<?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 Ian McCullough on Medium]]></title>
        <description><![CDATA[Stories by Ian McCullough on Medium]]></description>
        <link>https://medium.com/@ipmcc?source=rss-3567854fbab1------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*JRfXC72VtZuvg6RF62-XkQ.jpeg</url>
            <title>Stories by Ian McCullough on Medium</title>
            <link>https://medium.com/@ipmcc?source=rss-3567854fbab1------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 01:20:54 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ipmcc/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[So, This Was Wild…]]></title>
            <link>https://medium.com/@ipmcc/so-this-was-wild-3306396a45ff?source=rss-3567854fbab1------2</link>
            <guid isPermaLink="false">https://medium.com/p/3306396a45ff</guid>
            <category><![CDATA[scams-to-avoid]]></category>
            <dc:creator><![CDATA[Ian McCullough]]></dc:creator>
            <pubDate>Thu, 16 Apr 2026 18:39:45 GMT</pubDate>
            <atom:updated>2026-04-29T15:50:58.286Z</atom:updated>
            <content:encoded><![CDATA[<p>This morning, I was sitting down, getting started in on some work, waiting for a carpet installer, and generally minding my own business. My phone rings. I don’t recognize the number, but I’ve been getting quotes on a lot of little home improvement projects recently, so a lot of companies call me whose numbers I don’t recognize.</p><p>I answer. The person on the far end says that they are Sergeant Ryan Foster of the Allegheny County Sheriff’s Department, and they’re calling because I’m supposedly wanted on charges of failing to appear for a confidential grand jury selection hearing yesterday, and contempt of court. They claimed that I was sent a jury selection notice in a certified letter and that I signed for it on March 13th, 2026. I have no memory of this, I live alone, but at this point in my life, I accept that it’s <em>physically possible </em>that I signed for some certified letter and don’t remember doing so. I said, “I have no recollection of this, and I take jury service seriously, so this feels very odd. How do we sort this out?”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/286/1*heF2_oWeHlWdMg0T4sxn-A.jpeg" /><figcaption>Actual Certified Mail (Source: Google Image Search)</figcaption></figure><p>The story from here gets hinky pretty quick, but I don’t need any legal trouble at the moment (or ever, really), so I go along with it. The woman on the other end, who sounds very much like an African-American woman (this will come up later), says that I need to post bond for myself. OK, fine. They say it has to be cash. I’ve bailed people out of jail before, and my experience has been that law enforcement typically doesn’t take personal checks or credit cards, so while this is a little weird, it doesn’t feel <em>impossible… yet.</em></p><p>The script she was using was <strong>thorough </strong>and she <em>definitely</em> sounds like she’s done this before<strong>. </strong>There was a whole story with dates and times for when the jury notice was delivered, the postman who delivered it, and how they verified it, etc. They also had pretty convincing background audio of what sounded like a police squad room. They had my (correct) address. They had my phone number. They had pretty much everything. Docket numbers, penal code numbers, and she claimed her name was Sgt. Ryan Foster and gave me a badge number. As far as the phone script, these folks have thought of every detail. Unfortunately for them, they hadn’t really thought much past the phone call.</p><p>They also claim that there is a “Make &amp; Maintain Contact” order on me, so I supposedly had to stay actively on the phone with them — the entire time I’m showering, using the toilet (I made sure to give them the full sonic experience), getting dressed, going to the bank, going to the “bail bonds” place, going to the courthouse, etc… which, conveniently, prevents me from texting anyone, calling anyone, or making most other attempts I would normally make to explore the veracity of this call, in the moment. They also claim that because the jury summons references a “sealed” case, that I was also under a “gag order”, and I was not to talk to anyone about what was going on. I explained that since I’d never seen the summons, I had no information to leak, but they were adamant that I not talk to <em>anyone</em> about this, “for [my] safety.” They said something like, “If anyone knew you were going to be on a grand jury, they could call CrimeStoppers on you, and at that point, it would be out of your hands; You’d be booked.”</p><p>I explain that I will need to get off the phone with them in order to call my lawyer, but they said that I only had the right to an attorney if I was being detained and that the “sealed” grand jury stuff meant that I couldn’t talk with even my lawyer about it <em>unless</em> I was already in police custody.</p><p>In this mindset, I go to the bank and take out the money, still on the phone with this woman the whole time. She had given me the addresses of three <strong>supposed</strong> ‘bail outlets’ where I could pay, that weren’t on government property, and this was important because if I were on government property I could be arrested and booked, and if that happened, it would take up to 72 hours to get bailed out, and Sheriff’s deputies don’t work on weekends, so I’d likely be in the clink for ~5 days.</p><p>I go to the first address she gave me, 601 Penn Ave, and it’s the 7–11 across from Gaucho. Here’s what I find:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WG9M7-Sn_rFWs_QzYeg3aQ.png" /><figcaption>Not a Bail Bonds place (Source: Google Street View)</figcaption></figure><p>I tell her, “This is a 7-Eleven, not a bail bonds place. What’s going on here?” And she says, “No, you must’ve taken the address down wrong. It’s 610 Wood St, not 601 Penn. It’s inside a Dollar General.” I’m thinking, “Who pays surety bonds at Dollar General?” but I walk up to 610 Wood St, and sure enough there is (sort of) a Dollar General there, but it’s literally <strong>out of business.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Q_zglkjNrQJCVRhWXaL8cg.png" /><figcaption>DG. Note the “For Sale” sign in the upper left. (Source: Google Street View)</figcaption></figure><p>I peered in the windows and it was completely dark, and empty, front to back, side to side. Then she says, “There’s another bail bonds place at [8-thousand something] Penn Ave.” I said, “That’s gotta be eight miles from here! I’m not doing that. But you know what? I’m just a few blocks from the county courthouse. I’m pretty sure you’re a scammer at this point, but how about I just go to the Sheriff’s office there and turn myself in?”</p><p>At this point, while they’ve pretty much all but confirmed to me that this is a scam already, then they seal the deal by freaking out at the idea of me going to the <em>actual</em> Sheriff’s office. “If you do that, you could very well be in jail for 72 business hours! I’m tryna help you here, Ian!”</p><p>With her still on the phone, trying to convince me to go to a Dollar General 8 miles away, I go to the county courthouse a few blocks away. When you go in, you have to go through a metal detector and put your cell phone and everything else in a bucket. It’s just like the airport.</p><p>The guy at the metal detector asks, “What are you here for today?” I say, “I’m looking for the Allegheny County Sheriff’s Department.” He asks, “What business do you have at the Allegheny County Sheriff’s Department?” I start to give him the three sentence recap, but he interrupts me and says, “Yeah, that’s a scam. Happens all the time. But it’s Room #317. They’ll tell you the same thing.” I tell him, “I hear you, and I was pretty suspicious it was a scam already, but I want to hear <strong>them</strong> say that.” (Note: She’s still on the phone!) He says, “Suit yourself!” and waves me through the metal detector.</p><p>I arrive at the Sheriff’s office and it’s a bevy of activity. Lots of guys in shackles coming in and out. I explain the situation to the two officers that are sitting in the foyer of the office and immediately they say, “Yeah, it’s a scam. We get something like three or four of these a week.” They ask if I’ll give them my ID. I hand it over. After all, if there <em>is</em> a problem, I want to get it taken care of. He bangs on the keyboard for a few seconds and says, “Yeah, there’s nothing here, but let me ask: <em>Did you actually send any money to anyone?” </em>I explain that I did not. He says something like, “Good! Don’t! This is <strong>not </strong>how these things work. If we have business with you, we don’t give you warning calls, or text messages, and we <em>definitely </em>don’t ask you to wire money from a Dollar General. We come and get you. You <strong>will not</strong> have <strong>any</strong> warning when we come for you!” While that response was a little more ‘Dirty Harry’ than was entirely necessary, I appreciated the official confirmation that there are no outstanding warrants for me.</p><p>Then, they ask for the phone number that’s calling me, I look down at my phone, and realize the woman has hung up on me… after being on the phone with me for almost <strong>3 hours!</strong> The guy says, “Block the number. Don’t try to mess with them, or tell them you talked to us, but also, don’t <em>ever</em> send cash to anyone for bail unless you’re in the courthouse or at a business that says ‘Bail Bonds’ on the front sign.”</p><p>Then the woman there asks me, “Out of curiosity, did they give you a name?” I said, “Yeah, ‘Sgt. Ryan Foster’. Do you have a Ryan Foster?” and she said, “Wow! Yes, we absolutely do. What did he sound like?” and I said “They sounded like a young African-American woman. Which seemed odd for the name, but what do I know, right? It’s 2026.” She said, “Sgt. Ryan Foster is a big white dude. There’s no way anyone would ever mistake him for a young African-American woman over the phone.” Googling for a hot second, Sgt. Foster appears to be this dude here:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/201/1*0RBrSYnxtAeSPxDcx4V2kg.jpeg" /><figcaption>Highly unlikely to be the woman I was on the phone with. (Source: ACSD Facebook)</figcaption></figure><p>In the end, I avoided any significant harm, but I wasted my entire morning. I had to get a friend to go over to my house and wait for the carpet installers. I had to pay $19 to park downtown. I paid my friend $40 to cover his Ubers and time. Essentially, I’m out $60, a little bit of gasoline, and a half day of my time. Oh yeah, I also got to have a panic attack on the streets of downtown Pittsburgh while on the phone with a scammer. Super cool.</p><p>Anyway, if this happens to you? <strong>Don’t fall for it.</strong> I really can’t believe that I <em>almost</em> fell for this. I normally catch onto these very quickly, but this one had me going for way too long. I don’t think I would’ve actually put money into a ‘bail kiosk’ in a Dollar General, but if I got this close, I have to assume that other people go through with it.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3306396a45ff" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Maker Economics]]></title>
            <link>https://medium.com/@ipmcc/maker-economics-6fc6b3530988?source=rss-3567854fbab1------2</link>
            <guid isPermaLink="false">https://medium.com/p/6fc6b3530988</guid>
            <category><![CDATA[electronics-embedded]]></category>
            <category><![CDATA[electronics]]></category>
            <category><![CDATA[makers]]></category>
            <dc:creator><![CDATA[Ian McCullough]]></dc:creator>
            <pubDate>Wed, 01 Apr 2026 15:18:07 GMT</pubDate>
            <atom:updated>2026-04-02T11:43:59.846Z</atom:updated>
