Chart of Works
Content
Projects, writing, links and ideas
Map
Explore islands and open their pages
An explorable map of my work
Chart of Works represents my projects, writing, links and ideas in an interactive world you can explore.
Read overviewTHE COMPLETE SURVEY
A field guide to my work.
Projects I’m building, writing, and links to my work elsewhere.
Projects/ 04
Seren
Input
A website URL
- 01Map the site
- 02Check the evidence
- 03Prioritise findings
Result
Prioritised security findings
Security reviews for AI-built websites
Seren maps a website, checks its security and explains which findings a founder should address first.
Read overviewTaffy
Input
An architecture draft
- 01Write a section
- 02Get Archie’s feedback
- 03Revise and export
Result
A revised Markdown document
Architecture documentation with AI review
Taffy is a macOS app for writing architecture documents, getting feedback from Archie and revising individual sections before a design review.
Read overviewTrellis
Example goal
Save $12,000
$400 / month30 months
$650 / month19 months
Illustration: fixed savings, no interest.Compare
Monthly savings and time to the goal
Personal finance planning and scenarios
Trellis compares the effects of different decisions on cash flow, investments and savings goals. Its planning tools run locally; an AI agent is in development.
Read overviewWriting/ 04
Agentic nines
Assigned work
Tasks with an agreed definition of success
- 01Define success
- 02Measure outcomes
- 03Account for rework
Operating cost
Accepted outcomes plus the effort to recover failures
Reliability measured in useful work
A proposal for measuring AI by useful outcomes and the work needed to correct failures, with evals building confidence in judgment across varied tasks.
Read articleEngineering context
Request
What you ask a model for
- 01Output
- 02Apps
- 03Good apps
- 04Context gathered
Result
What you need, using context the model already has
What comes after context engineering
Good results from AI models still depend on people supplying detailed context. I argue the next step is models that gather that context themselves, whether building an app or changing an enterprise codebase.
Read articleMeta-loop
AN UNWRITTEN NOTE
Meta-loop
This note is still unwritten.
Visit the noteOutcome engineering
AN UNWRITTEN NOTE
Outcome engineering
This note is still unwritten.
Visit the note