Experimental agent control
Stopline
A browser-agent decision gate. A model classifies a proposed action and application policy decides whether it may run. The repository documents the mechanism and its limits.
Work you can inspect.
I publish small engineering artefacts and teaching material to make a design decision easier to examine. Each one states what it does, what it leaves unproven, and where to look next.
Local coding agent
Metron
A small terminal agent for local AI development. It limits the tool surface, proposes a patch, and waits for explicit approval before applying it.
Free teaching exercise
AI-assisted code review kit
An illustrative shipment-mapping exercise with runnable JavaScript, a worksheet, and a worked answer. It examines an unsupported default that the original fixtures miss.
Apply the review exercise to a real process.
I also offer a proposed private workshop for teams that want to examine their own review expectations. The page describes an interest enquiry. It does not reserve a date or accept payment.
Review the workshop details