            <content:encoded><![CDATA[<p>Anyone who follows me here knows that I’m on a bit of a tear with the ‘maker stuff’ this year. Specifically, most recently, I’ve been working on an artsy project that uses NeoPixels to make a wall clock. The longer story of the project is <a href="https://medium.com/@ipmcc/good-vibes-tiny-puters-pcbs-and-3d-printing-c5c3f704205d">over here</a> but I’m here today to talk about the unusual economics of NeoPixels, and possibly a lot of other ‘maker bits’.</p><h3>The First Attempt</h3><p>As described in the other article, I needed to acquire a lot of these NeoPixels. After I convinced myself, without trying, that I could still solder 2x2mm surface mount parts by hand, I bought 100 of <a href="https://www.adafruit.com/product/6428">these</a> for $30:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/970/1*tKuFtkJ8sm5lKVxqBOhWAQ.jpeg" /><figcaption>NeoPixel Nano 2020 RGB LEDs — WS2812B (Source: Adafruit.com)</figcaption></figure><p>When they got here, I soon realized that it was substantially harder for my aging eyes and hands to work with parts this small (absent a pick &amp; place machine and a reflow oven), but I’d already ordered PCBs to fit these tiny parts, so I was stuck with them for at least part of the project.</p><h3>Baby Steps…</h3><p>After working with the 2x2mm parts, I decided that anything further — like, say, the hands of the clock — would be based on the <a href="https://www.adafruit.com/product/3094">more common 5x5mm parts</a>. That’s not so bad, right? And again, 100 LEDs for $30.</p><p>I understand economies of scale, and I realize that if I were ordering 100,000 of these, they’d be cheaper than three for a dollar. But I don’t want 100,000 of these. I want a few hundred… <em>maybe</em>. So I ordered the 100 pack from Adafruit (and then got all bitter that it took more than one day to arrive, <em>thanks Amazon for ruining me.</em>)</p><h3>Then I Saw This!</h3><p>Subsequently, in my travels, I came across <a href="https://www.amazon.com/dp/B088BTYJH6">this listing on Amazon</a>:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Al8CvUvBxRSF3Zimq_G5OQ.png" /></figure><p>And I was left thinking, “So… wait… I can get 256 of these LEDs for $17?! Overnight?! Why did I just pay $30 for 100 of them!?!? And wait 5 days!!” Now, I must confess that desoldering 256 LEDs with a heat gun was not exactly the <strong>most</strong> fun way I’ve ever spent a half an hour, but it was far from the worst either.</p><p>Now, I have a little dram jar with 256 5x5mm LEDs in it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*I6N6pigHXD3aYKPqcTT-mA.jpeg" /><figcaption>A shot full of LED! (Source: me)</figcaption></figure><p>At 256 for $17, that’s about 7 cents per LED, which is a far bit better than 100 for $30 – ~75% cheaper, in fact!</p><h3>Why?</h3><p>Well, I’m sure that the company making the 16x16 matrix of LEDs is buying them in the hundreds-of-thousands (if not millions) quantities, so that’s great for them, and they seem to be passing the savings along to me. I just can’t escape the inevitable conclusion: <em>Either the 16x16 matrix company is selling those things at a loss for some unknown reason… </em><strong><em>or</em></strong><em> Adafruit, SparkFun and other parts vendors are selling these things at a </em><strong><em>huge</em></strong><em> profit.</em></p><p>Now, I don’t begrudge any business making a profit by selling things to people for a price they’re happy to pay. And let’s be clear: I was not overly bitter about paying $30 for 100… until I saw these giant matrices. Like any red-blooded American, I always want everything to be cheaper. The spread here is just <em>so</em> wide that I thought I’d put something out in the world (this article) and see what everyone else had to say about it.</p><p>The individual/bare parts on DigiKey aren’t price-competitive. Even the matrix products sold on DigiKey are <em>wildly</em> more expensive than the ones on Amazon. Even at LED quantities of 10,000+, buying these matrices from Amazon and desoldering them is<strong> still</strong> winning handily over buying bare parts.</p><p>It’s also not exactly clear who makes these. Most of the listings on DigiKey point to either AdaFruit or SparkFun — both lovely companies I am grateful for, serving the maker-sphere, but neither company is exactly ‘Texas Instruments’ with respect to making and selling semiconductors in truly massive quantities.</p><p>The real mind-blower of this corner of the market is the <a href="https://www.adafruit.com/product/1938">Adafruit through-hole NeoPixels</a>. These things are still pushing a buck a piece even at quantities of 500. It’s bonkers! I would love to buy a couple hundred of these to just have on hand, but I’m not going to drop a buck a piece for the privilege. Sorry, Adafruit.</p><h3>Now What?</h3><p>So, what do you think? Have I found some secret ‘hole’ in the market whereby I can get these things well below market price if I’m willing to sit here for a half hour with a heat gun desoldering them from their circuit boards? Or was a just getting ripped off buying bare parts at qty 100?</p><p>What’s been your favorite ‘hack’ for saving money doing maker projects?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6fc6b3530988" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Good Vibes, Tiny ‘puters, PCBs, and 3D Printing]]></title>
            <link>https://medium.com/@ipmcc/good-vibes-tiny-puters-pcbs-and-3d-printing-c5c3f704205d?source=rss-3567854fbab1------2</link>
            <guid isPermaLink="false">https://medium.com/p/c5c3f704205d</guid>
            <dc:creator><![CDATA[Ian McCullough]]></dc:creator>
            <pubDate>Mon, 30 Mar 2026 14:15:09 GMT</pubDate>
            <atom:updated>2026-05-07T13:33:01.689Z</atom:updated>
            <content:encoded><![CDATA[<p>I’ve recently been working on a fun little project involving AI, printed circuit boards (PCBs), and micro-manufacturing with 3D printers. But first let me rewind for a second…</p><h3>Background</h3><p>I wrote an <a href="https://medium.com/@ipmcc/working-with-very-small-computers-1995a4dc0f7e">article</a> a while back about working with the RP2350–Zero and RP2040-Zero embedded single board computers. One of the folks who read that article reached out to me, and we rapped for a bit about a project that he’s working on, and how he had been considering making custom PCBs, or rather having a fabrication company make PCBs <em>for him</em>. Personally, I no longer make any attempts to fabricate PCBs myself, and I advise everyone else to skip that exercise as well. (See <a href="https://medium.com/@ipmcc/the-one-about-making-pcbs-12ef5973323c">another one of my pieces</a> for a deep dive on that.) I asked for his blessing, received it, and began to work on a very similar project, while leveraging AI, 3D printing, and custom circuit boards from <a href="https://www.pcbway.com/">PCBWay</a>. Maybe if he likes what I come up with, it’ll create more chances for me and this fellow to work together.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3Zts5YyUZ7DPf_biaSsqiA.jpeg" /><figcaption>PCBWay did a bang-up job on this project. (Source: <a href="https://pcbway.com/">pcbway.com</a>)</figcaption></figure><p>Speaking of working together, I also have a local friend who is very, <em>very</em> into AI-assisted coding — so-called ‘<a href="https://en.wikipedia.org/wiki/Vibe_coding">vibe-coding</a>’. He’s a great friend to have (in many ways, but especially…) because he tries them <em>all</em>, then tells me which ones are good, and which ones are less good, so I don’t feel like I’m wasting my time.</p><p>I don’t want to harp on this, but I do want to come completely clean: Until very recently, I was a <strong><em>deep-seated</em></strong> AI skeptic, but at this particular moment, I had this challenge from a reader, and I had my other friend constantly yakking in my ear about how great <a href="https://anthropic.ai/">Anthropic’s</a> <a href="https://code.claude.com/docs/en/overview">Claude Code</a> is, and I love a good 3D Printing challenge, so I decided to give it a try!</p><h3>The Project</h3><p>The basic idea here is to make a <strong><em>round</em></strong> <em>clock face</em> out of PCBs printed in arc segments. I want the clock to be big enough so as to be meaningful as a wall clock — not so small that it’s a desk clock, but not any bigger than it would need to be to read from across a room. My reader’s clock project was 1m/~40in wide, and that, to me, is beyond ‘wall clock’ and into the realm of ‘electro-mechanical art’. I decided to try for more of a standard-size wall clock. I decided on about 400mm/16in in diameter. And yes, you <em>can</em> get fab companies to make you a circuit board that is a circle with a diameter of 1m, but I’m sorry to say, no matter where you shop, you’re not doing to like the price. I priced it out, and we’re talking well into US$1,000+ per-board territory.</p><p>Even at my chosen size of 400mm, printing a single board would <em>still</em> have been prohibitively expensive, so I decided I’d do it in parts and come up with a way to reliably put them all together.</p><p>Because this project brings together so many different disciplines and because I’m lazy, I decided to “<a href="https://en.wikipedia.org/wiki/Fast-track_construction">fast track</a>” it, in other words, to start building before the design was complete. I started right in with…</p><h3>PCB Design using KiCad</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*cdZpkjGIHYKiqurkmroxIQ.png" /><figcaption>My favorite tool for PCB design. (Source: <a href="https://upload.wikimedia.org/wikipedia/commons/thumb/5/59/KiCad-Logo.svg/960px-KiCad-Logo.svg.png">Wikipedia</a>)</figcaption></figure><p><a href="https://www.kicad.org/">KiCad</a> is many things to many people, and frankly, you could do a <strong>lot</strong> worse. I’ve seen people try to design PCBs using Keynote/PowerPoint slides. That said, one particular thing it’s not really good at is giving you the best tools you might want to create precise, repeatable designs, with irregular features (i.e. not everything is arranged at 0°, 45° or 90°). In most cases, this isn’t a big deal. If you look at my previous article about designing a bridge board for these RPXXXX-Zero modules, I did that without any of the procedural tooling or AI assistance that I’m using in this jaunt. Once I got into <em>this</em> project, it became clear to me that it really demands precise, repeatable, ideally procedural, tools and I felt like I couldn’t do that by hand. So, let’s get into what I <strong>did</strong> do!</p><h3>3D Design using Autodesk Fusion</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9kUc4m-BjVQruTOMh-lzBg.jpeg" /><figcaption>Autodesk Fusion — no more 360 (Source: Google Image Search)</figcaption></figure><p>When CAD-ding, my ‘weapon of choice’ for the last year or two has been Autodesk Fusion. (In case there’s any confusion, it used to be called “Fusion 360”, now it’s just “Fusion”, but it’s the same product.)</p><p>It’s quite a capable package, although it can sometimes be a little persnickety. It’s not the best I’ve ever used, but it’s <strong>far</strong> from the worst, and the price was right: For hobbyist/non-commercial projects like these, it’s free. There’s a rich community in the forums to answer your questions, and there’s a broad library of YouTube videos and blog pieces that will show you how to do almost anything you can dream up.</p><p>It’s probably worth mentioning: Fusion has its own PCB design package. I haven’t used it much, and frankly, I haven’t heard particularly good things about it, so I wasn’t itching to use it. I already know how to use KiCad, and I’m doing something pretty out-of-the-ordinary here, so I figured I’d go with what I know.</p><h3>So, to recap:</h3><ul><li>I want to make a <strong>wall clock</strong>.</li><li>I’m going to make it roughly <strong>16in/400mm in diameter</strong>.</li><li>I want it to be <strong>round</strong>.</li><li>I want it to have 60 <strong>NeoPixel LEDs</strong> around the outside where the second hand on a normal clock might tick.</li><li>I decided to make the circle out of <strong>ten</strong> <strong>arc segments</strong>, each spanning 36°, and having <strong>six</strong> “tick LEDs”.</li><li>In the spirit of ‘<strong>fast-track</strong>’ development, I’m going to wait until later to figure out how the hour and minute hands are going to work.</li><li>Similarly, I’m not going to think about what I’ll use to drive it, but you can confidently wager that it’ll be either an <strong>RP2XXX-Zero</strong> or an <strong>ESP32</strong> (tempting for WiFi), or both.</li><li>I will <strong>3D print</strong> the frame and housing on my <a href="https://us.store.bambulab.com/products/p1s">Bambu P1S</a>, but there’s not much point in doing that until I know I have working PCBs, and until I understand how I’ll mount the PCBs in the 3D printed parts.</li></ul><p>If it wasn’t obvious, I want to front-load the PCB design, because that’s what has the longest lead times.</p><h3>Arriving at My First Real Strategy</h3><p>I started out working by hand in KiCad. I’ve done it a bunch before. It’s produced really great results. For my 16-inch wall clock, I need the pieces to fit together <em>precisely</em>. I was thinking roughly about 0.15mm of ‘slop’ between the pieces as they fit together. I also decided to force myself into this by making the segments fit together like jigsaw puzzle pieces. I figured that would be a good way to crack the whip on myself so I always know that it has to be <em>exactly right</em>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2XTM8hV31PL5CWwUMo0ubw.jpeg" /><figcaption>Puzzle pieces… like this! (Source: me)</figcaption></figure><p>KiCad is a powerful package, but it is <strong>not</strong> Adobe Illustrator, and it is <strong>not</strong> Autodesk Fusion. To be fair, it’s not <em>trying</em> to be either of those products. It’s just that I have experience using those kinds of products, and virtually every day that I work in KiCad, I miss features from CAD and vector illustration programs, and what I’m left with are inconsistent parts. Remember, this project involves <em>ten</em> different circuit boards, that all have to fit together well — even small errors are going to accumulate when you’re trying to make a ten-piece loop. With that in mind, my sense is that certainly, those 10 parts have to be 100% consistent between themselves. I know from experience that KiCad is not really great at keeping multiple PCBs in sync. In fact, it doesn’t even really try.</p><h3>Enter Claude…</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XVryBmvhUtHHEFevcYADOg.jpeg" /><figcaption>New best friend? New best employee? Time will tell, but off to a good start. (Source: <a href="http://claude.ai/">claude.ai</a>)</figcaption></figure><p>My AI-enthusiast buddy told me that Claude is the front runner supporting most of his technical work, so when thinking about this project, I thought, “Why not?!” Unfortunately, I didn’t keep transcripts of all my interactions with Claude, so this will be a high-level overview.