial</a> is at this, as we speak.</p><p>What does this mean in practice? It means that before you apply LLM-s in a given industry, there are opportunities across the stack that are obvious, yet untouched. You cannot build a new thing on old foundations falling apart.</p><p>One area that is interesting is legal tech. Instead of building another contract assistant for the space and trying to accelerate question answering, start from a centralized ERP for the industry. Where do the contracts live ? Where is the data stored and how is it connected to other pieces of data? Is it easy to retrieve , is it easy to connect? Is it standardized? <a href="http://corpora.us/">Corpora</a> is on this path with the vision of becoming a unified legal ERP. They started to go after the startup segment vs the legal firms. Whereas <a href="http://harvey.ai/">Harvey</a>, <a href="http://robin.ai/">Robin </a>and others are moving in a different direction.</p><h3><strong>The data cleansing layer and the ecosystem for data</strong></h3><p>Let’s dive into the next layer above compute and storage: data quality.</p><p>For decades, we’ve been searching for a solution to effectively cleanse and annotate data, and now, Large Language Models (LLMs) have finally unlocked this potential. Prior to LLMs, the data quality arena was dominated by sub-optimal solutions that relied heavily on human intervention, resulting in inadequate accuracy, slow turnaround times, and high costs.</p><p>As someone who has spent the last three years building Ntropy’s categorization business, I’ve experienced these challenges firsthand. Before the advent of LLMs, achieving satisfactory results on out-of-distribution samples was a near-impossible task. Our initial transformer had less than 100 million parameters, which limited its capabilities.</p><p>Interestingly, a valuable byproduct of data cleansing is the creation of a data network that unlocks hard-to-match efficiencies in terms of inference quality, speed, and cost. I’ll be discussing this further while wearing my Ntropy hat. In essence, if you can master data cleansing and annotation at scale, making it both inexpensive and fast, you’ll gain a competitive edge and long-term advantages in terms of both economics and performance.</p><h4><strong>Tooling: the pro-sumer space</strong></h4><p>Finally there is an opportunity in the tooling space across different problem verticals that is absolutely unique and overlooked. For instance 3D art and graphics generation. <a href="http://customuse.com/">Customuse</a> is a perfect example where over 3M avid creators will not go to ChatGPT or Claude for a new upgrade, but will be using the option made for their favorite platforms and all the bells and whistles it comes with. Customuse’s users are monetizing the assets generated, creating the dynamics Sam Altman wanted so much for the GPT store. Once you have filtering, curation and monetization, the ecosystem is hard to leave.</p><h4><strong>Recommendations</strong> <strong>and curation: the fuel of ecosystems</strong></h4><p>When the cost of creating content goes down massively and the barriers to entry are tiny, curation is everything. We saw this with the rise of social feeds, and algorithmic feeds and are just about to experience further levels of this with LLMs.</p><p>On the filtering side, there is a real opportunity in the browser space to embed content filtering and recommendations with LLM-s that have not been possible before. Once this story plays out, we will see the evolution of advertising where the option picked as the most personal to you gets a placement and attribution. After I have adopted Arc and see no way back, I think winning as a new browser with a sick filtering toolkit is a possibility still. It does not have to be an incumbent at all.</p><p>If I have ADD, my browser should be blocking all possible tabs once it detects I am in a flow state, and recommending the right music on Spotify, while at the same time filtering new content for me, and planning my next meeting according to my calendar. Why would a personal assistant live outside these premises, where I already spend my online life?</p><h4>The ecosystem opportunity in e-commerce</h4><p>Finally, ecommerce ecosystems are also ripe to see the benefits of LLMs. Amazon is not the sole everything store. In fact, I order more “stuff” on Doordash these days and there is a whole generation of people whose first credit and shopping experience has been a shiny new BNPL app.</p><p>Especially in regions where credit is not prominent and the new players are taking over, the leverage they have with merchants and directing consumer choice is massive. Take <a href="http://tamara.com/">Tamara </a>and <a href="http://tabby.ai/">Tabby.</a> Once you become the place for discovery and recommendations and there is an algorithm that can perfectly understand your intent, going elsewhere will be similar to living the “for you” feeds and Google search for Duck Duck Go. Dark place with things that do not matter to you. Not sure you want to hang there.</p><p>I think <strong>embedded search will take over generic search </strong>and Perplexity is amazing but has no chance of winning. This means you will search in your documents, within your assistants and apps and the recommendations will be so intuitive, just going on and generically looking for something will become a behavior of the past.</p><p>The <strong>everything app opportunity for commerce </strong>is still there and can 10–100x the existing businesses of current players, while significantly improving their margin game. Yet again, we need cheaper and faster inference.</p><h4>Who wins enterprise?</h4><blockquote>There are many companies building LLM tooling for the enterprise. The opportunity is obvious. This is new technology. There is so much to build. <br>Observability, output validation, prompting layers, connecting tools and API-s, inference, routing, cost and speed optimization, evaluation suites and more. Yet, the market is divided into two.</blockquote><blockquote>Companies, that know what they are doing and have huge teams to deliver this custom and in-house.