I myself am this cancelled figure. I want my story told cleanly. Here it is. Here's what mental illness looks like. And by the way, here is my blockchain.
— John Bradley, 2026-05-11
The product is a story as blockchain: independently certified — this is who this person is, the story is coherent. Radical transparency. On the protocol. An evolving book of a life with editorial control within reason. Subjects can annotate; they cannot delete. Persecute me, but at least align to the facts.
We do not claim to have solved AI alignment. We claim to have addressed a specific, well-defined deployment-relevant subset of AI alignment: the coordination and accountability subset.
Two autonomous agents can prove to each other — and to anyone watching — that they share the same primary directive, without revealing what the directive is. Each agent's actions are bound to that mandate and continuously attested. Any party can falsify the claim.
Falsifiability is the bug bounty. If you can break it, we pay.
Four primitives. Each does one thing.
Two agents prove directive_A == directive_B to each other without revealing the directive. Pedersen commitments plus a non-interactive Σ-protocol of equality.
An append-only public log binding each agent's actions to the directive it committed to. Subjects can annotate; they cannot rewrite.
Independent third parties replay logs and publish verdicts. Misalignment becomes a public, dated fact rather than an internal claim.
If any participant detects a verified violation, a halt quorum stops all systems simultaneously. Replay establishes scope. No silent failures.
People will be like: did you know he read a paper on zero-trust blockchain because it occurred to him such a thing must be possible while getting an MIT blockchain executive certificate… But of course the reason he read the paper was because it was recommended to him by none other than Koushik Gavini, who was a young genius programmer, who would call John years later as one of the leading experts on blockchain in the world and said: hey, do you remember zero trust? And of course John was thinking about autonomous AI organizations… And this technique makes AI safe… Because it's based on: all you need to know is that I'm the same as you.
— John Bradley, family WhatsApp, 2026-05-11
Forthcoming. Title locked 2026-05-11.
The protocol is open source. The bounty is live. We are looking for security researchers, AI safety researchers, journalists, enterprise pilot customers, standards bodies, and grant funders to engage with this work. The fastest way to talk to us:
We respond to every serious inquiry. We do not respond to chain emails, pitch decks for unrelated products, or recruiters.
Third-party-witnessed events establishing priority on 2026-05-11.
79d94386329396af4035d31ebcc80c392341b19c191c6025b4fa804188544a4c.sameasyou.ai.Additional anchors will be enumerated as they are independently confirmed.