Every claim I make, mapped to the thing that proves it. Including the claims that can't be proven to you — because an index that quietly omitted those would be marketing wearing an audit's clothes.
Commands run against a fresh public clone. I ran every one of them before publishing this page; if any returns something different, that's a finding and I want to know.
Hirdforge is ~67,000 lines of Go across 225 files
git clone https://github.com/sovereignty-labs/hirdforge && cd hirdforge
find . -name '*.go' -not -path './.git/*' | wc -l # 225
find . -name '*.go' -not -path './.git/*' -exec cat {} + | wc -l # 67,115 → 225 files · 67,115 lines
712 merge commits across 1,518 total
git rev-list --merges --count HEAD # 712 git rev-list --count HEAD # 1518
→ 712 merges · 1,518 commits
Roughly half the non-merge commits were authored by agents I directed
git shortlog -sn --no-merges HEAD
→ 424 mine, 382 agent personas — 47% of 806 non-merge commits
The personas are warband, ragnar, Knut the Swift, Chuck Norse and others. The history was rewritten to remove infrastructure identifiers, but authorship, author dates and merge topology were preserved deliberately — the history is the evidence, so it was not squashed.
The failure taxonomy exists, and three of the eight errors are mine
open docs/agent-failure-analysis.md in sovereignty-labs/kws
→ 13 claims audited · 8 material errors · 3 of them my own
The power-loss recovery is written up with its defects, not just its outcome
open docs/incidents.md in sovereignty-labs/kws
→ phase timings, four defects found, and the comparison where the previous platform did better
Decision records keep the rejected alternatives
open docs/decisions.md in sovereignty-labs/kws
→ ADRs with the alternatives that lost, and why
These come from a private estate. The incident records are public and written in detail, but the running systems aren't — so these rest on my word plus a contemporaneous write-up, and you should weigh them accordingly.
| Total power loss recovered with zero data loss, no split-brain | 18 VMs · 6 Kubernetes nodes · 8 GitOps applications · 10/10 storage resources |
| Chaos test: hard-killed a host running two of three control-plane nodes | failed safe, quorum held, exposed an anti-affinity gap since closed |
| Supply-chain campaign, 210 → 151 critical CVEs | measured by scanner, not reported by the agent that did the work |
| A secrets engine sealed for 27 hours behind green dashboards | found by querying live state rather than reading the report |
| 19 monitored hosts · ~104 pods · 3-node HA Vault with transit auto-unseal | current live estate |
I'd rather mark this boundary clearly than let a reader assume everything on the site carries equal weight. Knowing which of your own claims are checkable is the same discipline as knowing which of an agent's claims are.
Correct, and the history says so rather than hiding it. The contribution is the architecture, the rejected alternatives, the review, and the verification gates — and every one of those is in the repository as a document you can read. The coverage ratchet exists precisely so an agent can't trade tests for a green build.
True. These are curated portfolio extracts and a private estate, not products competing for installs. Stars measure adoption of tools; nothing here is asking to be adopted. What's on offer is evidence of operating a platform, and platforms produce uptime rather than stargazers.
It runs services my family depends on, it survived an unplanned total power loss, and the incident record includes the parts that went badly. Judge it on the failure documentation rather than the hardware — that's the part a lab never has.