About · The manifesto
“Open source is a graveyard of good intentions.”
Stephan Joseph, founder — posted publicly on X, August 19, 2026
He wrote it under a post by DHH — who had just sworn off monetizing his open-source project — arguing that good work dies without a funding structure to pay real people to build and maintain it. DHH replied, in full: “Stay tuned.” Within days, the Omacom Foundation was announced: roughly $12.6M to fund that project and the open-source work it depends on, for years.1
SaneCheck is that argument, built. Not for one project — for the record itself. The world’s corpora — the published literature, the archives, the canons — are full of good intentions and dead papers, and nobody’s job is checking them. Checking is now cheap. It was never funded. So we built the funding structure, the checker, and the first two campaigns, in that order of honesty: artifact first, ask second.
What we are
A fund
Braided strands — grants, compute credits, dollars — never commingled, aimed at one campaign slate, reported on one public ledger. The house holds the aim, not a pot of money.
A checker
SaneCite, open-sourced: it reads the source behind every citation and shows its evidence. Free forever. Anyone can re-run it. Meet it.
A method
The certification pass: enumerate everything, check everything, publish the ledger. Proven on Campaign Zero — 328 works, 7,504 passages, about $449 in machine cost.2
How we keep ourselves honest
- Whole corpora, never cherry-picked targets. The moment a checker aims at a pre-selected villain, it becomes the thing it was built to fight.
- Methods published before the run. Raw outputs published with the result. If you can’t re-run it, it’s an opinion, and we label opinions as opinions.
- Flags are flags, not verdicts. A citation error is not fraud. We never print it as fraud. The villain is a behavior — prestige published without verification — never a person.
- Right of reply, attached to the record. Anyone named in a result can answer it, in the record, permanently.
- No single funder above ~25% of a contested campaign. No governance token. No exceptions that “just this once.”
- Sunset rule. A campaign that misses its published completion state twice loses eligibility. No zombie campaigns — we of all people don’t get to become one.
- Our own numbers get checked too. Every figure this house publishes ships with its source and its re-run path. If one of our results fails a re-run, we say so at the top of the page — not in a footnote.
Don’t cry. Check.
The literature is not going to fix itself, and outrage has a perfect record of fixing nothing. Checking is a job. Jobs get funded. Fund this one.
Sources — every number on this page
- The exchange is public record: X, August 19, 2026 — Stephan Joseph’s reply in DHH’s thread on funding Omarchy past its founder; DHH’s reply, “Stay tuned,” the same day. The Omacom Foundation was announced within days (August 2026) with ~$8M in founding checks from technology operators, reaching ~$12.6M shortly after, including 1Password and 37signals ($600K), to hold trademarks, fund infrastructure, and fund the open-source projects Omarchy depends on.
- Campaign Zero case file: Via Patrum — 328 works of the Church Fathers translated and certified against the Greek and Latin originals; 7,504 passages; full-catalog synthesis ≈$449 at Cloudflare list prices (October 2026).