A RUNNING COLLECTION
What I’m studying, the questions I’m working through, and links worth coming back to.
My current focus is reliable AI-assisted development: how to move faster while keeping the software understandable and the results verifiable.
I’m also learning about agents, plugins, and RAG, and how these capabilities can help AI do useful work.
Video / Matt Pocock
A great talk on making AI more predictable in development, with attached skills to help put the workflow into practice.
I recommend this talk for better ways to make AI more predictable in development. The attached skills make it especially useful for exploring how to put those ideas into practice.
Open the video on YouTube for the talk and its accompanying resources.
Article / Ryan McCue
McCue makes the case for explicit ownership of how AI-assisted work is specified, executed, verified, and delivered.
A question for team discussion: who is responsible for improving the whole workflow, from defining the work to knowing it is ready to ship?
The article is a useful starting point for thinking beyond tool access and individual prompting habits.
Article / Ryan McCue
McCue argues that AI-assisted changes should come with evidence of behavior, with verification expectations scaled to the risk of the change.
A question for team discussion: what evidence would help a reviewer understand this change without having to reconstruct all of the author’s reasoning?
McCue’s examples include targeted tests, screenshots or demos, and risk and rollback notes where appropriate.
More reading on team AI usage: Ryan McCue’s blog ↗