I saw a photo of an SF billboard with what looked like a broken URL: https:// {64659, 123310, 75584, 8138, 38271} It led to Listen Labs’ Berghain Challenge, where you act as a nightclub bouncer an...

I have written before about publishing my own collection of agent skills through npm. At the time, the appealing bit was the small command: npx skills install @abuxton/skills Put a useful SKILL....

I owe you an apology for the quiet August. I have not posted much because personal life took priority: I got married. It was a very good reason to step away, but it is still nice to be writing agai...

Birgitta Böckeler at Thoughtworks spent four weeks running AI coding models locally on Apple Silicon and wrote up what she learned in two detailed memos: factors affecting viability and practical e...

I thought it interesting to take a look at Microsoft’s Waza and compare it to whatever else exists for evaluating AI agents and skills. That request landed at a good moment. I have been quietly bui...

I have access to both GitHub Copilot and IBM Bob through work, and every now and then I do the same slightly unfair but very revealing thing: I give them roughly the same prompt and see what comes ...

I have just read Salvatore Sanfilippo’s post, Control the ideas, not the code, and I think it is one of the sharper pieces yet on what AI is actually changing in software engineering. Not because ...

There is a course on ergo.org called Computation as a Universal and Fundamental Concept that I have been sitting with since I read it. I think it is one of the more interesting things I have come a...

Joe Beda’s post, The animatronic rubber duck, is excellent. Not because it claims AI has become a magical collaborator, but because it finds a better description for one of the most useful ways to ...

I have just read Lars Faye’s post, Agentic Coding is a Trap, and it is worth your time. Not because I agree with every implication in the strongest possible form, but because it names a tension th...