</blockquote><blockquote>Companies, who do not have the know how, yet have a budget. These are a huge and expanding TAM, where you have to upgrade existing software and stacks with new capabilities or sometimes re-write it from scratch.</blockquote><p>However today, the go-to-market looks more like handholding and doing everything at once vs delivering platforms they can use themselves. That is okay. After all, VC-s have been long haunted by businesses looking like consulting due to lack of scalability. Now is the time in history, where you do not need to linearly add humans to scale a consulting business. In this moment in time, consulting enterprise customers is not a bad path at all. In fact, it can get repeatable and is inevitable.</p><p>The consulting business itself is upgrading into this and embracing LLMs.</p><h3>Epilogue</h3><p>In the next few years we will see winners I did not even think to mention here and you do not even think to ask about. That is what’s fun about the future though.</p><p>In the meantime, I do believe that The Rippling model of building a business has never been more valuable and more feasible. My prediction of <strong>stack as a service</strong> is here. Parker did not need LLM-s to pull this off, others will do.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=684115178408" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[With AI, the opportunity is in the shleps]]></title>
            <link>https://medium.com/@foundress/with-ai-the-opportunity-is-in-the-shleps-63a44015208a?source=rss-7d289ad87fc3------2</link>
            <guid isPermaLink="false">https://medium.com/p/63a44015208a</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[llmops]]></category>
            <dc:creator><![CDATA[Naré]]></dc:creator>
            <pubDate>Sun, 21 Jan 2024 12:40:45 GMT</pubDate>
            <atom:updated>2024-01-21T13:03:26.568Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QL80NhLmUvaJSMuyVXxzpQ.png" /><figcaption>Here is the image of the imaginative creature “Shleps,” a whimsical blend of a sheep and an octopus, set in a playful and magical grassy field.</figcaption></figure><p>Some time ago PG published the essay on <a href="https://paulgraham.com/schlep.html">“Shlep blindness”</a> and why there are many great ideas untouched, because hackers hate doing the tedious stuff, the work that does not involve <a href="https://paulgraham.com/schlep.html">writing code and pushing it to servers.</a></p><p>Many a hacker has overcome this now, balancing between doing things that do not scale and figuring out operational leverage, while cooking software businesses in any vertical imaginable. My very good friend Rishi has taken over fish farms globally and is <a href="https://www.observe.tech/">optimizing feed.</a> This is way beyond writing code and waiting. It includes sitting on boats in Chile and learning Spanish to impress the farmers and incorporate their feedback. Then flying to Scotland or Tasmania and doing the same.</p><p>With large models it seems that now more than ever, writing software, fixing bugs, writing unit tests, re-training models, most of the unpleasant stuff is becoming way easier or automated. Here we are cramming a decade of predictions around automation into one year, 2024, getting excited about automating every single workflow imaginable.</p><p>Yet, to get these models to work, <strong>there’s more shleps than you can think of.</strong> In every domain, for every task you have an assembly line of things you need to do that can seem very bespoke, that break all the time. This is not glorious at all.</p><p>Let’s dig into prompt engineering. Well, in order to engineer prompts, it is not just about engineering software. It is about a detailed understanding of the domain and the task you are trying to get done, the intricacies of the model and how it works, and then some communication and quirks on top.</p><p>Or let’s talk retrieval. Most enterprises are psyched about RAG. At least on paper or on X. But it involves a lot of work, including managing non-trivial ETL processes, model customization and finally “train-deploy-eval” cycles that are not plug-and-play today.</p><p>Then, if we talk about Vertical AI, meaning applying AI to specific domains and gaining better performance on those domain specific tasks and workflows, most of the opportunities lie in the shleps too.</p><p>Accounting automation. Sounds fantastic. The reality is to build a product that works, a model is the beginning or maybe the magic dust. You need the rest. Lost of work and domain specificity.</p><p>One of the things I am most excited about in terms of <strong>“tooling that needs to be built”</strong> is the tooling that will make it easier to extract knowledge from domain specialists. A chat interface is not it. But it’s the best we have got so far. Karpathy knows.</p><iframe src="https://cdn.embedly.com/widgets/media.html?type=text%2Fhtml&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;schema=twitter&amp;url=https%3A//twitter.com/karpathy/status/1748784260318990496%3Fs%3D20&amp;image=" width="500" height="281" frameborder="0" scrolling="no"><a href="https://medium.com/media/4441269c49b1d1a9b727112c9e9e3d30/href">https://medium.com/media/4441269c49b1d1a9b727112c9e9e3d30/href</a></iframe><p>The most exciting companies either enabled by AI or enabling AI are going to be built by those, who are ready to embrace the shleps.</p><p>P.S.</p><p>Above is a generate photo by ChatGPT. I asked it to generate an image of shleps.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=63a44015208a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to live off the edge of the cliff and make the most of it]]></title>
            <link>https://medium.com/@foundress/how-to-live-off-the-edge-of-the-cliff-and-make-the-most-of-it-59b76f670417?source=rss-7d289ad87fc3------2</link>
            <guid isPermaLink="false">https://medium.com/p/59b76f670417</guid>
            <category><![CDATA[entrepreneurship]]></category>
            <category><![CDATA[psychology]]></category>
