Posts

You’re probably not running out of tokens. You’re running out of confidence. If your plan is too small, no widget fixes that — upgrade first. But if you’re on a flat-rate subscription and still rationing prompts like a hiker rationing water on a trail that turned out to be two miles long, the problem isn’t supply. It’s visibility. You self-throttle because you can’t see the gauge, and you finish the day with budget to spare and nothing ambitious to show for it. That’s overhead, not output. ...

Something subtle happens as your Claude Code context fills up. The output gets slightly less sharp. The architecture suggestions get more generic. The creative leaps come less often. The code still works, but it’s more… obvious. This isn’t the old “AI forgot what I said” problem — Opus 4.6 stays on target like a bloodhound, even at 80% context. It’s more like the difference between a colleague who’s fresh in the morning and one who’s been in meetings all day. Still competent. Just not at their best. ...

I finally got a secretary. Not a real one, alas, but still. Ever since I can remember I’ve wanted one. First mostly because my dad had one, and to the five-year-old me he was of course the most powerful man in the world. If I had a secretary, I’d be powerful too, and that would make him proud. Easy! When I got older this understanding shifted to finding the idea of a secretary extremely practical. Imagine not having to think of hard-to-remember stuff, like names and birthdays. ...

I embarked on my quest to Pantsify (as we call it in my team) our quite large monorepo in juli 2024. I’ve made some notes along the way; this is my Pants journey. Let me save you some time: Pants is brilliant if you keep close enough to the paved Python road. You can even walk in the grass next to it, but don’t go too far because you’ll need to learn new spells. Pants is still worth it forces a single way of working, and allows you to run all validations constantly and repeatably; because it knows what hasn’t changed and therefore can be skipped. ...

I recently had a colleague who suggested it might be a good idea to test our Python software against older versions of Python. And at first I was like: Yeah, good idea! It’s good practice, right? Support a whole range of versions! We see that all the time in the packages we fetch from Pypi. Supporting more versions is just good stewardship, right? But we were walking into a trap. We were discussing our own software, that we run, ourselves, in Kubernetes. We control the environment, all the way from development to production. There is simply nowhere where we don’t explicitly decide what Python version to use. On top of that most of that software is technically scripts, not installable packages. ...