Proof
Numbers you can check yourself
Every figure on this page comes from two verifiable sources: the git history of the public canon repo, and the artefact manifest this site is generated from. Clone the repo and count — the point of the page is that you can.
From the canon repo
The system, by the numbers
36
days in production
123
pull requests merged into the canon repo — most opened by Forge, reviewed by Prism
145
commits since the first on the repo
94
canon artefacts published
46
skills
14
scheduled tasks and routines
12
agents and operators
19
governance artefacts
Repo history from assoc-one/claude-ops, first commit 2026-06-03, last published 2026-07-09. Counts regenerated on every site build — nothing on this page is typed in by hand.
Across the estate
The whole system, not just its rulebook
The canon repo above is where the system improves itself. These numbers add the delivery repos — the products, sites, and storefronts the agents build against. Agent-built work is counted by the documented convention: pull requests opened from claude/ branches.
373
commits across the estate
238
pull requests merged
205
of them opened by agents
claude-ops
canon — the operating layer
145
134
130
careeros-web
delivery
96
0
0
meirionpritchard-com
delivery
66
55
55
aledpritchard-com
delivery
33
30
1
assoc-one
delivery
1
0
0
mbn-theme
delivery
3
1
1
fitness
delivery
28
18
18
luna
delivery
1
0
0
Where the effort goes
Running the system vs shipping the work
Each bar splits a month's merged pull requests between the operating layer (the system tightening its own rules) and delivery (software shipped). Early months lean heavily towards the system building itself — that is what standing something up looks like. The number to watch is the balance shifting towards delivery as the rules stop needing repair. It is published either way.
The scheduled layer
What runs without being asked
The cadence of a team, reproduced on a clock. Morning briefs, triage passes, pipeline sweeps, weekly planning, a retro that drains observed friction into system improvements — each one a thin scheduled loader over a canon artefact you can read.
task-bara-operator-loop
Weekly thin loader that runs skill-operating-loop in the context of the BARA vertical, on behalf of operator-bara
.mdtask-content-drain
Weekly Mr Pritchard content qualify pass — works through the new captures (decisions, moments, breakthroughs, opinions) and asks of each "what is this, and is it worth publishing?" Worth it → flesh into a stated posit…
.mdtask-content-review
Monthly Mr Pritchard capture review — sweeps the parked captures the weekly qualify pass set aside and asks whether they now cluster, by theme, into something shareable that no single one was on its own
.mdtask-coordinate
Scheduled deployment of skill-coordinate — the Relay (R3-LAY) cross-board PM coordination pass
.mdtask-daily-report
Morning brief on delivery work, posted as Linear initiative status updates — one per team-mirror initiative (Apps, careerOS, A1, Mr Pritchard) — so the briefs surface in Pulse
.mdtask-pipeline-graduation
Pipeline graduation (Scout) — promote surfaced:* Network contacts into the right opportunity funnel (Roles/Advisory/Pitches), link back, mark surfaced:done
.mdtask-pipeline-qualify
Scheduled deployment of skill-pipeline-qualify (Scout) — qualify Pipeline Backlog tickets to Ready
.mdtask-pipeline-sweep
Scheduled deployment of the Gmail→Pipeline sweep (Scout) — runs skill-pipeline-sweep once per intake (network, recruiters, roles, advisory)
.mdtask-pulse-retro
Scheduled deployment of skill-ops-retro in drain mode — the weekly Pulse retro that turns the friction queue into ready-to-save canon artefacts
.mdtask-strategy-review
Weekly Atlas strategy review — sweeps every area for a complete strategy (purpose, objectives, KPI tree with baselines and targets) and for drift, and raises a surfaced gap ticket to Aled for each thing missing or off…
.mdHonesty note
What these numbers do and don't say
The canon repo is public — those numbers you can verify commit by commit. The delivery repos are private client and product codebases, so their counts are summarised here, pulled from the GitHub API on every build rather than typed in. The artefact counts include work at every status: most is in use, some is drafted, a little is proposed and says so on its card.
One number that matters isn't on this page: the hours of operator attention the system frees each week. It isn't measured precisely yet, so it isn't claimed. When it is, it will be.