            <category><![CDATA[builders-and-developers]]></category>
            <category><![CDATA[builders]]></category>
            <dc:creator><![CDATA[Naré]]></dc:creator>
            <pubDate>Sat, 13 Jan 2024 11:02:25 GMT</pubDate>
            <atom:updated>2024-01-13T11:13:05.305Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Kd1TMBxTnqw_iCSCctJeYA.png" /></figure><p>I hate talking about how being a founder is so unique and the challenges of it. Although I know it is.</p><p>Being a founder is a massive privilege.</p><p>A lot of the time it’s not one big leap of faith you have to take, though. There are lots of little ones and then some big ones along the way. This trains your optimism muscle like nothing else. But this also means living off the edge of the cliff most days, weeks, months.</p><p>So how to actually manage living off the edge off the cliff and not losing your mind?</p><blockquote>Off the edge of the cliff is where all possibility is. That’s a fact. Paths unknown, not walked by others is where you find most satisfaction, opportunity, creativity. Like the black parts of the map that turn green in Age of Empires , if you are old enough to remember the game.</blockquote><p>Below is my short list of tips on how to make living like this more bearable, more fun and in fact more productive.</p><ol><li>This is your choice. Daily reminders that you have chosen this route are quite practical. They are key. Maybe it is choosing the broader idea. Maybe it is the choice of a specific go-to-market motion or the people you work with or even your board and investors. Can be the market, the copy of your website. You can zoom in or zoom out infinitely. Once you do this, you become more intentional and being intentional makes taking jumps more controlled, hence they feel better.</li><li>Stay curious, but avoid the what ifs. Curiosity is why we are all here. When you stop being curious, you age. Not to be agist, but you come closer to an ending. However <strong>what ifs </strong>can be all-consuming and they are not productive. What if you did not hire a certain person or hired earlier, what if you chose a different feature set, what if you started fundraising earlier? Again you can zoom in and zoom out as you will. These are toxic. Instead, if a <strong>what if </strong>is bugging me a lot, I tend to design curiosity driven mini experiments that won’t take much resource but will get me my answers. If you are building, you are constantly looking for answers. Negative, positive. You need signal.</li><li>Seek negative signal. Seek no-s. Sounds counter-intuitive. Most people will tell you that you have to be ok with no-s, that you do not have to take them to heart. I would say you should proactively seek them. Because when you are making large bets you need to assign probabilities and count for uncertainty. Getting to <strong>no-s</strong> narrows down focus. Focus is what wins you the pieces of the pie and the whole pie in the end.</li><li>Assign probabilities and plan around them. This helps with emotion control and damage control when you are taking bets. Reward yourself for the right guesses and refine the art of measuring what is to come. In the end you become exceptionally good at figuring out what is to come and/or at making it happen.</li><li>Do not take yourself too seriously. This is a key part of the culture at my company, <a href="http://ntropy.com/">Ntropy.</a> Once you start to take yourself seriously, living off the edge becomes not feasible. What if you fall and lose everything? What would others think? What would you think? I think this is a known technique to fight fears, and I have been personally deploying it since I was very very young. Imagine the worst case scenarios. If you are like me you will picture even a couple of them, you can paint every detail, then get over it and move on. It is not that bad. You are not that important.</li></ol><p>Would love to hear how others make it off the edge and what you have learnt along the way.</p><p>Living like this is a lifelong commitment, with parameters that get slightly adjusted depending on where you are and what you are doing.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=59b76f670417" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[On Conviction]]></title>
            <link>https://medium.com/@foundress/on-conviction-6b08f7b0dd59?source=rss-7d289ad87fc3------2</link>
            <guid isPermaLink="false">https://medium.com/p/6b08f7b0dd59</guid>
            <category><![CDATA[personal]]></category>
            <dc:creator><![CDATA[Naré]]></dc:creator>
            <pubDate>Sun, 31 Dec 2023 18:47:36 GMT</pubDate>
            <atom:updated>2023-12-31T18:47:36.060Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5cswbNUpdXPC4nXkt_-ZTg.png" /><figcaption>Nothing is as it seems</figcaption></figure><p>This year I got to see someone very close to me, someone I love in a casket in a little church in LA. Somewhere he probably never even went. Somewhere far from home.</p><p>I had promised him I would see him before, but I did not make it. He had promised me he would stick around till the New Year. But he did not make it. This is one of those rare moments when I would actually physically give anything to make us keep our promises.</p><p>I first convinced myself that he was asleep. After all, he looked so peaceful, with a smile. Then I saw his hands. They did not look real, he did not look real. It was not him any more.</p><p>After the service there was a black car, parked right outside. I made sure I stood far from it so I could not see it or even hear it take him away. Take hope away. Take everything away.</p><p>In that moment everything he ever was, everything he would be, everything he would do and he wished for was gone, taken away.</p><p>I have not been able to sleep without thinking about it ever since.