</p><p>I started by giving Claude one of the very-rough parts I had made by hand. I explained the project in plain English, and Claude seemed to understand fairly well. Claude also claimed to be fluent in KiCad’s API. I explained the project, in plain English, and it professed to understand. Before long, I started over working with <em>just </em>Claude, and abandoned trying to use my hand-built board. And so, we started iterating. I’d say things like, ‘Orient these labels such that the text baseline is tangential to the arc of the clock face. Orient these through-hole parts such that their pins are aligned radially.” We were off to the races!</p><p>In about a half a day I had a baseline implementation that would <em>probably</em> produce a 10 piece ring that could be manufactured, but of course, I had to go and get fancy. 😀 Since these ten PCB segments form a circle, and because I’m into aesthetics, I wanted as many things as possible in this design to be either radial, tangential, or circumferential. I already knew that, out-of-the-box, KiCad didn’t have the tools I would need to make this task easy to do by hand, but I had hope that with Claude’s help, I could probably come up with something viable using the API.</p><p>Claude confidently declared that it was up to the challenge, and that it has in-depth knowledge of KiCad’s Python API that it would use to address this challenge. With my head held high, my typin’ fingers ready, and my voice dictation setup ready to go. (As an aside: I sometimes prefer to talk to Claude verbally, other times I like to type to it. Generally, if the thought is fully formed in my head, I want to speak it, but if I’m still figuring it out as I go, I prefer to type it.)</p><h3>The Basic Low-Level Electronic Design</h3><p>At the risk of oversimplifying it, <a href="https://www.adafruit.com/category/168">NeoPixels</a> are individually addressable daisy-chain-able RGB LEDs. They each have 4 pins. V+ and GND, and then DATA_IN and DATA_OUT. My ring of PCBs would need three inputs and three outputs, and then would need to daisy-chain the 6 LEDs within the board, running traces from LED N’s DOUT pin to LED N+1’s DIN pin and so-on.</p><p>Then the DOUT of the last LED on an arc segment goes to the DIN of the next arc segment (through a set of <strong>castellated </strong>joints). It’s blindingly simple electrically, but a total PITA in terms of the irregular PCBs.</p><p>This daisy-chain of NeoPixels will be driven by a microcontroller. Given the ‘low frequency’ of this application, I could probably just ‘bit-bang’ a GPIO pin, but there are plenty of NeoPixel libraries, so I’ll just use one of those. Either way, I’ll decide later. As mentioned, I’ll be using one of these (probably the ESP32):</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4aggKA5atlZ1469PYJFg-g.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/409/1*ps4VQDcZ4TFSCDqf-toEWg.jpeg" /><figcaption>RP2040-Zero (Source: Amazon) and a ESP32 module I have on hand (Source: <a href="https://www.amazon.com/dp/B0CNYK7WT2">Amazon</a>)</figcaption></figure><h3>The PCB Design</h3><p>We’ve gone far enough without some truly nerdy graphics, so behold this — Arc #1 of my clock rim:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*K6Buc5wk2hzcLGe7jt9a0A.jpeg" /><figcaption>One Arc board. (Source: <a href="https://www.ipmcc.com/">me</a>)</figcaption></figure><p>And yes, that is <strong><em>all</em> output by Claude</strong>, using the KiCad API, with no human post-processing at all. When it came time to send the boards off to PCBWay, I didn’t even do <em>that</em> by hand; I just had Claude write scripts that would generate the production outputs, and put them in a ZIP file, ready to be uploaded. Oh yeah, and this whole time Claude isn’t just ‘giving a man a fish’, it’s ‘writing scripts into a Git repo so the man can fish any time he wants.’ Claude is also a great resource for figuring out best practices. I’d explain what I was trying to do, and with which tools, and Claude would tell me exactly how to store this project in Git, what should be in my `.gitignore`, and so on.</p><h3>The ‘Way’ to PCBs</h3><p>After completing my design, I had Claude generate the production outputs, (Gerbers, drill files, etc) sent them off to PCBWay, and worked with their customer service folks to make sure that my design was proper for fabrication. They were <em>very</em> helpful and detail-oriented. I’ve used other providers in the past, but the PCBWay folks really went above and beyond in letting me know things like, ‘this edge cut impinges upon this trace’, and so forth.</p><p>Living in the U.S., and having done this a few different times, I know that it always costs an arm and a leg to get boards back from East Asia — sometimes the shipping even costs more than the boards, and the ongoing US tariff shenanigans don’t help either. With that in mind, even though the boards are very affordable, you really want to <strong>get it right the first time</strong>. Ideally, if I could fabricate these boards in my garage (<a href="https://medium.com/@ipmcc/the-one-about-making-pcbs-12ef5973323c">I cannot</a>), I could probably iterate through three or four different prototypes per day and eventually find my way to a working design. But when I’m going to send it off to East Asia and wait ~10 days to get it back, only to find out that I made some horrible mistake? That’s not really what I want.</p><p>Please know: These are not your garden variety PCBs like my <a href="https://medium.com/@ipmcc/working-with-very-small-computers-1995a4dc0f7e">RPXXXX-Zero prototyping adapter</a>. Just about everything on them is curved: the outline of the boards, all the traces, most of the text, etc. Furthermore, I didn’t want to do ten runs of 5 boards a piece, so I had Claude panelize them for me. It took considerable effort. Claude was confused by the curved boards at first, but in time, we were able to converge on something that would allow the creation of all ten arc pieces in a single board, which meant that with a minimum quantity of 5 boards, I was now getting 5 clocks’ worth of boards. This would come to be important, because not <em>everything </em>went perfectly right the first time. Check it out:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4-KtB_zxS4lZgyNr2RFbhQ.jpeg" /><figcaption>Ten small PCBs arranged as one! (Source: <a href="https://www.ipmcc.com/">me</a>)</figcaption></figure><p>I really wanted to test PCBWay’s capability to produce unusual PCBs like these. They did not disappoint!</p><p>Quick aside: Those little nubbins that hold the boards together in a panel are called “mouse bites”. These mouse bites are where Claude had some of the most trouble with the PCB generation, but we got through it with a lot of hand-holding, and many refining prompts. I would estimate it took <strong>six hours</strong> of back and forth with Claude to get the panelization just right, but by that point, I was all-in on the all-AI part of this project, so I slogged through. FWIW, one of the things I like to do when working with Claude on something that doesn’t work out of the gate is, when we get to an acceptable result, I ask Claude:</p><blockquote>Wow, that took a lot of time! Could you please write a text file explaining the prompts I <strong>could’ve</strong> given you that would’ve gotten us to this end faster? Put it in the docs/ folder.</blockquote><p>Its replies to this kind of query are often <em>very </em>informative, and they help me learn how to work more effectively with Claude in the future.</p><h3>What I Got Back</h3><p>My friends could tell you: I was like a kid in a candy store when I got these boards back the other day! I was positively <strong>giddy</strong>! But I won’t make you wait in suspense any longer; Here‘s what PCBWay sent me:</p><figure><img alt="5 Panelized Beauties!" src="https://cdn-images-1.medium.com/max/1024/1*NXTBTLc1Uw299SQmGLQM1A.jpeg" /><figcaption>5 Panelized Beauties! (Source: me)</figcaption></figure><p>I just couldn’t wait, so I broke one of the panels out into its constituent boards, and with my proverbial fingers crossed, I started to put together the puzzle pieces… and now we come ‘full circle’ (LOL):</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_IoRhyTHlCFnr6vpmGA_lQ.jpeg" /><figcaption>They went together perfectly! On the first try! (Source: me)</figcaption></figure><p>I couldn’t be happier with how these came out. They really are a thing of beauty. I owe the perfection of the design largely to Claude, but I owe the perfection of the execution entirely to PCBWay. These boards are downright <em>beautiful</em>. They were also my first experience using a white solder mask and black silkscreen, but since the point of these is to display with colored LEDs, I figured having a neutral solder mask and silkscreen would be appropriate.</p><h3>The Body of the Clock</h3><p>I plan to give this process a full rundown in a later article, but the nickel tour is this: Claude is pretty good at PCB design… it is <em>less </em>good with non-trivial 3D part design. That said it is just as predictable and deterministic which was a big part of the reason I chose to use the AI. Claude proclaimed that it was also fully versed in the <a href="https://www.autodesk.com/products/fusion-360">Autodesk Fusion</a> API (and that’s a good thing because their API is positively <em>gargantuan</em> in scope.)</p><p>Claude did fine with basic sketches and extrusions, but it became <em>far</em> more difficult when parts got complicated or irregularly shaped. Sometimes I had to get downright <em>remedial</em> with it. Under the covers, what it was doing was taking my instructions, and translating them into Python programs based on the library <a href="https://github.com/gumyr/build123d">build123d</a>. From there, it would export those parts from the Python code into STEP files, which can then be imported (using the API, of course) into Fusion.</p><h3>Mistakes Were Made…</h3><p>Not everything went perfectly. Which is a demure way to say:<strong> I</strong> made some mistakes. Let me share, so hopefully you don’t have to learn the hard way like I did.</p><h3><strong>Lesson #1: Component Choice, or ‘Size Matters’</strong></h3><p>NeoPixels come in several forms, but the most common one is the <a href="https://www.adafruit.com/product/1655">5x5mm SMD variant</a>. There are also <a href="https://www.adafruit.com/product/2686">3.5x3.5mm versions</a>, and the <a href="https://www.adafruit.com/product/4684">2x2mm SMD one</a> that I used. There’s even <a href="https://www.adafruit.com/product/1938">a through-hole version</a>! (Although at a $5 each I’m pretty sure the through-hole ones aren’t flying off the shelves.)</p><p>As I was planning this out, I cut out a 5x5mm piece of paper, went to my soldering station and nudged things around and thought, “…this is totally fine. I’d better <em>go smaller</em> and design for the 2x2mm ones.” Hilarity ensues.</p><p>I’ve been working with embedded electronics since I was a teenager. When I was 16, at my <a href="https://www.onsetcomp.com/">workplace</a>, after hours, with access to a nice stereoscopic microscope, high-end, SMT-specific soldering and<em> desoldering </em>stations with all manner of fancy tips, a truly staggering array of tweezers, and the sharp eyes and steady hands of youth. I firmly believe that, in <em>that</em> context, at <em>that</em> age, these 2x2mm parts would’ve been quite tractable for me. However, now that I’m pushing 50, with an above-average, but neither high-end, nor SMT-focused soldering station, a simple elementary-school-style magnifying glass, my failing eyesight, and my shaky hands, it was quite tedious. If I were going to do it over again, I would either go with the 5x5mm version or maybe even the through-hole version. (Although probably not the through-hole, just based on price.)</p><p>At the end of the day, it’s been difficult, but not impossible, to work with these 2x2mm parts. I just wish they weren’t so darn fiddly… or that I had a stencil… or that I had a pick &amp; place machine.</p><p>In a way, I’ve already learned a related lesson in another context. In my travels with 3D printing, I learned quite quickly (through trial and error) that any fasteners smaller than an M3 are probably not going to work. M3 will probably work for most things involved in 3D printing (because if the force involved would break a steel M3 screw, the 3D printed plastic is probably going to give up first) so I keep a bunch of those on hand, but if you have room in your design to use M4 or M5, <strong>you probably should</strong>. I should’ve thought of this lesson when I decided on 2x2mm SMDs, but here I am, making the most of what I have.</p><h3>Lesson 1B: Quantity Matters</h3><p>When building electronics, one is always well-advised to have extra parts on hand. I ordered 100 of these 2x2mm LEDs, initially thinking: ‘Oh, this’ll be enough for most of <em>another</em> clock!’ Well, having assembled and debugged all ten arc segments now, I’m sad to admit: I have used <strong>all</strong> of the hundred LEDs. There are even a few whose brightness doesn’t seem uniform with the rest of them, and I’d love to replace them, but I’ve used all the LEDs I have. I would order more, but having realized that these 2x2mm LEDs are best left for the youngsters, I don’t want to be stocking dozens of these things. It turns out you can cook tiny parts to their death pretty easily with a 400°F soldering iron.