2026-11-03 — The product thinker's rulings for an experiment testing a long autonomous run (given 2026-11-03): during the experiment a release is cut whenever every gate passes, and any refusal waits for the product thinker; each API key has a hard monthly cap and no separate daily limit; sessions use the person's subscriptions first and move to a key when a subscription window reaches 95%, returning once it resets; the run may use the API keys the person has named to it; agents may plan a draft whose open questions each have one defensible answer on the record, and hold every other draft for the person; and the run keeps the person informed through one private daily digest page (what merged, what stalled, spend per key, what waits for them).

Frontmatter

.abcd/work/decisions/dcn-2611031615094092-the-product-thinker-s-rulings-for-the.md
open on GitHub ↗ · commit history ↗

mentions7

• dcn-26110317200123992026-11-03 — The run builds on the person's subscriptions: the build loop's roles are routed to the host, so each stage runs as a sub-agent of the coordinating session on its subscription, since the claude runner's --bare launch always bills a key; the keys stay a fallback at 95% of a subscription window, per dcn-2611031615094092. The main work moves to the account that always runs the latest build, and a second account keeps the latest release.
• iss-2611031731385043record-lint's prose_citation_resolves refuses a record id whose record exists, minted, on another open pull request's branch and not yet on main: on 2026-11-03 a decision citing dcn-2611031615094092 (in #1381, queued) failed the handover PR's preflight, and the easy way past was to cite the ruling in words, which drops the id and with it the trace (the product thinker ruled ids must always be used). The forward-looking marker keeps the id, but it was written for a record not minted yet, and nothing removes it once the record reaches main, so stale markers accumulate and the next reader cannot tell a pending citation from a dead one. <!-- record-lint: forward-looking -->
• iss-2611050840341934A major record captured mid-cycle refuses the next release cut (GuardFindings), while a drain under the baseline rule takes only nitpick and minor, so on a scheduled cut with no one present one such record holds the release. The ways past it, a fix, a wontfix or abcd capture defer, are each a judgement no unattended stage is given. dcn-2611031615094092 settled this for one experiment only, ruling that a refusal would wait for the product thinker and that a daily digest would list what waited for them. For scheduled cuts beyond that experiment, nothing says who is to make the judgement.
• itd-2611031342011261abcd keeps a per-session ledger of every agent it launches
• itd-2611040344170206Unattended work runs on a named profile, one per way of paying
• itd-2611041549483696An unattended run writes one private daily digest page
• spc-2611041701500441abcd keeps a per-session ledger of every agent it launches