</p><p>We stayed in LA that whole month, and while I was working, playing with my kid, spending time with family and enjoying the sun and the ocean, I kept thinking about the wishlist. The wishlist of a person who made it on NYE 2022, with a lot to tick off, that he did not manage to. Health goals, travel goals, business goals, life goals. First these thoughts made me sad. A type of sadness you would not wish upon anyone. Helplessly sad.</p><p>Gradually the feeling has evolved. I am finding it hard to describe it right now, but it is that feeling that made me want to write about conviction.</p><p>Conviction has Latin roots. It has come to mean a strongly held belief. The more educated you are, the more you know, the more experience you get and the more people you meet, the harder it is to build conviction. Yet, the more valuable it becomes once you do.</p><p>The way humans work is that they reward you for being right in a very obvious kind of way. Being right is a path that most have walked, hence the outcomes are predictable. For instance doing well at school is universally considered to be the right thing to do, you cannot go wrong with that.</p><p>Marrying and having kids is another societal dogma that is widely regarded as the expected and right path.</p><p>No one will judge you or think you are out of your mind for wanting to be a doctor, a lawyer, an accountant or a consultant either. After all, those are well examined ways of adding value and making a living in a modern world.</p><p>There are many things in life that are way more ambiguous and way harder to get right.There are less probabilities that can be assigned with certainty.</p><p>Less people are thinking about them, likely no one or very few have gone there and mostly these are portrayed as scary. All life is a game of probabilities, yet some are way better explored than others, hence easier to decide on.</p><p>Then there are things that introduce Knightian uncertainty, unknown unknowns where the probabilities are impossible to see.</p><p>Starting a business can be one. It can be calculated depending on how much you know the domain and the problem you are solving, but also presents a lot of unknowns you have no ways of measuring. Relationships with people, ability to lead, ability to attract capital, market conditions and so on.</p><p>If you decide to not go to school, not get married in your 20s and early 30s, not do whatever others conceive of as the right paths, you wake up scared. Facing true uncertainty.</p><p>If you quit your job and leave for a new country where you do not know anyone, you wake up scared.</p><p>When you decide to divorce or pivot your business, you wake up scared.</p><p>There are many of these forks in fact. Some are more controversial than others. More Knightian in fact.</p><p>While they fill you with angst, these forks are exactly the best times to build conviction. To start learning and examining the world and yourself, finding gaps and building beliefs.</p><p>Once you have built one, sticking to it and putting all your “eggs” there is the best thing you can do. You do not hedge when you have conviction.</p><p>This is not to say you have to pick something no one else sees, you have to go against the grain.</p><p>Often really amazing ideas are out there, lying on the ground, staring at you and no one is bothered to pick them up. Or maybe it is too much hard work to actually get it done. Or perhaps it takes optimism and you are not wired that way. Most of us are not. We prefer to expect the worst, see the wrongs, and appear smart.</p><p>The “Man in the Arena” became a meme this year. Yet it is a beautiful reminder on what counts and an example of having conviction.</p><p>Conviction is not the same as passion. It can get you to feel things, and it definitely helps you get yourself out of the gutter, but it takes you further, it is less abstract, less emotional, sturdier.</p><p>When you have conviction you can see things happen before they do, physically see them, be there, touch it, before anyone else can.</p><p>Athletes often experience this when they have a goal. Nothing to do with visualization and techniques that are similar. This is a way more logical path of absorbing knowledge and examining the world around us in order to form beliefs.</p><p>Working with people who have conviction is fantastic. The hard things become easier, the impossible stuff gets done. Choosing people who have conviction is an art of its own. Meeting them is luck.</p><p>As we are stepping into 2024 with a lot of losses, having lost and continuing to lose so much human life everywhere around the world, as I am standing here staring at my own loss in 2023, listening to the silence of someone I loved dearly, I want us to step up.</p><p>Start learning about things around us, start going for things we believe in and surrounding ourselves with people who do. Once you have found something you have deep conviction in, dive in . With your time, your capital, your emotions. After all there is not much else to do out here in the universe.</p><p>Someday we are going to be a list of stuff that we have not managed to get done. Boxes unticked. It is not sad, if we went for what we truly believe in, what we are ready to vouch for, where we are ready to appear silly or embarrass ourselves.</p><p>There are few things in life that you can know for sure and very few things you are ready to embarrass yourself for. Once you know you are right, make it happen. Fight for it.</p><p>As one of my old favorites, Rene Magritte would say, there is nothing confused, only the mind. I have not actually heard him say it. They say he did.</p><p>Happy New Year .</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6b08f7b0dd59" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Why invest in A.I. ?]]></title>
            <link>https://medium.com/@foundress/why-invest-in-a-i-9dbe7dd59945?source=rss-7d289ad87fc3------2</link>
            <guid isPermaLink="false">https://medium.com/p/9dbe7dd59945</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[vc]]></category>
            <category><![CDATA[investing]]></category>
            <category><![CDATA[thesis]]></category>
            <dc:creator><![CDATA[Naré]]></dc:creator>
            <pubDate>Tue, 04 Apr 2023 20:32:13 GMT</pubDate>