</p><h3>Lesson #2: Stencils</h3><p>This is an <a href="https://en.wikipedia.org/wiki/Stencil_printing">SMT PCB stencil</a>. There are many like it, and this one is <strong><em>not</em></strong> mine, nor can I show you a picture of mine because I <em>failed to order</em> it.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/560/1*Qw51OySsNyrygI3QjH4lpQ.jpeg" /><figcaption>An SMT Stencil — Not mine (Source: <a href="https://www.mokotechnology.com/smt-stencil/">MOKO</a>)</figcaption></figure><p>In the days of yore, I worked with stencils, <a href="https://en.wikipedia.org/wiki/Pick-and-place_machine">pick &amp; place machines</a>, and a ‘real’ <a href="https://en.wikipedia.org/wiki/Reflow_oven">reflow oven</a> (that took up most of a room), but since I was probably going to be doing this by hand no matter what, I didn’t bother to order the stencils that allow you to uniformly deposit solder paste onto your board. Don’t get me wrong: PCBWay does provide these, but it’s a few extra bucks, and I kinda thought, “I got this! Why bother?” After applying solder paste using a stencil (or ‘dead reckoning’ with a syringe), you put the parts onto their footprints, and then put the whole thing into a reflow oven. The reflow oven then heats up and cools the PCB in accordance with a specific curve that is apparently quite important. More on this below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*55aS8U6L-LlT0ut-hOqoFg.png" /><figcaption>The reflow heat curve. (Source: <a href="https://en.wikipedia.org/wiki/Reflow_oven#/media/File:RSS_Components_of_a_Profile1.svg">Zithan</a>)</figcaption></figure><p>Trying to use a syringe to deposit uniform, tiny amounts of solder paste onto pads that are 0.7x0.7mm? Without leaving glorps of paste all over the place? Without creating bridges between pads? It’s difficult, and <strong>not</strong> particularly enjoyable. Possible? Yes. Fun? No.</p><p>Learn from my mistake: If you’re going to try to do SMT PCBs, just get the stencil. Even if you never end up using it, it’s just a few extra bucks, and trust me: there is <strong>no</strong> Olympic medal for hand-placing SMD parts on to glorpy, hand-pasted PCBs. Although, I must confess, that would be hilarious to watch.</p><h3>Lesson #3: Reflow</h3><p>Another question I asked AI was, “Is it possible to use a home cooking oven as a reflow oven?” The answer was that <strong>yes</strong>, strictly speaking, it <em>is</em> possible, but that it is very fiddly, and at least in my case, downright destructive.</p><p>The advice I got was to preheat the oven to 350°F, with your board in it — this is meant to mimic the ramp up and preheat phases of the curve. Then to jack it up to 450°F, watch for the paste to melt — it’s cool, the solder turns from grey to silver, and the parts just magically ‘sit down’ onto the pads — and then immediately shut the oven off. Sounds simple, right? Well, just try not to get distracted by the TV, or a friend, or a bathroom trip, or anything else, while it’s in there. I would guess the board was in the oven for ~3 minutes total. It came out looking like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vwKCIX-JqM-iA3-c_SCIbw.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yLysfv-Du6BRyIKAk-b0Uw.jpeg" /><figcaption>I mean… the parts are soldered on… sort of? (Source: me)</figcaption></figure><p>I’m seeing some major problems here. First, my beautiful, pristine, white solder mask was completely burned/sublimated away. The second problem, and I have to admit that the AI warned me about this: You can clearly see that the heat in a home oven is uneven: some places on the board are obviously much more ‘scorched’ than others. Although I did screw it up, and I admit that, the comments about home cooking ovens being uneven heat sources are obviously true.</p><p>Next I tried to solder the LEDs by hand. This was a complete disaster. One arc segment took me about 90 minutes. Did I (eventually) get a working segment board out of it? Yes. Do I recommend this process? <strong>Absolutely not</strong>!</p><p>For my next attempt, I used a heat gun. I wasn’t particularly optimistic, and even Claude told me that this wasn’t the best idea. I wouldn’t want to reflow a complex board this way (and it’s absolutely <strong>not</strong> viable for things like BGA (ball grid array) chips) but since I was doing one LED at a time, applying paste by hand, setting the LED in the paste, etc, the heat gun actually worked <strong>really, really well</strong>. A Satinsky/Kelly/Hemostat clamp helps too, so you don’t melt the skin off your hands.</p><h3>Recap: Lessons Learned</h3><ul><li>Get the stencil. Even if you think you won’t use it. It’s a few extra bucks, but having it could’ve saved me a great deal of grief and fiddly-ness.</li><li>Don’t try to reflow in your kitchen oven. AI tells me that it’s possible. Bloggers tell me that it’s possible. But I tried it, and it’s simply not something I will try again. Seeing my beautiful board come out looking like it sat too long in the tanning bed was disappointing (not to mention a waste of 6 LEDs.)</li><li>Bigger is better. There’s something in me that always wants to make things smaller, but my life was so much better once I set a ground rule of “M3 or bigger”.</li><li>Test fixtures. When you’re not confident in your assembly process, having a way to test boards is really, really important. Only you will know what’s appropriate for you and your application, but making a test fixture helped me validate each board I populated. It’s been worth its weight in gold.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uDNzF_1N-lBhy8EiWqTaYQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Pd7XOpKu3rd5R6V2pQBSOA.jpeg" /><figcaption>My test fixture! ‘Pogo pins’ are quite useful. (Source: me)</figcaption></figure><ul><li>When hand-solder-pasting pads without a stencil, <em>less is more</em>. More means bridges. Use less.</li><li>When building a large circuit made of daisy-chained devices — especially if you’re doing it by hand — go one at a time. Only add the second when the first words. Only add the third when the first and second work. And so on. I didn’t do this at first, and I paid the price in having to desolder and re-solder <strong>many</strong> LEDs before getting things to work right.</li><li>Try not to hit yourself with a hammer… this should be self-evident, but here I am. I had to ‘drill &amp; drain’ this fingernail, which is never pleasant. I could also have just taken Advil, and waited for it to fall off, but I’m impatient, and I didn’t want to lose a nail. Groan.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*N4X2QU3CipXjmFnJReQWEQ.jpeg" /><figcaption>Try not to do this. (Source: me)</figcaption></figure><h3>In Closing</h3><p>Perhaps obviously, since there’s not yet a YouTube link to a beautiful, completed, working, time-keeping, WiFi-enabled clock, I still have a <em>lot</em> left to do on this project, and I expect to follow up on this once I have more working. But every project like this starts with an idea, then typically a breadboard prototype, then a first prototype, a bunch of other BS, and then… finally, you <em>might</em> end up with a functioning product! Learn from my mistakes, get help from experts, bite off digestible chunks, but by all means: If you have a product in your head? You can make it! It’s just a question of how much you’re willing to learn along the way. For me, learning is joy. I hope it is for you too.</p><p>If you have any questions, please feel free to email me: vibepcb-at-ipmcc-dot-com. The only thing better than learning is sharing your learnings with someone else. 😁</p><h3>And because this is <em>very</em> important to me…</h3><blockquote>I want to be 100% transparent: This article references some commercial products and services that were provided to me free of charge. I’m quite happy with the results, I’m happy to sing their praises, and I would happily pay for these products and services myself in the future. Also, please know that if I were <strong>not</strong> happy I would also <strong>say so</strong>… I’m just kinda that way. This is <strong>not a commercial</strong>, but I do feel it’s important to be clear about these things. If you have any questions about this, feel free to contact me at vibepcb-at-ipmcc-dot-com.</blockquote><p><strong>PS: </strong>Special thanks to @merichar for coming through for me with some desoldering wick, on a Sunday, while she was on the road traveling, and when I needed it the most! Thanks, Meg!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c5c3f704205d" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Working With Very Small Computers]]></title>
            <link>https://medium.com/@ipmcc/working-with-very-small-computers-1995a4dc0f7e?source=rss-3567854fbab1------2</link>
            <guid isPermaLink="false">https://medium.com/p/1995a4dc0f7e</guid>
            <category><![CDATA[embedded-systems]]></category>
            <category><![CDATA[electronics]]></category>
            <dc:creator><![CDATA[Ian McCullough]]></dc:creator>
            <pubDate>Tue, 03 Feb 2026 16:30:06 GMT</pubDate>
            <atom:updated>2026-02-09T12:14:06.554Z</atom:updated>
            <content:encoded><![CDATA[<p>Lately, I’ve been playing around a bunch with these new boards — well, new to me anyway — called RP2040-Zero and RP2350-Zero. The company that seems to have officially come up with these boards in the first place is called <a href="https://www.waveshare.com/">Waveshare</a>. I have gotten a couple of first-party Waveshare branded boards, and they’re lovely, but I’ve also gotten a bunch of cheap knockoffs on Amazon. My most recent haul were these: <a href="https://www.amazon.com/dp/B0DXL12W59">RP2040-Zero</a> &amp; <a href="https://www.amazon.com/dp/B0F53R76GJ">RP2350-Zero</a>. They’re essentially stripped down, aggressively miniaturized versions of the Raspberry Pi Pico and Raspberry Pi Pico 2, respectively.</p><p>These boards are astoundingly inexpensive and quite able to handle a large variety of computing tasks and embedded tasks. For instance, I recently ordered a <strong>six-pack</strong> of the RP2040 version, and it was $16. For all six. The 2350 version was a little more. I got a three-pack of those for $14. I’ve never had one fail, except by my own stupidity, so I’ve been running with the cheap knock-offs for now.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7faoTURkNryG9V7nXEzRkg.png" /><figcaption>Waveshare RP2040-Zero</figcaption></figure><p>The two boards are the exact same size, roughly 24mm x 18mm. They’re also reasonably comparable in terms of capabilities. The 2040 is a dual-core ARM Cortex-M0+, with ~256K of SRAM and 2MB of flash storage. The 2350 is slightly faster and has a bunch more stuff, such as twice the SRAM and twice the Flash, and some sort of RISC co-processor — I haven’t gotten to dig into that part yet.</p><p>You can program these things in standard C/C++, MicroPython, or Arduino’s version of C/C++. Getting off the ground is pretty easy if you use a lightweight toolchain like <a href="https://micropython.org/">MicroPython</a>/<a href="https://thonny.org/">Thonny</a>, and slightly more effort, if you want to use <a href="https://code.visualstudio.com/">VSCode</a>, or write in C/C++.</p><p>From a development standpoint, debugging support is… not great. Which is to say, there really isn’t any debugging support beyond <em>printf</em> and flashing an LED. I’ve seen indications that the truly motivated can cobble together some user-space debugging ability for MicroPython, but the methods I found looked really onerous. But for three bucks, what can I say?!</p><p>Now, I need to talk about the physical package. As mentioned, this is roughly 24mm x 18mm, and it has pins on three sides and a USB-C connector on the fourth side. There are nine pins each along the 24 mm sides, and five pins across the back. Which brings me to my first recommendation, which is if you’re going to buy these boards, buy them <em>without</em> the pins already soldered in.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_uiCSI7U-Ry3oUf8lrvx3A.jpeg" /><figcaption>Standard breadboard for prototyping</figcaption></figure><p>When you have the pins soldered in on all three sides, you can no longer use these modules on a standard breadboard for prototyping. Even if you don’t solder in the back pins (on the short side), it is still somewhat awkward for breadboarding. Its width means that in trying to center it over the gutter of a breadboard, it consumes three columns on one side of the gutter, and two columns on the other. It’s not symmetric. On the side with two columns left, you’re working with a limited number of connections. This is what I’m talking about:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jiwxQuOBQ1Ci_gCAP2ZnhA.jpeg" /><figcaption>Off center. Meh.</figcaption></figure><p>Three columns on the left, two on the right. It’s awkward, and it turns your prototype into a total rat’s nest if you happen to need more than two connections to any given pin on one side.</p><p>So, we started out with: (2 * 9 + 5) = 23 pins. Then we had to give up <em>easy</em> access to five of them so we could plug it into a breadboard at all. I ginned up a little cable that, if the five pins at the back are installed the opposite way (i.e. pins up, as in the picture), can draw those pins out to other rows on the breadboard using jumpers:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*JkzBjZj6BU5yzbnfi3IRmw.jpeg" /><figcaption>My little jumper device.