Engineering Blog

At first I was bemused by Retort, Adrian Cockroft's whole stack coding agent evaluation tool. It is growing on me, and might be useful as a starting point for custom evals.

Every application has some kind of workflow. Understanding larger workflows, making them cohesive and understandable to users and developers is non-trivial. I re-discovered Ward Cunningham's Swim the Test approach, this is the Further Reading section of future posts.

I am making small contributions to tiny open source projects I have just started using, where I would let it be before coding agents.

Synthetic Test Driven Development is when you have a coding agent generate the tests and the code, and maybe also do the refactor step with an agent.

I don't like what I now call Moralistic Programming. 'You are not a good developer if <you don't use my pet favourite practice>'. I see some of that resurfacing in discussions around coding agents. I value Situational Awareness over best practices and fingerpointing..

Next annoying variation in my blog writing workflow is adding images to a post. I wanted to add drag and drop. I let a coding agent generate something that almost worked, and then did the finishing touches by hand, simplifying the UX and code as I went.

Brian Marick once distributed stickers with 'An example would be handy right about now'. This is also handy in cases where writing a test is not directly obvious

QWEN 3.6 models are multi modal, but getting that to work requires an additional file and configuration in llama.cpp and enabling images in the Pi.dev models configuration.

Checking out a repository on a fresh machine often involves some fiddling around. An init scripts costs only one prompt and gets you going now, and months later.

In which the writer reflects on believing AI is conscious