My manager asked me last week how I had fixed an issue. Not what the fix changed, and not whether it was correct. How it worked, in the code. I could answer the first two questions without hesitating. I could not answer the third, so I told him that.
Then I told him what I actually think, which is that we should not be required to know every implementation detail of everything we ship. We should be required to have enough tests in place to catch it when something breaks.
Everything you need to run it well on a phone, and the popular advice that quietly makes it worse.
I do real work from my phone — a Django site, a scraper pipeline, deploys to a Hetzner box, all from a Pixel 8 Pro. Claude Code runs on it and has done for months.
Getting there took a day of benchmarking, most of which produced results I didn’t expect and a few that reversed things I’d assumed. You don’t need to repeat any of it. Here’s the setup, in order, and then the list of sensible-sounding tweaks that measure out to nothing or actively hurt.
Ten years of cheap electronics, and a new €3 charge that quietly makes the whole habit pointless.
I’ve been buying from AliExpress since November 2015. My first order, for the record, was a €1 bag of replacement screws for a MacBook — the sort of thing no shop near me stocks and no sane person pays shipping for. That pretty much set the tone for the next decade.
I pulled my whole account history the other day and counted 322 orders. Just under €5,000 over ten years, which sounds like a lot until you look at what it actually was: jumper wires, GPIO headers for a Raspberry Pi, solder, a handful of sensor boards, plastic sleeves for my game cartridges, little displays you genuinely can’t buy here in the Netherlands without paying three times the price. Supplies, basically — the raw material for a hundred small weekend projects. There were a couple of real splurges buried in there too — a brand-new iPhone, the odd used MacBook SSD — but mostly it’s a long, boring list of things that cost a euro or two.
For all that it comes to ten years and nearly €5,000, the median order was about €5. Almost half of everything came in under that, fifty-odd orders were under €2, and the cheapest was a literal one cent. The most expensive single thing in a decade was that new iPhone, at €499 — the exception that proves how small the rest of it is. That’s the shape of it: a steady drip of tiny parcels, around thirty a year, with the occasional proper purchase thrown in.
On 1 July 2026 the EU changes the rules in a way that makes that kind of shopping stop adding up. So I’m winding the account down. Here’s the thinking.
A from-scratch RAMSES-II receiver on the cheapest hardware that works — and the one detail that makes the signal look encrypted when it absolutely isn’t.
You can sniff and decode your Honeywell evohome / T-series heating straight off the air with a Raspberry Pi Pico and a ~€1 CC1101 radio — no nanoCUL, no ESP32, no pre-built firmware. Here’s the whole thing from scratch: the exact wiring, the working code, and the one detail that makes the signal look encrypted when it isn’t.
Here is a jar of jelly beans. How many are in it?
Take a guess and hold onto it. We’ll come back to it — and to the slightly unsettling fact that nobody, including the people who set the puzzle, actually knows the answer.
A side project of mine is a video-game price-comparison catalogue. It pulls product data from a handful of retailers — affiliate feeds, public web APIs, a couple of cookie-based scrapers — and shows you which of them currently has the best price on a given title. About 27,000 editions across ~12,000 games.
I picked up three LiitoKala 34B flat-top 18650s to populate the Waveshare UPS Module 3S sitting under my Raspberry Pi 5 home server. The cells are sold as a Panasonic NCR18650B rewrap — the printed cell label reads NCR18650B Li-ion. They came from the LiitoKala official store on AliExpress at about $2/cell after coupons and AliExpress coins (closer to $4 at sticker), which is well under half what genuine retail-channel Panasonic cells go for.
I started using Claude Code in December last year. In four months, I’ve delivered what my team would have needed two or three years to ship — at least. That sounds like a happy ending. It isn’t quite.
Early 2025 AI was a toy. ChatGPT could write a function, sometimes correctly, often not. It was impressive in demos and frustrating in practice. A year later the picture is unrecognisable. AI is now genuinely useful — useful enough that I can build almost anything I want in a single day, regardless of how complex it is.
Last year I went to the Google Cloud Summit with a few colleagues. We saw a lot of impressive projects built around agents and LLMs, and walked out a bit deflated. As a data team — analysts, engineers, but not “pure” software developers — we could see the value clearly, but we’d seen this movie before: the moment we wanted to build something real, we’d be queuing up for backend developer time, or pricing out an external consultant to implement agents for us. Same bottleneck as always, just with a fancier label.
A year later, that bottleneck has quietly dissolved. We built it all ourselves — fast — with Claude Code as the agent in our IDEs. The projects we thought we’d have to outsource, we shipped on quiet afternoons.
And it isn’t just the work-shaped projects. I’m finally shipping ideas that have sat in my personal backlog for literally ten years — the kind of thing where every time you open the notebook you remember why you closed it last time. Now they’re real, deployed, and — the part I keep being surprised by — they’re solid. Tests, sensible error handling, security defaults that hold up to scrutiny. Not vibey demos. Things I’d hand to a friend without a list of caveats.
When people ask how much faster I work with the agent, my honest answer is that “faster” is the wrong axis. I built this chart for an internal talk last week to explain why.