            <atom:updated>2023-04-04T20:32:13.995Z</atom:updated>
            <content:encoded><![CDATA[<h3>Why invest in A.I. ?</h3><h4>2023 edition generated by ChatGPT</h4><p><strong><em>(Disclaimer: this post is fully generated and edited by ChatGPT prompted by my 2017 post on investing in A.I. and one sentence on latest developments in LLM-s).</em></strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/512/1*4VJ9Kd4zKZ3TxNU8FfDEGA.png" /><figcaption>generated by Stable Diffusion 2.1</figcaption></figure><p>The rapid advancements in artificial intelligence (AI) since the release of GPT-4 have brought us closer to general intelligence than ever before. With AI becoming a vital layer in software and the ongoing AI-ization of various industries, it’s time to revisit our understanding of AI’s potential and reevaluate software businesses in this ever-evolving landscape.</p><p>Since the original post in 2017, the hype around AI has both fallen and risen, with valuable lessons learned along the way. Today, we have a clearer picture of the opportunities and challenges AI presents to businesses, investors, and society as a whole.</p><p>In this updated post, we’ll explore the current reality of AI-first businesses, addressing how the landscape has evolved and considering new perspectives on investing in AI.</p><ol><li><strong>Ubiquity of AI:</strong> AI is indeed becoming more ubiquitous as a technology, and while it may not be a distinct category in itself, it represents a significant shift in the enterprise software stack and the software stack as awhole. AI is now a crucial component in the development of software transforming industries and creating new opportunities to create a proprietary advantage at the earliest stage.</li><li><strong>Large enterprises becoming AI-first: </strong>While major tech giants like Google, Amazon, Facebook, and Microsoft have successfully integrated AI into their operations, most companies still struggle to make the shift. This gap presents opportunities for AI startups to innovate and capture market share</li><li><strong>The importance of data and specialized LLMs: </strong>Data is undeniably critical for AI development, but it’s no longer a scarce resource. AI-first businesses can now tap into previously inaccessible or underutilized data from various industries, creating new solutions and disrupting traditional models. The rise of specialized large language models (LLMs) also demonstrates how tailored AI solutions can outperform foundational models in certain tasks, offering unique value propositions in niche but fast-growing markets.</li><li><strong>Commoditization of point solutions and the rise of platform startups:</strong> As AI technology becomes more accessible and widespread, point solutions are increasingly being commoditized. This trend will lead to the emergence of platform startups that offer a suite of AI-driven services and tools, enabling businesses to harness the power of AI more effectively and efficiently. These platforms are well-positioned to capture significant market share by offering comprehensive, end-to-end solutions that cater to a vertical user base with a broad range of products</li></ol><p>In conclusion, the AI landscape has evolved considerably since 2017, with a new era of AI-first businesses emerging as a result. These businesses represent a unique investment opportunity given they meet certain criteria and can rapidly expand end user products while powering the data flywheel.</p><p>These new startups will resemble big companies way earlier than before, yet they will move faster, burn less cash and will launch things in batches.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9dbe7dd59945" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Ntropy Core Principles]]></title>
            <link>https://medium.com/ntropy-network/ntropy-core-principles-cd98e46675d1?source=rss-7d289ad87fc3------2</link>
            <guid isPermaLink="false">https://medium.com/p/cd98e46675d1</guid>
            <category><![CDATA[cutlure]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[company-culture]]></category>
            <dc:creator><![CDATA[Naré]]></dc:creator>
            <pubDate>Wed, 08 Feb 2023 10:52:48 GMT</pubDate>
            <atom:updated>2023-02-08T10:55:25.275Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Retrofuturistic image of a woman going to space made using Stable Diffusion 1.5" src="https://cdn-images-1.medium.com/max/512/1*Sp2fwxzXOaLJD2SUJ8lXmw.jpeg" /><figcaption>Retrofuturism. Stable Diffusion 1.5</figcaption></figure><p>We went through a cultural reset recently and had to rethink and rephrase who we are and what we stand for. I thought sharing this publicly will help those who join us on this journey.</p><p>Below is a set of principles that every Ntropian stands by. They are not fit for everyone and that is fine. Some we follow religiously already, others we are still working on.</p><p>As a whole, they unite us, as Ntropians, in achieving our goal.</p><p><strong>The goal is</strong> to build a company of generational importance which has a global impact that outlasts ourselves.</p><p><strong>Every Ntropian</strong> is a co-owner and has a vested interest and role in our success. This needs to feel like and is going to be your chef-d’œuvre. We are going to make sure every single person on this team has the means and the freedom, including economic freedom, to realize their wildest ambitions.</p><h4><strong>How we think and act</strong></h4><ul><li><strong>Optimists</strong></li></ul><p>We believe we have an unfair/asymmetrical advantage in what we do and are disproportionately positioned to dominate. This means we get into solving problems with a predisposition of being able to crack them.</p><ul><li><strong>Humble</strong></li></ul><p>We do not take ourselves too seriously. This is the first thing you will experience becoming a part of this team. We have perspective and we are very aware how faulty each and all of us can be.