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1010/1*rlO_73VxOSfcFAW3Zd8h3Q.jpeg" /><figcaption>…and then in its natural habitat.</figcaption></figure><p>But that’s only the beginning! Behold the bottom of the RP2040-Zero:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/589/1*PdHmAOojTdY8AIp_md3cRw.png" /><figcaption>Bottom side of RP2040-Zero</figcaption></figure><p>We can see that in addition to the 23 pins with through-hole spots, there are ten more surface-mount pads on the bottom, and it turns out that these are nine more GPIO pins and another ground pad. The documentation vaguely suggests that if you <em>really need</em> these pins, then you can endeavor to solder wires directly onto those pads. Interesting, but also somewhat of a challenge being roughly <strong>half-pitch</strong> relative to normal 0.1in/2.54mm through-hole spacing. I was not excited about this task.</p><p>You may also notice that the main pins are ‘castellated’. Castellated PCBs (or “castellated holes” more precisely) feature plated, half-cut through-holes along the board edge, creating semicircular, solderable, vertical connections for mounting these modules directly onto another board. In other words, you can solder the board to another board without using through-hole pins. Think of it like the entire board is, itself, capable of being surface-mounted. Neat! I’d never worked with a castellated board before.</p><p>With my 18 pins in the breadboard and my five-pin extension jumpers to handle the back pins, I got to work. I immediately ran up against what feels like it was almost inevitable: I needed <strong>one</strong> more GPIO than was led out to the through-hole pins. I thought about multiplexers, shift-registers, and other hacks to expand the number of GPIO pins, but in the end, the novelty of its castellation, and the presence of 9 more GPIO pads on the back… it just got the better of me. Similarly, I was extremely frustrated trying to prototype with these on a breadboard because of the 3 row/2 row issue mentioned above. I know it’s a little thing, but it always felt “off” to me.</p><p>With that in mind, I set out to create some sort of solution that would make these boards easier to breadboard, and, hopefully, could provide some access to these back SMT pads. The first part is a breadboard adapter. It looks like this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*t0n1CUkQ4wNQUXcwpuLd8g.jpeg" /><figcaption>The edge-card adapter</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GR4TlWN7_mMFgmiwATnDCA.jpeg" /><figcaption>One column on each side! Beautiful!</figcaption></figure><p>Looking end-on, you can see that this adapter only occupies one column on either side of the breadboard gutter. There is a <a href="https://www.amazon.com/dp/B0BPP6HK45">36-position edge-card socket</a> (think like old-school Nintendo cartridges) mounted to a trivial PCB that routes the contacts of the edge connector out to through-hole pins. Then I made this board that plugs into the edge-card slot:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4eoSV6LtZB7i9zrL2k0hEw.png" /><figcaption>3D Rendering of the board.</figcaption></figure><p>OK, so these adapters solve the ‘all 23 through-hole pins to the breadboard’ problem. Great! Here’s what a real one looks like from the front:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*y3iKGIObQMF6OvHWtV2C8Q.jpeg" /><figcaption>Purdy!</figcaption></figure><p>The castellation mounting was a new experience for me, but as long as I used enough flux, it just went right into the joint through the magic of capillary action. When I was younger, I was quite deft with a soldering iron. As I’ve gotten older, between my dwindling eyesight, my somewhat less steady hands, my lack of a microscope and a good work-holding setup, etc. life has conspired to make fine pitch soldering more difficult to do well. I’m not proud of my soldering job here, but it’s functional.</p><p>This edge-connector board also creates an opportunity to draw out those ten pads on the back of the board. I looked at doing some sort of spring-tension daughter card that would push up against the pads of a module that was attached to the edge card (with castellation and no pins). Long term, I’d love to revisit that, but that approach swallowed a day and a half of my effort, and I couldn’t really get anything that I could work with. Those half-pitch SMT pads are just… well they’re freakin’ tiny. But, slowly, deliberately, with much profanity, and a crappy magnifying glass I happened to have lying around, I produced this abomination:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*H_9knWC9dZ87JXH-JplMcA.jpeg" /><figcaption>Not my best work.</figcaption></figure><p>This is wire-wrap wire, running from each of those bottom pads to the 10 through-holes at the left, which are, in turn, led out to pins on the edge connector.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*YNSXwomR5Y1xBV1TkkzhrQ.jpeg" /><figcaption>Put it all together…</figcaption></figure><p>The reason I wanted to write this up is that, when I first saw these boards, and their 10 backside surface-mount tabs, I was frustrated. I was asking myself, “Why would they even put those pads there?” That being said, I <em>was</em> able to do it. It took me about 1h15m to do one board, so I’m not likely to make a living building these, but I have faith that there is some possible solution. This was my process:</p><ul><li>Verify that the module works by hooking it up to my computer via USB-C. Yay. Works.</li><li>Solder wire-wrap wire (30AWG) to each pad.</li><li>Have almost every single solder joint fail, and have to be redone.</li><li>At this point, I have an RP2350-Zero board with 10 <em>extremely delicate</em> wires coming off the back of it.</li><li>Then I super-glued the wires to the back of the RP2350-Zero so that moving the wires around would no longer put stress on the tiny solder joints.</li><li>Test all 10 wires with a volt meter/continuity check. Verified everything is connected. (These 10 solder joints all look incredibly janky.)</li><li>Mount the RP2350-Zero board to the adapter board with castellations. Fun!</li><li>Run the wires from the SMT pads up to the 10 through-hole pads on the other side of the board.</li><li>Solder them.</li><li>Coat <em>that</em> whole area with super-glue too, for good measure.</li><li>Test all ten wires again, this time against the full path from the RP2350-Zero board to the breadboard. Verified everything is connected.</li><li>Off to the races!</li></ul><p>Now… I know some of you are thinking, “Wait. So you bought these tiny RPi Zero boards instead of the standard RPi Pico/Pico 2, but then immediately turned around and mounted them to a much bigger PCB? Why?” TBH, I don’t really know why. For whatever reason, I fell in love with the tiny form factor and the fact that these boards are roughly equivalent to the desktop computers I was using when I got to college(!). The other thing I’ll say is that I needed what I needed for the application I’m working on. I wanted to use what I had on hand, and because they’re so cheap, I had a bunch of these on hand. Naturally, then I went to a bunch more effort to get these adapter PCBs made. In for a penny, eh?</p><p>I also think that this edge card breadboard adapter is actually a good idea in the abstract. The march of progress yields electronics that are smaller and smaller, and yet people still want to prototype on breadboards and get their circuits working <em>before</em> they productize their design. Cooking something up from whole cloth, sending it off to Hong Kong to get fabricated, and just hoping and praying that it works when it gets back to you? That’s a slow, expensive process. My point here is that this edge connector breadboard setup could be a boon for other components as well.</p><p>If you’re reading this and you’re interested, please reach out and, for a few bucks, I would be happy to either mail you some of the extra PCBs I have, or give/license you the KiCad files, or something like that. If you’re interested, email me at rpzeroadapter-at-ipmcc-dot-com.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1995a4dc0f7e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The SpaceMouse™ and Other Input Devices for CAD]]></title>
            <link>https://medium.com/@ipmcc/the-spacemouse-and-other-input-devices-for-cad-e7f550e29d74?source=rss-3567854fbab1------2</link>
            <guid isPermaLink="false">https://medium.com/p/e7f550e29d74</guid>
            <category><![CDATA[cad]]></category>
            <category><![CDATA[3d-modeling]]></category>
            <category><![CDATA[spacemouse]]></category>
            <dc:creator><![CDATA[Ian McCullough]]></dc:creator>
            <pubDate>Sun, 25 Jan 2026 14:17:44 GMT</pubDate>
            <atom:updated>2026-05-21T20:29:06.037Z</atom:updated>
            <content:encoded><![CDATA[<p>Recently, someone on one of the Discord servers I watch asked about a “3D mouse” to accelerate CAD design work. We quickly figured out that he was talking about the <a href="https://3dconnexion.com/us/spacemouse/">3DConnexion SpaceMouse</a>. I bought one of these second-hand a while back, specifically the <a href="https://3dconnexion.com/us/product/spacemouse-compact/">SpaceMouse Compact</a>, so I figured I’d share my experience. It turned into a rather long discussion, so bear with me. 😄</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*99wBFRSJ-Ltf8ArDCqiH8w.jpeg" /><figcaption>3DConnexion SpaceMouse Compact (Source: <a href="https://3dconnexion.com/us/">https://3dconnexion.com/us/</a>)</figcaption></figure><p><strong>Why did I get it?</strong> I saw that many Fusion YouTubers use them, and they make <em>really</em> sweet looking videos where everything pans and zooms in one fluid motion. While I’m not a YouTuber, I figured that I might be able to use it to simply navigate my models more quickly and easily during my mainline work. My prior workflow was just going bonkers with my standard wheel mouse, zooming way out, zooming way in, drag the canvas around, and to do <em>all</em> that using my ‘normal’ mouse. I must confess, it looks a lot cooler when you can do it in one sweeping motion.</p><p><strong>Why is it on the left? </strong>Because it’s <strong>not</strong> a real mouse. By that I mean you can’t use a SpaceMouse to replace your normal mouse, and all the software support for it is <strong>application specific, </strong>by which I mean that (at least on macOS) there is no ‘standard’ mapping for what ‘moving a SpaceMouse’ would do in a general/abstract context. If you’re right-handed, you will typically be simultaneously using your right hand to continue to operate your ‘normal’ mouse, which leaves your left hand free to operate the SpaceMouse.</p><p><strong>What about the cost?</strong> Bought new, these start a little under $200, and go up from there as you add features. With a little haggling, I got mine from a fellow on Facebook Marketplace for $60. I’ve actually seen a surprising number of these devices come through those marketplaces. Not all of those are as cheap as mine was, but they’re not full-price either. Definitely take a look at the used market.</p><p><strong>What about building one, by 3D printing parts and assembling it?</strong> Being <em>that</em> type of person myself, I <em>definitely</em> looked into this. It is possible, and there are <a href="https://www.instructables.com/DIY-Space-Mouse-for-Fusion-360-Using-Magnets/">step-by-step plans out there</a> for doing so. After I found a cheap one online, I lost interest in building one from scratch rather quickly. I learned this lesson years ago when I built a CNC gantry router from scratch: If you want to use a tool like this, buy it; if you want to maintain and troubleshoot a tool like this then, by all means, go forth and build it yourself!</p><p><strong>Now for the less fortunate part:</strong> It seems that I am <em>strongly</em> right-handed, and I’ve been mostly unable to make frequent, beneficial use of the SpaceMouse using my left hand. As mentioned, it’s a <strong>companion</strong> to a regular mouse, not a replacement, so you really aren’t going to want to move one hand back and forth between the normal mouse and the SpaceMouse. (In fact, when I was trying it out, I often used both things at once.) I’m sure with enough practice, I could train my left hand to use it effectively (after all I play guitar standard/righty, and I deftly touch-type with both hands).</p><p><strong>What happened?</strong> I gave it the ‘old college try’ for about a week. But even after a week of practice, I was still navigating <em>much</em> faster with the normal mouse. Once the SpaceMouse became inadvertently buried under a stack of papers on my desk, it quickly became a case of ‘out of sight, out of mind’, and I haven’t used it in probably 3 or 4 months until I saw this person asking about it.</p><p>I strongly suspect that, given that there are a good number of these incredibly niche/esoteric products flowing through the Internet secondhand markets, it might be that my experience is actually <em>not unusual</em>. Put differently: People buy it, they give it a try, and for whatever reason, they don’t connect with it. Eventually they get tired of trying, give up and sell it. The guy I bought it from was a game designer; He had bought it to fly around in Unreal/Unity/etc 3D environments during game development, but had just found that he wasn’t ever using it, so he sold it to me.