GloriousFlywheel Orgwide Enrollment Reality Check 2026-04-18

GloriousFlywheel Orgwide Enrollment Reality Check 2026-04-18

Snapshot date: 2026-04-18

Purpose

Re-truth the current adoption surface after the recent source-repo and template-work rollout.

This note is intentionally narrower and more current than the 2026-04-17 adoption stack:

  • it uses live GitHub state from tinyland-inc and Jesssullivan
  • it focuses on current workflow enrollment, not just repo existence
  • it separates real ARC and repo-owned runner use from template noise, hosted-only paths, and raw compatibility runners

Method

The scan used the recent non-fork repo surface:

  • tinyland-inc: 118 non-fork repos, 58 pushed in the last 30 days
  • Jesssullivan: 147 non-fork repos, 30 pushed in the last 30 days
  • combined recent non-fork surface: 88 repos

For those 88 repos:

  • inspect .github/workflows on the default branch
  • mark repos with no Actions workflows separately
  • inspect workflow text for:
    • GitHub-hosted runners like ubuntu-latest
    • shared ARC labels like tinyland-*
    • repo-owned runner variables like PRIMARY_LINUX_RUNNER_LABELS_JSON
    • raw compatibility surfaces like literal self-hosted
    • reusable workflow consumption from tinyland-inc/ci-templates
    • explicit cache/bootstrap hooks such as setup-flywheel and nix-setup

One important exception was checked separately:

  • Jesssullivan/MassageIthaca still defaults to main on GitHub, but the current product authority branch is dev/main
  • Jesssullivan/MassageIthaca#189 merged to dev/main on 2026-04-17
  • that branch does use PRIMARY_LINUX_RUNNER_LABELS_JSON

So the default-branch census is the right baseline for orgwide truth, but it slightly undercounts that one product repo.

One more authority caveat also matters:

  • Jesssullivan/XoxdWM is the non-fork repo in the downstream denominator
  • but key self-hosted fast paths are currently gated to the tinyland-inc/XoxdWM fork instead
  • and both repo surfaces currently report 0 accessible Actions runners

Topline Numbers

Across the 88 recent non-fork repos:

  • 47 have no GitHub Actions workflows at all
  • 41 do have workflow files

Across those 41 workflow repos:

  • 20 are hosted-only on the scanned branch
  • 8 consume tinyland-inc/ci-templates in some form
  • only 2 clearly use repo-owned runner vars on the scanned branch:
    • Jesssullivan/scheduling-kit
    • Jesssullivan/acuity-middleware

The first-pass “mentions shared or self-hosted” scan overstates enrollment, because it includes:

  • the source repo itself
  • the template repo itself
  • repos that only call hosted reusable workflows
  • repos with compatibility or gated self-hosted paths that are not their normal authority path

Real Enrollment Versus Noise

Repos With Real Shared ARC Or Repo-Owned Lanes On The Default Branch

These are the live recent repos where the default-branch workflow surface actually shows GloriousFlywheel-backed or repo-owned runner usage:

  • tinyland-inc/GloriousFlywheel
  • tinyland-inc/blahaj
  • tinyland-inc/lab
  • tinyland-inc/tinyland.dev
  • Jesssullivan/scheduling-kit
  • Jesssullivan/acuity-middleware
  • Jesssullivan/yt-text

That is:

  • 7 of 41 workflow repos if the source repo is included
  • 6 of 39 workflow repos if GloriousFlywheel and ci-templates are removed from the enrollment denominator

That is roughly 15% downstream enrollment on the active workflow surface.

Product-Branch Exception

Jesssullivan/MassageIthaca is not cleanly represented by the default-branch scan.

Live dev/main now contains:

  • runs-on: ${{ fromJSON(vars.PRIMARY_LINUX_RUNNER_LABELS_JSON || '["ubuntu-latest"]') }}
  • beta and alpha Playwright validation on that same repo-owned runner path

So if the active product branch is counted, downstream enrolled repos become:

  • 7 of 39

That is still only about 18%.

The important point is not that MassageIthaca is missing. It is that branch authority is fragmented enough that a default-branch census and an active-product-branch census do not say the same thing.

Broken Activation Surface

Jesssullivan/XoxdWM is also not cleanly represented by the default-branch non-fork scan.

Live main on the Jess repo still defaults key paths to hosted runners, while the self-hosted fast path is gated to github.repository == 'tinyland-inc/XoxdWM'. But the more important update is that the Jess repo now has USE_SELFHOSTED=true, and both the Jess repo and the tinyland fork currently report 0 accessible runners via the repository runners API.

So XoxdWM should not currently be counted even as a broader cited-repo runner exception. It is better described as a broken self-hosted activation surface.

What Is Still Mostly Hosted

The recent active workflow surface still has a large hosted-only block.

