Gloriousflywheel Bazel Dogfood Tranche 1 Status 2026 04 22

GloriousFlywheel Bazel Dogfood Tranche 1 Status

Correction on 2026-04-24: This note is preserved as historical tranche-status evidence from before the pooled-substrate reset reframed the product contract. It is not the active execution surface. Read older tranche language here as historical proof context, not as the current top-level plan.

Date: 2026-04-22

This note records the live status of the first authoritative Bazel dogfood tranche after the source-repo proof package and lab benchmark canary landed.

Why This Note Exists

The tranche-1 set is no longer just a planning shape. It now has enough real proof to describe the remaining work honestly:

  • the source repo has a bounded Bazel cache-first proof path on main
  • lab has a real benchmark and cache-evidence canary on tinyland-nix
  • two repo-owned JS/Bazel package consumers are already green on canonical default branches
  • the downstream tinyland-inc/scheduling-kit mirror is no longer red for stale workflow drift
  • tinyland.dev is not missing runner truth; it is an intentional hybrid-by-policy boundary

That means the remaining gap is not “does tranche 1 have any proof at all?” The gap is whether the proof is routine, measured, and honest across mirrors, hosted exceptions, and product-facing repos.

The repo now also has an operator path for that question through just bazel-dogfood-tranche-status.

That operator path is now paired with a hosted routine path through the Tranche Proof Status workflow, which renders the same bundle on main, through workflow_dispatch, and on a daily schedule with text and JSON artifacts. That hosted path now uses owner-scoped GitHub App tokens for tinyland-inc and Jesssullivan as its sole auth contract, rather than keeping the older broad tranche token as a standing compatibility bridge.

The source-repo proof itself now also has a dedicated routine workflow through Source Bazel Proof on main, workflow_dispatch, and a daily schedule, so the tranche anchor no longer depends on surviving churn inside the wider Platform Proof matrix.

Tranche 1 Repo Status

tinyland-inc/GloriousFlywheel

  • proved as the source-repo cache-first package
  • landed via #324 on main
  • proof command: nix develop -c bazel build --config=ci-cached //:deployment_bundle
  • role in tranche 1: anchor the product claim on the source repo before downstream expansion

tinyland-inc/lab

  • proved as the benchmark and cache-evidence canary
  • landed via tinyland-inc/lab#148
  • authoritative proof: workflow run 24764027904 completed green on tinyland-nix for bazel-fast-tests shard 1
  • role in tranche 1: benchmark, cache-evidence, and runner-capacity truthing surface

Jesssullivan/scheduling-kit

  • canonical repo default-branch CI is already green
  • authoritative proof: run 24608528050 on main at 5653e2dd91425ad15af755a928bb4b32440ca36b completed success
  • uses the shared JS/Bazel package contract on the canonical repo
  • mirror-integrity proof also now exists: tinyland-inc/scheduling-kit#12 merged as 7899a7e, and downstream mirror CI run 24765310127 completed success on main
  • role in tranche 1: repo-owned shared JS/Bazel package consumer

Jesssullivan/acuity-middleware

  • canonical repo default-branch CI is already green
  • authoritative proof: run 24645985930 on main at 2a7437fe58307a3afbe865dd16d078aca44d34ff completed success
  • explicit publish-path proof also now exists: Publish Package dry-run run 24765511512 completed success on main@2a7437f
  • both CI and publish paths use the shared JS/Bazel package workflow with Bazel target //:pkg
  • role in tranche 1: repo-owned shared JS/Bazel package consumer

tinyland-inc/tinyland.dev

  • not a converged runner-authority package consumer
  • current main reads as hybrid-by-policy: the Nix and Attic builder path runs on civo-k8s, while Bazel CI, package publication, container publication, and staging deploy remain hosted by policy
  • role in tranche 1: keep the productization board honest about a real hybrid boundary inside the active repo set rather than pretending every tranche-1 repo is the same proof class

What Changed

Tranche 1 no longer needs to be described as “one source proof and a list of future candidates.”

It now includes four different proof classes:

  • source-repo Bazel proof
  • benchmark and cache-evidence canary
  • repo-owned JS/Bazel package consumers
  • downstream mirror integrity
  • hybrid-by-policy boundary repo

That is enough to move productization work from definition into contract hardening.

Remaining Gaps

  • make package-lane proof routine rather than dependent on manually reading green default-branch history
  • keep mirror and downstream surfaces honest when they diverge from canonical repo truth
  • keep proof state operator-readable through a live status command rather than relying on scattered issue comments
  • keep tinyland.dev classified as hybrid-by-policy until its hosted publication and deploy boundaries genuinely change
  • add stronger measurement and scorecard discipline so the tranche reads as cache-first product evidence, not just a collection of green repos

Next Moves

  • keep the source-repo Bazel package explicit and green on main
  • preserve lab as the benchmark and cache-evidence canary
  • decide whether stale downstream mirrors like tinyland-inc/scheduling-kit belong in the productization lane or in a separate mirror-integrity cleanup
  • only widen tranche-1 claims after the evidence becomes routine enough to be operator-readable without repo-by-repo archaeology

GloriousFlywheel