Open counts across teams, oldest still-open issue per team, and net 7d flow. COD = bug fixing queue, BET = beta feedback, DEV = active feature work. Counts cap at 200 — '+' indicates more exist.

Open PR mix (humans vs Dependabot/Renovate), oldest open human PR, recent merges + commits, time since last release. Source: honely-app/honely.

Latest 50 workflow runs. Failing workflows surface the most-recent failure per workflow name — green when nothing is failing.

Wall-clock inference from the recent runs sample — split by runner label. Estimated; Blacksmith doesn't expose a public minutes API, so this is directional not exact. Blacksmith figure is vCPU-minutes (label vCPU count × wall-clock).

Local sessions parsed nightly into Convex. Token volume is the real metric — Max plan is a flat £180/mo so dollar figures are 'what this'd cost at API rates', useful for tracking subscription leverage as Anthropic adjusts caps. Live usage-vs-cap %: pending Anthropic exposing it (see GH issues 32796 / 44328).

Active, paused and recently concluded experiments. Source: vault/experiments/active.md — edits land via PR. Strategy doc + ticket both recommend vault-first for v1; promote to a Convex table if edit friction becomes the bottleneck.