</p><ul><li><strong>Selfless</strong></li></ul><p>We think individually, but make decisions that benefit the group. Ntropy comes first. If you are unable to put the personal behind when needs be, you will struggle here.</p><ul><li><strong>Thorough</strong></li></ul><p>We care deeply about the work we do, about every detail or we do not do it at all. Not being thorough is costly.</p><ul><li><strong>Kind</strong></li></ul><p>We are kind to ourselves and our colleagues. We give feedback directly from a position of helping them win and us win collectively.</p><ul><li><strong>Fast and fearless</strong></li></ul><p>We are not afraid of making mistakes, failing, picking the wrong path, starting from scratch. We do it fast. The answer to a lot of things is just doing more better.</p><ul><li><strong>Ferocious</strong></li></ul><p>We will do whatever it takes. We have a thick skin and we do not get upset by things that do not matter in the grand scheme or do not bring us closer to our goal.</p><ul><li><strong>Gritty</strong></li></ul><p>We work hard. The only way to win is to outwork your competition.</p><ul><li><strong>Resourceful</strong></li></ul><p>We make more from less. If you are lacking a tool, a conference sponsorship or an expensive agency on retainer to get something done, you won’t be happy here. Doing more from less makes us creative and better.</p><ul><li><strong>Entropic</strong></li></ul><p>We often do things that haven’t been done before or that we ourselves are doing for the first time. This means we can survive and thrive in chaos while seeking structure, where it is necessary.</p><h4><strong>Ntropy is not the place for</strong></h4><ul><li>company politics of any kind</li><li>any shape or form of entitlement, ego patting and sense of being owed to</li><li>disrespectful treatment of each other and our colleagues</li><li>defensive discourse. We do things from first principles always. This means you are open to questioning your assumptions.</li></ul><h4><strong>How we communicate</strong></h4><p>How we speak is crucial to how we think, how we act and who we are. Getting this right creates leverage both internally and externally, when engaging with our customers.</p><ul><li><strong>Concise</strong></li></ul><p>Information density per sentence is important. If you can say it in a couple of words vs prolonged sentences, you should. If you need to pause before saying it, you should. If you can do things async vs jumping on a call, you should.</p><ul><li><strong>Useful</strong></li></ul><p>When you say things, they should be answering a question and covering a need. Whatever you say should add to the discussion at hand.</p><ul><li><strong>Authentic</strong></li></ul><p>Say exactly what you think. Ntropy is not the place for sugarcoating or political correctness</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cd98e46675d1" width="1" height="1" alt=""><hr><p><a href="https://medium.com/ntropy-network/ntropy-core-principles-cd98e46675d1">Ntropy Core Principles</a> was originally published in <a href="https://medium.com/ntropy-network">Ntropy</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The toy rabbit, the chase for freedom and the wholeness of time]]></title>
            <link>https://medium.com/@foundress/the-grey-rabbit-the-chase-for-freedom-and-the-wholeness-of-time-f915a89e0e1a?source=rss-7d289ad87fc3------2</link>
            <guid isPermaLink="false">https://medium.com/p/f915a89e0e1a</guid>
            <category><![CDATA[language-model]]></category>
            <category><![CDATA[childrens-stories]]></category>
            <dc:creator><![CDATA[Naré]]></dc:creator>
            <pubDate>Sun, 01 Jan 2023 20:46:11 GMT</pubDate>
            <atom:updated>2023-01-01T20:50:58.002Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/512/1*eIzFptxNYKSdQ1vKAOp3qA.jpeg" /><figcaption>the rabbit generated in Playdoh style , Stable Diffusion 1.5</figcaption></figure><h4>Disclaimer:</h4><p>This whole story and the illustration are generated by LLM-s. This was one of my Christmas presents for my 16-month-old, who has an adorable toy, a grey rabbit with ears that have been chewed on over and over again. As the rabbit has been there from day one to comfort my little one, I thought he’d appreciate a personal story about the feelings and life of the rabbit that stay behind the scenes of our daily life.</p><p>Moreover, the story can have endless versions of endings as it has been generated by ChatGPT. You do not have to settle for a season finale ever again or be unhappy with the last chapter.</p><p><em>P.S. all the characters are in Playdoh style and if it were a book it could be sold with playdoh to re-create your own characters.</em></p><p><strong>The big escape</strong></p><p>Bun Bun was tired of being a toy. As the favorite grey rabbit of one-year-old Gabriel, he spent his days being chewed on, thrown around, and generally mistreated. But he was determined to find more meaning in life, to be more than just a plaything for the little human.</p><p>So, one day, Bun Bun made the decision to run away. He hopped out of Gabriel’s crib and into the great unknown. It was a scary, but exhilarating feeling.</p><h4><strong>The journey</strong></h4><p>As he journeyed further and further away from home, he encountered all sorts of adventures. He met a little girl named Lisa who tried to adopt him, but Bun Bun knew that his heart belonged to no one but Gabriel. So he pretended to be her friend, only to escape when she was asleep.</p><p>He also met a fox whose fingers were burnt in a forest fire. The fox was weaving gloves for their Grandma, and Bun Bun offered to help. As a thank you, the fox took Bun Bun through the most beautiful parts of the forest and showed him a world he had never known. But even as he marveled at the beauty around him, Bun Bun knew that the forest wasn’t where he belonged.</p><p>So he continued his journey, eventually finding himself in the bustling city of Tokyo. As he got wet on the streets, a wise geisha took him in and gave him tea with pieces of sugar, his favorite snack. In return, Bun Bun danced and sang during the tea ceremony, and even got a chance to be the entertainment at a party. But he couldn’t shake the feeling that he didn’t belong there, either.