Calendar: couch surfing availability and claims at /calendar
Source: chat 2026-11-04
Get the default test run under 2 minutes
The default test run takes about 15 minutes, and 15 of its failures also happen on main in a cloud workspace, so every task ends with a long wait and a manual comparison.
Parallel full test run (--jobs, fulltest)
Add a parallel test runner: ./v3 test --all --jobs=N and a ./v3 fulltest alias, for the maintainer's periodic full run (10-core machine).
Smoke test set for per-task runs
Add a curated smoke test set so each task runs only the smoke set plus the tests for what it touches; the full suite moves to a periodic maintainer run (./v3 test --all). Update AGENTS.md and contributor docs with the process.
Fix tests that already fail on main
Fix the 15 tests that fail on main in a 2-core cloud workspace: 8 FrmClientTest GPG-agent failures plus ApprovalSeedCommandTest, LocalAppSmokeTest (3), ReadModelBuilderTimingTest, ReactionRecordFormatParityTest and TerminalOperatorUiCommandTest. A red default run currently needs manual comparison against main.
A /work page with status at a glance
Right now you can only see where work stands by opening each #work thread or reading /work.txt, and the only way to change status is typing a status: reply.