</p><p><strong>Conclusion? </strong>I think the bottom line is that before you buy anything along these lines, you should start by asking yourself what you’ll be able to do <em>with</em> it (or do <em>better</em>) that you can’t do <em>without</em> it. Unfortunately for me, that turned out to be pretty much ‘nothing’. I totally get why it’s 100% clutch with CAD YouTubers — it’s <em>really</em> good for that kind of thing — but for the core work of producing designs and modeling products? The benefit just wasn’t there for me. I do plan to hang onto it because it’s relatively small, and I’m curious to see if other use cases will crop up over time where it can perhaps help more.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UMllvig5HC8ziuDCItDFpQ.jpeg" /><figcaption>HOTAS System (Source: Amazon.com product page)</figcaption></figure><p><strong>What about </strong><a href="https://en.wikipedia.org/wiki/HOTAS"><strong>HOTAS</strong></a><strong> (Hands On Throttle And Stick)? </strong>Well… I haven’t tried <em>this</em> particular use for a HOTAS (I’ve used one extensively, but only for flight sim) so I don’t have any great CAD anecdotes to share, but I do think it’s generally fair to think of a flight stick as a <strong>2D </strong>control with a bunch of other buttons on it. (Maybe you could use the throttle quadrant as the third axis, but that would mean having both hands off of your normal mouse which feels like a step backwards to me, not to mention that it sounds like it would be profoundly awkward.)</p><p>On the other hand, a flight stick does share at least one capability with the SpaceMouse in that you can deflect the stick and then have it spring back to the center — they can be spring loaded. Maybe there’s some opportunity there. If I were going to try that, I’d be asking myself, after a week, “Am I actually working more quickly, or more accurately, etc. using this tool than I was <em>not</em> using it?” My sense is that I would not be faster with a flight stick, given that I have the purpose-built, much-more-capable SpaceMouse here, and even <em>that</em> device isn’t helping me work more effectively than I work with my normal mouse.</p><p>Even so, you’d probably also want to think about handed-ness here. Normally, most right-handed people, doing flight simulator things, will hold the flight stick with their right hand and adjust the throttle quadrant with their left. If you wanted to do CAD work using a HOTAS in conjunction with a normal mouse, are you then going to operate the flight stick with your left hand? And your normal mouse with your right hand? Or vice versa? If you’re considering that, I would recommend you play around with it, but I wouldn’t get overly invested in it. It just feels like the wrong tool for the job.</p><p>I haven’t exactly heard of lots of people running out to buy flight-sim rigs to do their CAD work, so it seems likely to me that this is not a well-established, or well-supported workflow.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/898/1*mgw9gi1WFEL8LBjcZxwG2A.jpeg" /><figcaption>The three-pedal device I use (Source: <a href="https://www.amazon.com/dp/B088WCG7L4?th=1">https://www.amazon.com/dp/B088WCG7L4</a>)</figcaption></figure><p><strong>Other ideas:</strong> I’ll admit it, I’m a sucker for an unorthodox input method. Over the years I’ve tried chorded keyboards, voice control, and currently I use a <a href="https://www.amazon.com/dp/B088WCG7L4">rack of three pedals</a> sitting at my left foot all the time that I use for various context-aware shortcuts. There’s no question in my mind that the pedal rack saves me <em>tons</em> of time, in the aggregate. I absolutely miss these pedals when I’m away from my desk and using my laptop without them. That’s something I unfortunately <em>cannot</em> say for the SpaceMouse: Once the initial anticipation and then excitement of purchasing it, and learning to use it, had worn off, I’ve never missed it one iota when I’ve been working away from my desk.</p><p><strong>The product:</strong> When I set out to write this, I did not <em>intend</em> to write a ‘negative’ review of the SpaceMouse. However, learning to use it well enough to improve my workflow? That didn’t really happen for me. If I had it to do over again, I concede that I probably would <strong>not</strong> buy the SpaceMouse. I would also recommend that anyone considering buying one try to find a way to audition one before buying. From a product perspective, the SpaceMouse is downright impressive: It’s beautiful. It’s got that certain heft. It feels solid under your hand. It’s responsive. It does what it says it does. That just doesn’t really help me much in terms of doing <em>my specific </em>work.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LeshH7v753zEtM1eVqNb2A.jpeg" /><figcaption>SpaceMouse Enterprise (Source: <a href="https://3dconnexion.com/us/">https://3dconnexion.com/us/</a>)</figcaption></figure><p>Their top model, the <a href="https://3dconnexion.com/us/product/spacemouse-enterprise/">SpaceMouse Enterprise</a> would be interesting to me because it’s also effectively a miniature, special-purpose keyboard with many standard modifier keys and 12 dynamically configurable keys, not to mention a small display to help you manage those configurable keys. I imagine I would like that one better, because (I suspect) I’d use it first for the keyboard/buttons, and then, if I already had my hand on it for the keys, I suspect I’d eventually get better at using the ‘stick’, if simply by osmosis. That said, $400 is a tough nut to crack, especially considering that I already have this SpaceMouse Compact that I essentially don’t use.</p><p><strong>In sum:</strong> If you’re making CAD videos for YouTube, you probably already have one of these, and if not, you should definitely look into it. For video production, it can produce some pretty slick results. If, on the other hand, you’re a designer, just trying to do standard CAD work, I’d recommend you look for a way to ‘try before you buy’, and see if you develop a feeling as to whether these devices are likely to improve or accelerate your workflow. Also, please do check out <a href="https://3dconnexion.com/us/"><strong>all</strong> of 3DConnexion’s products</a>, as they make good-looking ‘normal’ mice and keyboards too, and all their products are aimed squarely at the 3D/CAD user.</p><p>Happy designing!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e7f550e29d74" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The One About Making PCBs]]></title>
            <link>https://medium.com/@ipmcc/the-one-about-making-pcbs-12ef5973323c?source=rss-3567854fbab1------2</link>
            <guid isPermaLink="false">https://medium.com/p/12ef5973323c</guid>
            <category><![CDATA[electronics-engineering]]></category>
            <dc:creator><![CDATA[Ian McCullough]]></dc:creator>
            <pubDate>Fri, 15 Mar 2024 13:55:36 GMT</pubDate>
            <atom:updated>2024-03-20T15:19:43.121Z</atom:updated>
            <content:encoded><![CDATA[<p>So, I’ve been working on this project. What it is isn’t important, but I’m trying to make printed circuit boards (hereafter PCBs) for it. It has been a struggle.</p><p>Before you say it, yes, I know I can order PCBs from China/Hong Kong/Taiwan for cheap. The problem is that the turnaround time kinda bites. From the “best” fab house (as told on various forums and web pages) there’s still a ~8–10 day turnaround. I ordered a board, and it came through as specified, but it didn’t fit in the case I have. My bad. But it took over a week to figure that out, thanks to the turnaround time. So, yes, I’m aware that there is a “pro” option, but I want to iterate faster, and I stick to single-sided boards for just this reason.</p><p>OK, on to the meat of it. Here is a list of things I’ve tried and why they didn’t work.</p><ul><li>Toner transfer from glossy laser-compatible paper. This works about one in ten times. Produces decent boards, but takes 9 crappy boards to get it working right. I also ruined our clothes iron doing this and had to buy a new one.</li><li>Toner transfer from specially made ‘toner transfer paper’. I have never gotten a decent board from this paper. I don’t know why. It’s supposed to be better than the glossy laser photo paper, but in my experience, it’s not.</li><li>Pre-made MG Chemical photo-resist boards. They didn’t develop properly and when etched I ended up with a bare copper board. (I guess that made it easier to reuse… Always look on the bright side of life!) They’re expensive for someone like me who <em>obviously</em> doesn’t know how to, or have the equipment to, do this right.</li><li>Dry film. I’ve gotten some good and some bad from this. I’m trying to refine my process and I think it’ll work eventually, but I’ve made probably 15 bad boards with this process and one or two that were <em>half</em> good. I also managed to goop up my laminator when it ate a bunch of dry film. Rad.</li><li>Dry film part 2. I’ve tried using the sun to expose these. It works, but it’s hard to time it right depending on weather.</li><li>Dry film part 3. I’m using our dog-pee-detecting, hotel-jizz-detecting UV flashlight to expose it. It seems to be better than sun — at least more predictable and consistent.</li><li>Lastly, yes, as mentioned, I ordered some boards from Hong Kong and they took over a week to get here and didn’t fit into the case for the project. (My fault 100%.)</li></ul><p>OK, now we’re on to curing photoresist and etching.</p><ul><li>Ferric Chloride. As anyone who has ever worked with it will tell you, this stuff stains everything it ever touches except metal, which it simply eats alive (i.e. the whole point). It’s relatively expensive. It’s a pain in the butt to deal with. I don’t know how to get rid of it. (… well I actually <em>do</em> know how to get rid of it, but that’s <em>also</em> a giant pain in the butt.)</li><li>HCl + H202 + H20. Works but it always seems like a real pain to time. Lots of boards came out under or over etched with this one, too. Also high-concentration HCl is really nasty stuff, but at least you can buy it at Home Depot as ‘muriatic acid’ for pools. When you pour in this 37.5% HCl? It gives off visible fumes, and they’re really nasty. I caught a lungful once. I won’t do that again! Thankfully, I have the right mask and eye protection, but the end result of that is that I’m in my back yard looking like Walter White or Jesse Pinkman.</li><li>Sodium Carbonate (aka Washing Soda). Also quite prone to over development. In its favor, it <em>is </em>the least noxious of all the things I’ve tried.</li><li>Sodium Thiosulfate Pentahydrate. Very slow to develop, but not prone to over-develop. That said, one board took half a bottle of crystals, so I get only two boards out of a bottle, and it ain’t free. They say it’s reusable, but in my experience it just gets slower and slower. Oh yeah… this stuff also smells like horrible farts.</li><li>NaOH/Lye as developer. This stuff <em>goes to town.</em> If you want to use this, you have to time it right down to the second, and it’ll hurt you badly if it touches you. Remember that scene in <em>Fight Club</em>?… with the kiss? That’s this stuff. I have a jug of vinegar and a bag of citric acid crystals, and I’m not stupid, so I know how to deal with it, but I’d simply rather not, even with PPE.</li><li>Sodium Thiosulfate Pentahydrate + Sodium Carbonate. This is the current regime. It seems to share STP’s resistance to over-developing but gets things rolling along much faster than STP alone.</li><li>Ammonium Persulfate. I must confess: I bought this, but haven’t tried it yet. Maybe it’ll be the winner. Who knows? I’ll update if it is.</li></ul><p>I’ve also had a friend cut a board on his personal CNC machine that turned out great (Thanks, friend!), but again, I’m trying to iterate quickly, and it took something like 10 hours to cut a ~5x5&quot; board. I also don’t want to be the guy bugging this friend every two days to cut a board for me. End mills aren’t free, and he has his own life and job.</p><p>Another friend offered to help me by using a CNC PCB cutter at his work, but this created substantial discomfort for both myself and my friend, due to the operational costs and the extremely fast wear rates on the tools based on the feeds and speeds that machine works with. Although we do kinda sorta work together, I don’t want to get him into any trouble, so… He told me that if I have other options, I should use those first. Again, end mills ain’t free, and that cutter is pretty damn fancy, so if anything happened, questions would be asked, I’m sure.</p><p>I’ve even had some boards that came out crappy that I fixed up with wire-wrap wire where there were bad traces. But all of this is to say: I started down this road to iterate quickly. Instead of iterating quickly, I have dozens of unusable boards, and a cabinet full of noxious chemicals that will be a pain in the butt to get rid of, when the time comes. (God help us if we ever move.)</p><p>In the end, yes, I get the message… having boards fabricated by professionals in Hong Kong yields a far better product and doesn’t involve noxious chemicals (at least not in my house; I don’t know what it involves on their side… some sort of lithography I assume) and I get a silkscreen and a solder mask and all that good stuff essentially “for free”. I can use tiny, tiny traces, nestled in close, and be confident it’ll come out functional.