</p><p>So he set off once again, this time heading towards Russia. Along the way, he met a rabbit named Ella, who was a photographer. They fell in love and traveled together, taking photos of sunrises and sunsets. But even as he enjoyed this new adventure, Bun Bun knew that it wasn’t meant to be. In the middle of the night, he left Ella with a letter telling her that he would always love and remember her, but that staying there right now didn’t make sense.</p><h4><strong>Meeting the owl</strong></h4><p>Finally, during one of his escapes, Bun Bun found himself stuck in a mouse trap somewhere in Finland. Just as he was about to give up hope, a wise owl flew in and rescued him. The owl, who was a loner, told Bun Bun that true freedom wasn’t about running away from his fate or trying to find happiness in external things. It was about embracing who he was and finding joy in right now.</p><h4><strong>The journey back (version one)</strong></h4><p>With this newfound understanding, Bun Bun set off on a journey back home to Gabriel. When he finally arrived, he found the little human fast asleep in his crib. Bun Bun gently woke him up, and to Gabriel’s surprise, the rabbit began to cry.</p><p>“I’m happy,” Bun Bun said through his tears. “I can cry. That’s the biggest freedom of all.”</p><p>Gabriel chuckled and reached out to grab Bun Bun by the ears, just like he always did. And this time, Bun Bun didn’t mind. He was just happy to be home, with the one he loved.</p><p>As Gabriel took Bun Bun to his toy box to join his other friends, Bun Bun knew that he had finally found the freedom he had been searching for. It wasn’t about running away from his fate, or trying to be something he wasn’t. It was about being himself, and loving unconditionally. And that, he realized, was the greatest freedom of all.</p><h4><strong>The journey back (version two)</strong></h4><p>But as he journeyed back, he realized that there was another freedom he had been missing out on. The freedom to explore, to see new things, and to experience all that life had to offer.</p><p>So instead of returning home to Gabriel, Bun Bun made the decision to keep traveling. He visited new countries, met new animals, and had all sorts of exciting adventures. And even though he missed Gabriel and the life he had left behind, he knew that this was where he belonged now.</p><p>As the years went on, Bun Bun became a world-famous adventurer, known for his bravery and his love of adventure. He returned home to Gabriel every few years, always with a new story to tell and a new treasure to share.</p><p>And even though he knew that he could never fully escape his fate as a toy, Bun Bun was happy to embrace it in his own way. He was a toy, but he was also so much more. And that, he knew, was the greatest freedom of all.</p><p><strong>The journey back (version three and my favourite version)</strong></p><p>As Bun Bun set off on his journey back home to Gabriel, he knew that he had finally found the freedom he had been searching for. It wasn’t about running away from his fate, or trying to be something he wasn’t. It was about being himself, and loving unconditionally.</p><p>But just as he was about to return home, something strange happened. Bun Bun felt a sudden jolt, and then everything went black. When he opened his eyes, he found himself in a laboratory, surrounded by strange machines and people in white coats.</p><p>It turned out that Bun Bun had been part of a top-secret experiment all along, and the whole journey he had thought he had taken was just an elaborate simulation. The scientists had wanted to see how a toy would react to being given the freedom to make its own choices, and Bun Bun had passed their test with flying colors.</p><p>At first, Bun Bun was angry and confused. But as he learned more about the experiment and the people who had run it, he began to see things in a different light. He realized that he had been given a unique opportunity, one that few toys had ever had. And as he returned home to Gabriel, he knew that he had gained something even greater than freedom — knowledge.</p><p>From that day on, Bun Bun was a changed rabbit. He was still Gabriel’s favorite toy, but he was also so much more.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f915a89e0e1a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Transaction data is an uncut gem]]></title>
            <link>https://medium.com/ntropy-network/transaction-data-is-an-uncut-gem-47a6ee18f6c?source=rss-7d289ad87fc3------2</link>
            <guid isPermaLink="false">https://medium.com/p/47a6ee18f6c</guid>
            <category><![CDATA[ai-categorization-api]]></category>
            <category><![CDATA[fintech-infrastructure]]></category>
            <category><![CDATA[fintech]]></category>
            <category><![CDATA[transactions]]></category>
            <dc:creator><![CDATA[Naré]]></dc:creator>
            <pubDate>Tue, 21 Dec 2021 10:10:03 GMT</pubDate>
            <atom:updated>2021-12-21T15:52:04.541Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/800/1*5CXl7Nzuh5_EMk6n7CiJqw.jpeg" /></figure><p>Transactional data is generated every time money is transferred between two entities, one initiating the transaction order and the other receiving it. In the previous post we described the complexity of extracting information hidden in transaction data. However, if done right, the use cases are endless.</p><p>According to Mastercard’s recent<a href="https://b2b.mastercard.com/news-and-insights/research-and-reports/rise-of-open-banking-north-america"> report, </a>over 9 in 10 people in America use technology to manage their money. These are mostly digital products to complete simple financial tasks such as paying your bills, banking and so on. However, the need for more sophisticated solutions is becoming clear, especially among late millennials and GenZ-s. Various forms of financial forecasting, automated investing, flexible savings accounts, crowdfunding and crypto integration are some examples. At the core of all these possibilities lies the ability to extract insights from transaction data in an automated way.