Coding agents need a delivery system
A field report on coordinating coding agents, review, repository state, and human authority while shipping a real open-source project.
Technical notebook
Case study
Practice note
Technical guide
Field notes on useful AI systems.
Case studies, practice notes, technical guides, and looser ideas from the middle of building useful AI systems.
Featured field notes
Published · Updated
How to build an AI platform in a mid-sized or large company
Build around delivery bottlenecks, governed context, evaluation, operations, and reusable primitives.
Building LLM software lives between two worlds
Reliable LLM products need deterministic scaffolding and habits from both data science and software engineering.
More from the notebook
Earlier notes and working ideas.
Some entries are developed guides; others are simply a record of what I was thinking at the time.
AI failure modes: four lenses
A checklist for debugging the problem, context, model, and tools around an AI feature.
China’s open source wave
Personal analysis of open-weight models, developer ecosystems, export controls, and influence.
If you work in a non-tech company, this is how AI will change in the next 2 years
A dated forecast about embedded AI tools, platform capability, and governance.
AI changes work. Here’s how I think we should respond
Thoughts on routine work, human judgement, glue skills, and adapting organisations.
The economics of intelligence-as-a-service
Working questions about pricing, cost curves, value capture, and abundant reasoning.
Intelligence as a commodity
An early essay about cheaper model capability and the physical infrastructure underneath it.
Dataset creation at scale
A compact pattern for enriching datasets with structured LLM outputs.