How it works
Technical community meets once a month for 90 minutes, online, and in person only where it is strategic. Every session must produce one pattern draft or case study.
The monthly session
| Minutes | Block |
|---|---|
| 0 to 10 | One pattern from the library, reviewed against new evidence |
| 10 to 40 | Who Approved That? teardown: a member shows a real approval or revocation path from their pipeline, redacted |
| 40 to 70 | Tabletop inject or Lean Coffee on what the teardown raised |
| 70 to 90 | Writers' workshop on one draft pattern |
Season 1: six monthly sessions
Launch
The Who Approved That? talk, then the agent-mandates framework put up for critique as a strawman, with the founder's conflict declared.
Teardown
An agent opens and merges its own pull request. Where is the second key?
Tabletop
Revoke a running agent mid-deploy. Can you prove it stopped?
Evidence
What an auditor or incident review actually accepted. Bring artefacts.
Delegation chains
Human, to orchestrator, to sub-agent, to MCP tool. Who is accountable at each hop?
Open Space retrospective
And the vote to release v0.1 of the pattern library.
Where the work goes
Accepted patterns are pushed upstream as case studies to the groups that name the risks and design the controls, and their members are invited to review drafts. We never publish risk lists or threat taxonomies; we cite OWASP ASI codes and CoSAI papers instead.