Insights
Notes from the work.
What holds up in production, what doesn’t, and how we decide. Written by the engineers who do the work.
September 2026
Engineering practice
Why every AI project should start with a fixed-price feasibility study
A few weeks on your real data tells you more than any proposal. Here is what we test, and what we hand back.
6 min read
August 2026
Computer vision
Computer vision on the shop floor: what lighting does to accuracy
Models trained on clean footage meet glare, shadow and night shifts. How we plan for the conditions a camera actually sees.
7 min read
July 2026
Machine learning
Digital twins for healthcare operations: start with the queue
The most useful simulation is rarely the most detailed one. Why we model waiting first and everything else second.
9 min read
June 2026
AI agents
Where AI agents need a person in the loop
Agents are good at the routine and poor at the exceptional. A practical way to decide which steps a person should approve.
5 min read
May 2026
Engineering practice
What “you own the code” should mean in a contract
Code, models, data pipelines, credentials and documentation: a checklist for what should be handed over, and when.
4 min read
That’s every note so far — new ones arrive monthly.