Operator runbook.

morning · standup

1. /desk — single-screen briefing.
   • Are all four services green? If not, drill into the red chip's page.
   • Is realized PnL today positive or negative? Drill to /bot for the attribution breakdown.
   • Is the data lag inside the 24h band? If not, check /data for the slow event.
   • Stale-thesis count = N? If > 0, hit /writing/stale to triage.
   • Stale-cite debt > 0? Hit the at-risk strategies list to update citations.
   • Recent-visits card shows your most-recent detail-page hops — pick up where you left off.
2. /promotions — what got decided overnight. Filter by ?q=active and sort visually by citer-count (already pre-sorted desc); newly-superseded notes are flagged with the chain to their replacement.

intra-day · investigations

1. "Why did this trade happen?" —/bot's fills + provenance, or hit a specific order's citation set via /api.
2. "What features are pulling weight?" —/features ranked by best entry-score. Filter by prefix; click a row for its detail page.
3. "Which strategies use feature X?" —/features/<X> lists every strategy that rides it. Click any strategy slug → its catalog entry.
4. "Tell me about strategy Y." —/strategies/<Y> — provenance, feature chip-strip, cited notes, leaderboard entries, verification link.
5. "Is strategy Y's hypothesis still defensible?" — check the stale-cite banner on its detail page (auto-flag if any cited thesis went superseded). Cross-reference/strategies/compare for the A/B against baseline.
6. "I have a hypothesis I want to test." —/sandbox for browser mechanics on synthetic data;/backtest for the real engine. Result is good? Cross to/strategies/compare for the verified A/B.

end of day · hygiene

1. Did a thesis you cited turn superseded today? Find the strategies pointing at it on /strategies(filter by the slug); decide whether to re-cite or kill.
2. Any agent-proposed run land? /agents/runs has the leaderboard; promote-worthy slugs need a named-human gate per ADR-0019.
3. Open new /writing note for anything you learned that doesn't fit an existing thesis (status: exploratory until you're sure).

weekly · review

1. Sweep /promotions'sexploratory bucket — anything in there for > 30 days? Decide active or kill.
2. Sweep the leaderboard movement on/features: any feature dropped > 50% in best score? Investigate decay.
3. Sweep /strategies for any stale-cite warning chips; resolve.
4. Sweep /writing/stale for review-overdue notes; decide active / refresh / supersede.

keyboard · operator shortcuts

Hit ? for the full overlay. Key two-key chords (vim-style):g d → desk,g f → find,g s → strategies,g e → features,g p → promotions. On any list/leaderboard page,/ focuses the filter input.