Confirmed hosted-only examples on the scanned branch:

  • Jesssullivan/MassageIthaca on main
  • Jesssullivan/jesssullivan.github.io
  • Jesssullivan/tinyland-huskycat
  • Jesssullivan/server-fan-splitter-pcb
  • Jesssullivan/modular-atx-power-rack
  • Jesssullivan/zig-notify
  • Jesssullivan/hiberpower-ntfs
  • Jesssullivan/zig-ctap2
  • Jesssullivan/re-ems
  • Jesssullivan/zig-keychain
  • Jesssullivan/zig-crypto
  • tinyland-inc/tinyland-auth-pg
  • tinyland-inc/crs310-8g-2s-in
  • tinyland-inc/tinyland-auth-redis
  • tinyland-inc/huskycat-gl
  • tinyland-inc/network-topology
  • tinyland-inc/waifu-mirror

This is the blunt reality check:

  • there is still a large body of recent org activity that does not touch ARC at all
  • some of that is fine because those repos are small or incidental
  • some of it is directly in tension with broad “all our repos should use GloriousFlywheel runners and caches” language

Template Reality

The template layer still creates overcount risk.

ci-templates Is Not Equal To Enrollment

Recent repos consuming ci-templates include:

  • tinyland-inc/tinyvectors
  • tinyland-inc/vite-plugin-a11y
  • tinyland-inc/tinyland-rate-limit
  • tinyland-inc/tinyland-physics
  • tinyland-inc/tinyland-color-utils
  • tinyland-inc/tinyland-a11y-engine
  • Jesssullivan/acuity-middleware
  • Jesssullivan/scheduling-kit

But these do not all mean the same thing.

Hosted Template Consumers Still Exist

The following recent repos call hosted reusable workflows and should not be counted as runner-enrolled:

  • tinyland-inc/tinyvectors
  • tinyland-inc/vite-plugin-a11y
  • tinyland-inc/tinyland-rate-limit
  • tinyland-inc/tinyland-physics
  • tinyland-inc/tinyland-color-utils
  • tinyland-inc/tinyland-a11y-engine

tinyland-inc/tinyvectors now explicitly sets:

  • runner_mode: hosted

The other five repos above currently use npm-publish.yml, and that workflow is still GitHub-hosted.

So the existence of template consumption alone still overstates runner enrollment, even after the newer V2 template work.

Real Repo-Owned Template Consumers

The clean current template-backed repo-owned proofs are:

  • Jesssullivan/scheduling-kit
  • Jesssullivan/acuity-middleware

Those repos pass runner_labels_json: ${{ vars.PRIMARY_LINUX_RUNNER_LABELS_JSON }} through the shared package workflow on main.

Important Repo Corrections

betterkvm

betterkvm should not be counted as current ARC enrollment.

Its ci.yml still carries the explicit hosted pin:

  • NIX_RUNNER: ubuntu-latest

It remains a migration candidate, not a live dogfood proof.

tinyclaw

tinyclaw is still GitHub-hosted on its scanned branch.

It should remain outside runner-adoption claims.

remote-juggler

remote-juggler is still overwhelmingly GitHub-hosted on the scanned branch.

It is better described as a hosted CI repo with some Nix/cache-adjacent work, not a current GloriousFlywheel runner consumer.

XoxdWM

XoxdWM is not a clean non-fork default-branch enrollment:

  • many paths on Jesssullivan/XoxdWM still fall back to ubuntu-latest
  • the fast self-hosted path is real, but is gated to tinyland-inc/XoxdWM
  • Jesssullivan/XoxdWM now has USE_SELFHOSTED=true
  • both repo surfaces currently report 0 accessible runners
  • hardware paths still use raw [self-hosted, honey] on that fork-shaped path

It is better described as a broken self-hosted activation surface, not as a clean homogeneous contract.

yt-text

yt-text is one of the clearest live downstream shared-runner canaries, but it is still hybrid because docs and fuzz paths escape to hosted runners.

tinyland.dev

tinyland.dev still has meaningful ARC usage, but its authoritative CI surface is not homogeneous:

  • nix.yml uses tinyland-nix
  • bazel.yml, container.yml, publish.yml, and staging deploy paths remain hosted

It should be described as hybrid, not as a full platform-converged product repo.

Updated Claim-Versus-Reality Read

True

  • the source repo now has a real default-branch proof path and is green after GloriousFlywheel#234
  • the template contract work landed in tinyland-inc/ci-templates#7
  • Jesssullivan/scheduling-kit#64 and Jesssullivan/acuity-middleware#50 are real repo-owned downstream enrollments on main
  • Jesssullivan/MassageIthaca#189 did land on dev/main
  • Jesssullivan/yt-text remains a real shared-runner canary on main

Not Yet True

  • it is not true that recent orgwide workflow activity is mostly on GloriousFlywheel-backed runners
  • it is not true that ci-templates consumption implies runner adoption
  • it is not true that most active tinyland-inc and Jesssullivan workflow repos are on shared ARC lanes
  • it is not true that MassageIthaca is cleanly represented by its default branch today

Honest Bottom Line

If the question is:

“How close are tinyland-inc and Jesssullivan to the broad GloriousFlywheel claim on the active repo surface?”

the honest answer is:

  • the source platform is now more credible than it was on 2026-04-17
  • downstream enrollment is still narrow
  • recent non-fork repo activity is much wider than runner enrollment
  • the workflow surface is still dominated by no-workflow repos, hosted-only repos, template noise, and hybrid paths

On the active recent workflow surface, downstream enrollment is only about 15% on default branches, or about 18% if the MassageIthaca dev/main authority branch is included.

There is no defensible broader cited-repo runner ceiling above about 18% until XoxdWM has real accessible runners again.

That is progress. It is not orgwide convergence.

What Should Be Claimed Next

The safest current claim is:

  • GloriousFlywheel is now a live and credible source repo plus a small but real downstream runner platform
  • the first clean repo-owned downstream package lanes are live
  • the shared ARC canary set is real but still hybrid
  • one named hybrid repo still has a broken self-hosted activation surface
  • orgwide enrollment is still in early rollout rather than broad adoption
  • enrollment reporting should use separate runner-authority, template-mode, and branch-authority-exception counts rather than one blended repo number

That framing is accurate and defensible.

GloriousFlywheel