</p><p>But here I am, proverbially kicking myself in the balls, over and over, trying to do this at home. If anyone in Pittsburgh has mastered this home PCB etching process and wants to spread the knowledge, please hit me up in the comments because I’m almost at my breaking point.</p><p>Whew. I’ve got another dry film and dog-pee-light board in the combo chemical right now! Fingers crossed. (It wasn’t usable in the end. Making PCBs at home can’t really handle traces that ‘sneak out’ between IC pins.)</p><p>Oh! One last thing for the folks that made it to the end: If you’re using dry film and transparencies, spray mounting the transparencies to the board before exposing them yields fantastic results in terms of exposure. You obviously still have to develop and etch them, but it seems to prevent the ‘blurry board’ problem without causing other problems. Two pieces of glass and clamps is even better!</p><p>Update: I’m really starting to question whether photoresist is actually any better than toner transfer. It sure looks good when it goes into the developer, but when you go to rub off the between-traces material, it always seems to mess up somehow and take traces with it, making it worthless. And again I have 7 or 8 failures from the last three days and I’m starting to get fed up with this process. I only bothered to do it because it ‘promised’ to be better at fine detail than toner transfer.</p><p>Update 2: I <em>finally</em> got a viable board out of the dry film photoresist process. FINALLY. And by “viable” I mean I only had to repair 3 traces with wire-wrap wire. But I’m starting to think I might actually know a little more about what to do to make this work. But I hate, hate, hate working with FeCl. Next up? Trying ammonium persulfate etching. Everything I’ve read suggests that it’s much faster, and much less messy than FeCl (which I have started calling “fecal” because working with it is about as fun as working with crap, which I know more about than I’d like thanks to my four pets.)</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=12ef5973323c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Didn’t Get a Job…]]></title>
            <link>https://medium.com/@ipmcc/didnt-get-a-job-f948a7c3614c?source=rss-3567854fbab1------2</link>
            <guid isPermaLink="false">https://medium.com/p/f948a7c3614c</guid>
            <dc:creator><![CDATA[Ian McCullough]]></dc:creator>
            <pubDate>Thu, 08 Feb 2024 23:05:29 GMT</pubDate>
            <atom:updated>2024-12-19T12:14:08.277Z</atom:updated>
            <content:encoded><![CDATA[<p>I recently went through a hiring process. I thought I did pretty well. It was my sixth time interviewing with this company (who shall remain nameless, but it’s a ‘big tech’ company). I spent 7.5 years at Apple and 3 years at Uber, so I’m hardly a slouch, and I’m no stranger to ‘big tech.’ Why wasn’t I considered more seriously? I honestly don’t know.</p><p>I’ve made it to later phases with this company before, but I’ve never gotten an offer. Why? I don’t know. This time, I was told that, “all interviews were positive” but apparently not positive enough to move on to the next step. This feels like a lie, but I interviewed many hundreds of candidates in my time at Uber, and in the course of these things, I’ve told those candidates plenty of lies about how wonderful they were despite the fact that I was pretty confident we weren’t going to hire them, so what can I really say?</p><p>I asked if the recruiter could share whether my rejection was based on technical knowledge or “EQ”. They couldn’t share even that. They shared only that one interview, with one particular person, went <em>really well</em>. Sadly, I guess it didn’t go well enough to overcome the other interviews that showed a “lack of signal.” In my time interviewing candidates at Uber, a “lack of signal” meant that the <strong>interviewers</strong> had failed, and that we needed to do more interviews. I respect the reality that this company doesn’t do that (or claims they don’t, in my case, at least) but I remain deeply fatigued with their hiring practices.</p><p>I’m not intrinsically upset about this. In any hiring effort, some people get hired, other people don’t. Of the hundreds of people I interviewed while at Uber, I’d guess we hired maybe 10–15%, so for every gratified person there were probably eight or nine people who were similarly disappointed. My issue is that this company keeps reaching out to me<em>.</em> I <em>don’t</em> submit applications to them; they <em>reach out to me</em>. I’m, quite frankly, tired of the sense of disappointment and futility that comes from interviewing with them.</p><p>It seems that I don’t know the ‘password’ that I need to say (or not say) to allow this company see the value that I could bring to them. Many other companies have seen that value, and decided to cash in on it. I’ve been taking a health-related break from work for a while, but I’m 100% ready to get back at it. This, for a time, seemed like it might be that opportunity, but in the end it wasn’t to be.</p><p>If you’re a hiring manager, interviewing someone for the sixth time, maybe reconsider even doing that at all. I haven’t gotten past your gate-keepers over six tries in 20+ years. What are the chances that you’re going to magically like me the next time? It just seems unlikely. Never say never, but my patience for this company has significantly waned.</p><p>A word to the wise, I suppose.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f948a7c3614c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Keys to Unlimited Paid Time Off]]></title>
            <link>https://medium.com/@ipmcc/the-keys-to-unlimited-paid-time-off-4733b0ddf269?source=rss-3567854fbab1------2</link>
            <guid isPermaLink="false">https://medium.com/p/4733b0ddf269</guid>
            <category><![CDATA[workplace-culture]]></category>
            <category><![CDATA[vacation]]></category>
            <dc:creator><![CDATA[Ian McCullough]]></dc:creator>
            <pubDate>Thu, 20 Apr 2023 14:11:29 GMT</pubDate>
            <atom:updated>2024-02-20T15:20:20.688Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*asNk8KDxj8CxRHPGJZE_vQ.jpeg" /><figcaption>Photo by <a href="https://unsplash.com/@zhenhappy?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">PAN XIAOZHEN</a> on <a href="https://unsplash.com/s/photos/unlimited-vacation?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditCopyText">Unsplash</a></figcaption></figure><p>I’ve found that there are few workplace topics that draw nearly as much side-eye as “unlimited PTO”. Over the last eight years, I’ve worked for three different companies with unlimited PTO. Unlike many of my coworkers, I’m actually a fan of the practice. And before I get into the details, let me put it right out there:</p><blockquote><strong>I’ve taken 6 weeks (+/- a few days) per year at every one of these companies, and I’ve never caught an ounce of flak for it.</strong></blockquote><p>Let’s start with the mechanics of it all. Unlimited PTO benefits the company because they no longer have to carry accrued PTO (for all their employees) on their balance sheet as a liability. It’s an accounting thing. With respect to employees, you still have to submit your vacation to your manager for approval, and they can still say, “No”, even though it’s “unlimited.” The other mechanical difference is that you don’t get paid out for accrued PTO when you leave an unlimited PTO workplace. (And in my experience, most often people complain about this particular issue in the hypothetical sense. I can’t think of a single person I know who felt like they got ripped off because they didn’t get paid out for unused vacation time, but that’s just me.)</p><p>Now, let’s talk about the semantics. The pedants among us might think that “unlimited PTO” means you can get hired and then just never show up and get paid forever. That’s also not the case. Again, it’s an accounting thing. “Unlimited” in this case means there is no hard limit and no accrual schedule. It might be easier or more semantically correct to call it “Discretionary PTO” where the discretion is your manager’s. That said, even if you work for a company with an accrual system (where you earn X days of PTO for Y days of showing up) it’s still, ultimately at the discretion of your manager. You can bring your wants/needs to your boss and the two of you work something out. If you want some block of time off that’s unusually long, that’s usually going to require some combination of things like PTO, unpaid time off, short-term disability, a sabbatical, a leave of absence, etc.</p><blockquote>“Unlimited PTO? That just means <em>no</em> PTO, amirite?”</blockquote><p>This is the derision I hear the most. If you are someone who’s fundamentally incapable of advocating for yourself, yes, it could turn into that. If your boss is just a complete brute, yes, it could turn into that. If you’re a low performer relative to your peers, yes, it could turn into that. Those are all crummy situations to be in.</p><p>I’m going to share here my keys to making unlimited PTO work for you:</p><ol><li>Understand what the expectations are ahead of time. This happens by you asking the hiring manager, <em>before</em> taking the job, how much PTO <em>they</em> take, and what people on their team take: <strong>average</strong>, <strong>min,</strong> and <strong>max</strong>. Take what they say at face value. Don’t expect the worst, but also don’t expect that you’re going to somehow be different or special.</li><li>Get your work affairs in order before you go on vacation. (<strong>I cannot emphasize this enough. In fact, if you take nothing else from this piece, remember this!</strong>) If, the week before your two week vacation, you just work normally and make no additional preparations for your absence, be prepared for people to be correspondingly pissed off with you when you get back. Delegate! Get other people to agree to cover your responsibilities, attend your meetings, etc. It’s reasonable for people to excuse your lack of individual contributions while you’re gone, but if part of your responsibility is enabling others to do their work, then you need to make sure that this enabling still happens even when you’re out.</li><li>If it’s feasible for you to check in once a week, by phone or text, consider doing so, just to make sure the wheels aren’t coming off in your absence. It doesn’t have to be a full status report; it can be a text message like, “Everything going OK?” I know this sucks, and detracts from the ability to fully unplug, but again, the more you do #2 (prepare), the less important #3 is.</li><li>If you are not in the good graces of your manager, unfortunately, you have to be judicious, and you should try to get <em>into</em> those good graces. Ideally, if you want to take an above average amount of vacation, you really will benefit by already being recognized as an above average performer. If you’re teetering on the edge of being disciplined, or worse, let go, asking for a boatload of vacation won’t help your situation. Sorry to say it, but to the victors go the spoils. Real life isn’t equitable. If you want to take that month-long vacation, then you’ve got to cultivate a mindset in those around you, and especially those that manage you, that you deserve it.</li><li>Get your request in <strong><em>early </em></strong>— 6 months ahead of time, if possible. When it’s May, and you say, “I need the entire Thanksgiving week off this year, as I will be traveling.” it’s really uncommon for managers to say, “No”, because the proverbial ‘pain’ of your absence is so far off in the future. If they come back at you in November in protest, you can ask, “I gave you six months notice; What more could I have done to make this work?” What this is doing is saying, implicitly, “This is your failing, dear manager, not mine, but what can I do to help?”</li><li>Early submission is especially helpful in deadline-oriented businesses. “It’s April 15th, Product X is planned to ship on October 15th, I’m going on vacation the two weeks after that release date, cool?” …because deadlines <strong>often</strong> slip, but vacation plans are usually much less flexible, so unless your company wants to reimburse you for vacation expenses, they’re going to have an uncomfortable case to make.</li><li>…and, in closing, if it becomes clear that your company’s unlimited PTO policy is actually just complete BS, shine up your resume and start looking for another job, because if they dissemble about PTO, they probably lie about other important stuff too.</li><li>…and very lastly: Enjoy your vacation!</li></ol><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4733b0ddf269" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Animating Off One’s Nose to Spite One’s Productivity]]></title>
            <link>https://medium.com/@ipmcc/animating-off-ones-nose-to-spite-one-s-productivity-ac4fd49af2e?source=rss-3567854fbab1------2</link>
            <guid isPermaLink="false">https://medium.com/p/ac4fd49af2e</guid>
            <category><![CDATA[react-native-reanimated]]></category>
            <category><![CDATA[react-native]]></category>
            <dc:creator><![CDATA[Ian McCullough]]></dc:creator>
            <pubDate>Sun, 13 Nov 2022 16:38:25 GMT</pubDate>
