·801 words·4 mins
I test because I want to know what I’ve built is working. That’s very different than valuing testing.
·288 words·2 mins
I understand how batteries work. It doesn’t make them any less magical.
·1340 words·7 mins
Your instructions to an LLM can quietly bake in the failure mode you’re trying to prevent.
·1016 words·5 mins
Every good team needs its wildcard.
·1279 words·7 mins
An attention focus tells an agent where to look. A persona tells it how to think.
·3356 words·16 mins
Small differences in prompts produce wildly different outputs. You can fight that or you can use it.
·5655 words·27 mins
I treated Claude like a junior engineer, taught it principles, and watched it grow. Two weeks later, it’s on my short list.
·3775 words·18 mins
The ideas behind Parma, explained like we’re sitting at a bar and you asked me what I’d been working on.
·1061 words·5 mins
How Peter Alvaro’s lineage-driven fault injection clicked with me, and why I ended up at Antithesis.
·2030 words·10 mins
My wish list for a code management tool that actually understands how I work.