</p><p>Here are a few emerging use cases that we love the most:</p><ol><li>Climate positive purchases</li></ol><p>Where and how you spend your money directly affects the planet. To be able to change, we need to keep track of, and understand our spending habits.</p><p>High-resolution transaction data will play an important role in building a carbon negative future. There is a <a href="https://www.vox.com/the-goods/2020/3/5/21155020/companies-carbon-neutral-climate-positive">whole generation</a> of companies, such as <a href="https://www.ikea.com/gb/en/">Ikea</a>, <a href="https://www.henkel.com/company">Henkel,</a> <a href="https://www.brewdog.com/">Brewdog</a> and <a href="http://microsoft/">Microsoft</a> committing to this.</p><p>2. Revenue based financing for businesses.</p><p>A few decades ago, as an entrepreneur building something from scratch, you had to have great “friends and family” to access capital. Later on you had to look and act the part to gain the trust of your bank manager.</p><p>Now, you need to have years of experience being in debt, a credit score and fill out a bunch of paperwork. The experience is a pain. With a bit of luck, you may get the capital you need, but not on the terms or timelines that you would like.</p><p>Good quality transaction data can also provide a detailed view of the present and possibly the future state of a business. If processed by a machine-learning model, it can reduce the time to access capital from weeks to seconds. Indeed, transactions can be pulled from the bank accounts of your customers in seconds, sent through an enrichment engine <a href="http://www.ntropy.com/">(e.g. Ntropy API)</a> and fed into your underwriting pipeline for a final decision.</p><p>3. Embedded money.</p><p>Imagine you are the operating system, aka the CRM, the payment processor, the employee management system or the customer comms layer for car repair shops in the US.</p><p>You have saved them from the manual pen and paper operations and alleviated many of the operational pains associated with managing their business. Your software environment is technically the home for their business. However, when they want to grow, make changes, hire more, invest in inventory or get started, they need to go to a bank. This is an entity who knows nothing about them or their business and who has to start from scratch to understand their case. Banks spend large amounts of time and money to assess the credibility of applicants for their financial services and later charge that back to their customers in onboarding costs, communication and interest rates.</p><p>Instead, vertically integrating into the business of your customers, you can allow them to connect their cash flow and financial information without leaving your premises. You get very happy customers and<a href="https://www.fool.com/earnings/call-transcripts/2018/11/06/mindbody-inc-mb-q3-2018-earnings-conference-call-t.aspx"> 3–5x LTV.</a> They get time and resources to run their business instead of messing about with banks.</p><p>In order to make this happen, banking data needs to be machine-readable at scale and easily married to other meta-data about a business, including CRM information, customer service data and reviews.</p><p>4. Hyper-personal rewards.</p><p>As much as we hate the word “hyper”, getting things that you actually care for vs generic discounts and points is fun and convenient. Adoption of new financial experiences is generally driven by two factors: accessibility and convenience.</p><p>Fine-tuned rewards and recommendations available at the fingertips of your customers are excellent for boosting their loyalty, satisfaction and product experience. To do this, you need to know where they spend their money and time. That information is, you guessed it, hidden in their transaction data.</p><p>5. Making money last</p><p>Automated business expense management and personal finance management have so far been the primary use case for transaction data.</p><p>In order to allow your users, whether a business or an individual, to have an overview of how to make their money last, you need a granular understanding of where it is coming from and where it is going.</p><p>Traditional money management contains a lot of friction. It is retroactive, time consuming and inefficient. It works for people and companies who are ready to go the extra mile and put a lot of effort in making it work. For the rest of us, most of these tools fail. Hence, a hotbed of new opportunities for fintechs emerges.</p><p>However, as most fintechs, like a <a href="http://truebill.com/">Truebill </a>or a <a href="http://ramp.com/">Ramp</a> are moving towards empowering their users with a more <a href="https://b2b.mastercard.com/news-and-insights/research-and-reports/rise-of-open-banking-north-america">holistic</a> suite of money experiences, diverting focus on trying to build a transaction categorization engine in-house is not a poor strategy. The effort is extensive, the cost of mistakes is high and the risk of ending up with a suboptimal solution is apparent. For example, tagging data labeling as rideshare or restaurant costs as online education can result in immediate churn of a customer.</p><p>If you want your suggestions to be reliable, you need high quality and scalable transaction tagging and real-time notification for users to take action.</p><p>Finally, one of the biggest emerging use cases for interpreting transaction histories is within the creator economy and identifying potential high earners in order to fund and scale their businesses. We will elaborate on that in our next writeup.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=47a6ee18f6c" width="1" height="1" alt=""><hr><p><a href="https://medium.com/ntropy-network/transaction-data-is-an-uncut-gem-47a6ee18f6c">Transaction data is an uncut gem</a> was originally published in <a href="https://medium.com/ntropy-network">Ntropy</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>