            <atom:updated>2022-11-14T12:24:27.895Z</atom:updated>
            <content:encoded><![CDATA[<p>I’ve been working on a React Native-based app for a good little while now. It’s not the <em>exact</em> modality of mobile app development I would’ve chosen, but my company was already a couple years into it when I joined. Being cross-platform is important to us, but we have a very, very small engineering staff, so doing native development for two platforms really isn’t an option for us at this moment.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*FKgllzhNZd1lMf1d2sLb5g.jpeg" /><figcaption>(Image from Google Image Search)</figcaption></figure><p>A few months back, I pulled our mobile app’s repo and, all of a sudden, my source-level debugger no longer worked, and the app seemed to be broken in myriad other, genuinely puzzling, ways while trying to debug. Over time, I would come to learn that this was because of a package called <strong>“react-native-reanimated”, version 2</strong>. The authors of this package have, in the name of reducing latency, adopted React Native’s nascent <a href="https://reactnative.dev/docs/next/the-new-architecture/pillars-turbomodules">TurboModules</a> (don’t worry, you don’t need to know what TurboModules are to understand my gripe here — just know that they are new, esoteric, ostensibly marginally faster, but not without their challenges) and also run a completely separate JavaScript engine, on a separate OS thread, in order to achieve their desired ends. Unfortunately, the current debugging support in React Native (via Chrome) <a href="https://github.com/software-mansion/react-native-reanimated/issues/1223">does not support</a> source-level debugging in any app that uses this library.</p><p>For my part, my first experience of this was one of a jaw-dropping <strong>40:1</strong> reduction in my productivity: We had a bug that I spent, quite literally, <strong>20 hours</strong> searching for, using <em>console.log</em> statements, at my team’s recommendation. After becoming thoroughly fed up, and stubbing out all components that used Reanimated v2, I found and fixed the bug in just <strong>30 minutes</strong>, using the source-level debugger. The bug had <strong>absolutely nothing</strong> to do with animation, but our use of Reanimated v2 had disallowed debugging anyway.</p><p>I thought to myself:</p><blockquote>“These poor souls! console.log-debugging is for barbarians! They’ve never known the joy and productivity of source-level debugging! I shall teach them! Once they see how much more productive they could be, they’ll surely abandon Reanimated v2 and join me in the bliss of greater productivity!”</blockquote><p>I took my branch, in which I had stubbed out Reanimated v2, and I held a Zoom call wherein I showed them the magic of breakpoints, variable inspection, run-time expression evaluation, watchpoints, conditional breakpoints… the whole kit and kaboodle. They knew I was fired up about this topic, so I don’t pretend to know how genuine it was, but they all said things like, <em>“Oh, that’s pretty neat!” </em>and<em> “That would be very helpful.” </em>And yet, for some reason, their devotion to Reanimated v2 remained unshaken. I proposed that we downgrade to either the “included” React Native Animated, or to Reanimated v1. I was told by my engineering lead, “You cannot downgrade. Reanimated v2 is the future.” There is apparently no other option. (Never mind that Reanimated already has <a href="https://github.com/software-mansion/react-native-reanimated/commit/05c93b02ede93950904812239fbe95366e372ab6">v3 Release Candidates</a> out.)</p><p>In doing my homework on this, I was genuinely surprised to see that the Reanimated v2 developers are <a href="https://docs.swmansion.com/react-native-reanimated/docs/next/#known-problems-and-limitations">completely unrepentant</a> about this. They know very well that their library ruins debuggers for anyone that uses it, and therefore costs probably hundreds of thousands of hours of wasted productivity per year (based on the <a href="https://www.npmjs.com/package/react-native-reanimated">over 500,000 weekly downloads</a> of it); they <em>just don’t care</em>. They try to disclaim it by saying that “Reanimated 2 is in an early version”, yet here we are, staring down the barrel of v3. When does v2 stop being “an early version”?</p><p>If Reanimated v2 were a drop-in replacement for React Native’s Animated library, such that you could switch back to Animated when you needed to debug, or even if Reanimated v2 had a switch where you could say, “I’m willing to endure a dropped frame from time to time to save <strong>half a week’s</strong> labor…”, I would be fine with it. I concede debuggers aren’t perfect: the curse of debuggers is, of course, the <a href="https://www.rookout.com/blog/fantastic-bugs-and-how-to-resolve-them-ep1-heisenbugs/">Heisenbug</a> — a bug that shows up when you don’t have debugger attached, but doesn’t manifest when you do — but by all accounts that I’ve seen, the Reanimated folks seem to consider it a worthwhile tradeoff. Why else would they ever have released this? More importantly, how do <strong>they</strong> debug anything?</p><p>I may, some sad day, work on an application where a single dropped frame of animation is <em>so</em> important that it justifies immolating the modern developer experience (and by “modern” I’m referring to the 40+ years that source level debugging has been around.) I’ve certainly worked on apps with requirements in that neighborhood before. I worked on <a href="https://apps.apple.com/us/app/keynote/id409183694?mt=12">Keynote</a> chart animations for the better part of a decade, where dropped frames were completely unacceptable, but somehow the team that wrote the animation libraries managed to make them work with debuggers. All I can conclude is that these folks value 16ms of dropped frame more than the time and productivity of the developers that want to use their library.</p><p>The hardest thing for me, honestly, has been trying to understand my team’s devotion to this poison library. “It’s the future,” they tell me. If that’s true, then frankly, <strong>the future sucks</strong>. I’d like to think that Google will expand the debugging capabilities of Chrome to be able to accommodate multiple JS engines running on different threads, and to understand TurboModules, and whatever else it takes, but I’m not naive: React Native is a relatively small development community (compared to JS on web, C/C++, Java, Swift, Kotlin, etc.) and Google has other priorities. I’d also like to think that the developers of Reanimated will eventually come to terms with what they’ve done, and give the community a way to work around the sacrifices needed to guarantee perfect, frame-accurate animation, while letting us debug once again. I’m not holding my breath, because they already have a v3 release candidate going, and while there is some <a href="https://github.com/software-mansion/react-native-reanimated/pull/3446">apparent efforts in the area of debugging</a>, it’s still got problems (problems that were introduced by Reanimated in the first place). Beyond that, who knows how long it will take for every package that transitively uses v2 to migrate to v3 (as I write this, <a href="https://www.npmjs.com/package/react-native-reanimated">npmjs.com shows 746 dependents</a>). Maybe it, too, will become “the future” and bring salvation to us all. We can only hope.</p><p>In the meantime, I guess all I can do is: <em>console.log(‘Screw Reanimated v2… and the horse it rode in on.’)</em>.</p><p>As a final note: Before you comment to tell me I should check out <a href="https://fbflipper.com/">Flipper</a>, please know that I already have, and while it is a useful tool, that does lots of neat stuff, it is <strong>not </strong>the same thing as a source-level debugger (unless you’re using a very different Flipper than I am.)</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ac4fd49af2e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Firebase Local Emulators: My Arch-Nemesis]]></title>
            <link>https://medium.com/@ipmcc/firebase-local-emulators-my-arch-nemesis-31e54ac9658a?source=rss-3567854fbab1------2</link>
            <guid isPermaLink="false">https://medium.com/p/31e54ac9658a</guid>
            <category><![CDATA[firebase-cloud-functions]]></category>
            <category><![CDATA[javascript]]></category>
            <category><![CDATA[firebase]]></category>
            <dc:creator><![CDATA[Ian McCullough]]></dc:creator>
            <pubDate>Sat, 03 Sep 2022 20:24:52 GMT</pubDate>
            <atom:updated>2022-09-03T20:24:52.976Z</atom:updated>
            <content:encoded><![CDATA[<h4>The worst of both worlds.</h4><p>Fair warning: this is a rant, and unlikely to do any of my regular readers any good, but I need to get it out, because I’ve wasted 2.5 days now beating my head against this, and I’ve gotta send the vitriol somewhere.</p><p>I struggle to find the words to describe how frustrating it is to work with Google Firebase local emulators. I am trying to do what should arguably be the simplest thing one could ever want to do with this software: I have a React web app, and want to call one or two Firebase Cloud Functions from it.</p><p>Before you ask, yes, I have gone all the way back to copy/pasting examples from the documentation, and my experience is that <em>nothing</em> works reliably. The sample code never works out of the box. On top of the sample code never working out of the box, every single debug cycle looks like this:</p><ul><li>Quit Chrome.</li><li>Kill any running emulators.</li><li>Go to the functionsdirectory and manually do a tscinvocation to transpile the TypeScript into JavaScript.</li><li>Re-launch the emulators (with imported data exported from our prod instance.)</li><li>Go to the emulator console web site and manually reset the password on my account (Yup, every time! I get why they don’t bring over the passwords — security and all — but this is the step that I’m arguably most bored with.).</li><li>Attach the WebStorm debugger to the now-running functions emulator.</li><li>Log into the React web-site with my email and password (Yup, every time!).</li><li>Navigate to the page that would, ideally, trigger the cloud function.</li><li>Click the button and try to trigger the function.</li><li>Collapse into a weeping heap of anguish on the floor, because it still DOESN’T FREAKING WORK.</li></ul><p>I get that distributed systems have more moving parts than standalone desktop/mobile apps, so there’s always going to be some overhead. That’s still no excuse for the developer experience to be <em>this bad</em>. The whole reason for local development environments is so you don’t have to deal with all the cloud config and deployment in order to make incremental amounts of progress, and here, it appears, Google has in-sourced all the pain, and added even more on top.</p><h3>Other crap:</h3><ul><li>Standard .env.development.localfiles don’t appear to work with the emulators, so you’ve gotta hardcode all your secrets. Cool. Hope I don’t forget to remove them before I commit to Git!</li><li>Forget any ONE of the above steps, and you’ll be greeted with some inscrutable error message that does nothing to point to the actual underlying problem. My fave of these are the CORS-related crap. I’ll never get that six hours of my life back.</li></ul><h3>Here are some hot tips for Google:</h3><ul><li>Make functioning sample code. (Not a single snippet I’ve copy/pasted from their website didn’t require extensive edits before it would do <em>anything.</em>)</li><li>Keep it up to date. (They’re still using require… it’s 2022. WTF?)</li><li>Make sure it works, out of the box, in, let’s say, the top five IDEs that people use to write Firebase apps (VSCode and WebStorm would seem like a good start).</li><li>Use best practices. .env files for example.</li><li>Accompany these samples with exhaustive, bullet-proof documentation that explains, blow by blow, how to change the sample code to work with your Firebase project (i.e. API Keys, etc).</li></ul><p>In my experience with Firebase functions, it is literally 100x easier to test in production than it is to use the local emulators. Working locally should be <em>easier</em>, not <em>harder</em>. I value the ability to use debuggers, and to work offline, which is why I’ve been wading through this miasma in the first place, but after multiple days of beating my head against this, I’m just about done.</p><p>I despise software like Ruby on Rails that values magic over understandability. This is a different beast, but it’s similar; They’re trying to help, it’s just not working.</p><p>I started down this road because Google’s docs said that doing batch Firestore writes and deletes from a browser was not a “best practice”, and suggested bundling those operations up into cloud functions. 2.5 days after trying to follow their advice, I never want to write, test, deploy, or debug another Firebase Cloud Function ever again.</p><p>Larry Wall (creator of Perl) once said:</p><blockquote>Easy things should be easy, and hard things should be possible.</blockquote><p>My experience of working with Firebase Local Emulators is that easy things are <em>almost</em> impossible, and hard things are <em>actually</em> impossible. I give up.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=31e54ac9658a" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>