The series
Papers you can read now, and a build documented week by week.
The argument, paper by paper
Each piece is a PDF you can read or keep. The full six-dimension Index paper is in progress and will land here first.
Where the CDI lands
Placing the measurement layer against the methods, and what it reveals about each one.
Paper · PDFReporting AI to the board
A 27-page working guide to reporting AI upward: what a board actually asks for, the three tiers of evidence behind any claim, a specimen board page, aggregation rules for federated groups, and a ninety-day plan.
Paper · PDFWhen review becomes the job
What happens to review when the agent writes the code, and how to keep it substantive.
Paper · PDFWhat happens to the junior developer?
The path into the profession when entry-level work is the first thing an agent absorbs.
Paper · PDFArchitecture at agent speed
Enterprise architecture governance was built for a world where change was rare. Agent-assisted delivery ends that world.
Series · in progressThe Cognitive Delivery Index
The instrument in full. Six dimensions, five levels, and the governance theatre gap at the centre. Landing in Research first.
Notes from the field
Shorter pieces reading current events, platform moves and delivery patterns through the framework.
Clean data is not a control
Data quality answers the easy half of the trust question, and leaves the part that actually governs the output untouched.
Companion note · PDFThe harness now enforces the policy
Moving agent policy out of the prompt and into the harness is the right layer, and exactly where the governance theatre gap moves next.
Field note · PDFThe hours were never the product
What the first governed step looks like for an organisation that can neither stop nor wave everything through.
Field note · PDFThe loop got faster. Are you still in it?
Teams have noticed the loop outpacing them, and the instinct to make the output prettier is only half a fix.
Field note · PDFThe model was never the variable
Not one of these shipped failures was decided by which model was in the loop.
Companion note · PDFValue the goose, crack every egg
Why an unexamined mechanism is not a golden goose, it is an unaudited one.
Companion note · PDFThe platform governs the agent. Who governs the work?
Permissions, guardrails, traces and cost controls govern the agent. Governing the work is still open.
Analysis · PDFNine seconds in production
A CDF and CDI reading of the PocketOS incident, where a Cursor AI agent deleted a production database in nine seconds. What structurally failed, and the controls that would have stopped it.
Instruments and toolkits
The working documents that go with the Index: the companion guide, the workshop canvas, and the cheat sheets.
MCP went stateless: the control point moved
A two-page reference card on the 2026-07-28 MCP specification: what was removed, which headers are now required and where, what a gateway can and cannot enforce, and the six moves worth making.
Field guide · PDFThe AI governance plugin, governed
A seven-page field guide to Anthropic's ai-governance-legal plugin: how to run it well, an honest map of where it performs and where it stops, what to pair it with, and the wiring that takes each output from Level 4, Complete, to Level 5, Substantive.
Two axes, not seven rungs: the diagram
A one-page diagram redrawing the "seven layers of AI" mountain as a grid: what a system can do against what it is allowed to do, with the seven rungs as one diagonal through it and the corner the mountain cannot draw marked in rose.
Companion · PDFThe CDI companion guide
The full companion to the Cognitive Delivery Index, six dimensions, five levels, and how to score a team honestly.
Toolkit · PDFThe CDI workshop toolkit
A workshop canvas for running a CDI self-assessment with a delivery team.
Cheat sheet · PDFThe CDF DORA supplement
DORA tells you whether a team ships well. It does not tell you whether a control demonstrably changes what an agent does.
Cheat sheet · PDFThe CDI cheat sheet
Version 0.1 of the one-page CDI reference, superseded by the CDI Companion Guide. Kept because the URL and the PDF are already in circulation.
Cheat sheet · PDFWhen the regulator says “show us it works”
The FCA will not write an AI rulebook, it will supervise the AI you already run through the frameworks you already hold.
Cheat sheet · PDFUpskilling for the agentic shift
The people staying are the ones who can supervise the work, a cheat sheet for upskilling with governance built in.
Documenting the kernel build
The plugin's rebuild proved that governance has to be enforced structurally. The next chapter is the governance kernel, and its R&D is documented entry by entry in the devlog.
Devlog · ongoingThe governance kernel, in Rust
Policy as a compiled artefact, enforcement below the editor, and an audit trail built for evidence.
New material lands week by week
Research, the kernel devlog, news and the papers are all published here first. Get them by email, or follow the feeds.
The newsletter is moving in-house and relaunching soon. Until then, email us to be notified when it opens, or follow the RSS feed.
Or follow the work directly: Research · Kernel devlog · News · CDI assessment