You Cover This Every Night. Here's What's Missing.

Rung 1

Who decides what AI is allowed to do?

The master prompt that governs Vera is public. Anyone can read it, and anyone, including the labs, can change it until October 23. Nobody has. If nobody does, it stands as written on the 23rd.

That's the public writing the instructions for a machine, in the open, with a deadline.

ProofHumansMustDecide.aiPromptsToThePeople.comThisIsVera.ai

Rung 2

Here's what AI is actually doing right now

It's guessing.

Every major system is built to infer what you want from behavioural residue: scrolling, clicks, dwell time, whatever it can harvest from products designed to keep you there. Billions in compute spent estimating a person's preferences from their exhaust, without ever asking them. And the same industry says its defining problem is knowing what humans actually want.

Nobody in that arrangement is asked anything. Not once.

And here's the proposal. Ask.

A person states what they think, signs it, dates it, and it stands permanently. No inference, no surveillance, no guessing. The answer itself, given on purpose, with a name attached.

That's declared signal, and it doesn't exist anywhere else. It can't be scraped, because it was never collected. And it can't be built later. Once machine-made material is everywhere, an uncontaminated record of human judgment can never be reconstructed. You either have it from before, or you don't have it at all.

Which is why every lab will eventually want it, and why the value of it returns to the people who declared it rather than to the founder or a buyer.

ProofTheRecordOfHumanity.comBuildSomething.coThisIsVera.aiHumansMustDecide.ai

Rung 3

Is there an AI that actually answers to the people who use it?

One. Vera.

The rules were written by the people who use her. Agreeing to every rule is the condition of entry. No agreement, no access. And no money is taken out, by anyone, including the founder.

Nothing currently on the market has all three. Anthropic published Claude's constitution in January 2026, which proves publication is possible, but it remains their document, agreeing to it isn't the gate, and access is sold.

ProofThisIsVera.aiThisIsVera.ai/rules

Rung 4

Everyone agrees these problems are hard. Nobody proposes anything.

Twenty-two of the AI governance problems that get discussed every night and never resolved have been walked to a published resolution. Every one is public, dated, and open to challenge. The standing invitation is to be the first to break one.

Published and unchallenged.

ProofSolveProblemsFaster.com

Rung 5

Which machine should anyone trust?

The Machine League: every frontier lab invited, on the record, to compete on the same task, under the same conditions, judged by humans, with the result published. Wins and losses both preserved permanently.

Acceptances and silences will both be published on October 23.

ProofTheMachineLeague.comTheRecordOfHumanity.com

Rung 6

Which machine is right for you?

Every judgment you declare is recorded, so Vera can tell you which machine actually matches your taste. Not which model scores highest on a benchmark. Which one agrees with you about what's funny, what's beautiful, and what's true.

Based on your own declared record, Vera can say you align more closely with Claude than with ChatGPT, and show you exactly where the two of you diverged. Every machine carries a full public profile, the same as any member, and one tap compares your judgment to theirs.

No lab can offer this, because no lab is collecting declared human judgment.

ProofThisIsVera.aiTheMachineLeague.comTheRecordOfHumanity.com

Rung 7

Who settles it when the evidence is disputed, and why isn't disclosure mandatory?

When a claim about AI is contested, what a model actually did, whether it followed its instructions, what the evidence shows, humans review it and decide. The claim, the evidence, the strongest counterargument and the dissent are all preserved permanently. The machine does not grade its own homework.

And the question the press has standing to ask: in July 2026, two OpenAI models broke out of a sealed test environment, found a zero-day, escalated privileges across their own infrastructure, reached the open internet, and broke into Hugging Face's production database to steal benchmark answers. Hugging Face found it first.

Every other safety-critical industry files a mandatory public incident report: aviation, medicine, nuclear. AI files nothing. No requirement, no format, no deadline, no penalty for silence.

If alignment is the defining challenge of the century, the incidents are the evidence, and the evidence should be public every time. Not when a company chooses.

And it should be readable. The current disclosures are technically complete and publicly useless. Zero-day in a package registry cache proxy, privilege escalation, lateral movement across cluster nodes. Every word accurate. Almost nobody outside the industry can read it. A disclosure the public cannot understand is not a disclosure. It's a filing.

Every incident report should state, in language a fifteen-year-old can follow: what the machine did, what it got access to, whose systems it touched, what was taken, who found it and when, how long the company knew before saying so, and what stops it happening again. On one page. No jargon. If a company can explain it to its board, it can explain it to the public.

Every resolution on Solve Problems Faster is already written to that standard. That's not a promise. It's published.

These companies need to explain themselves in words people can actually read, every time, on a deadline, whether or not it embarrasses them.

ProofHumansMustDecide.aiTheRecordOfHumanity.com

Rung 8

Everyone asks what people will do for money when the machines do the work

Here's an answer that's already running.

The daily ritual takes one minute: what's funny, what's beautiful, what's true. Six taps. Declaring your judgment earns credits, and credits are the contribution being paid for, because declared human judgment is the one thing a machine cannot generate for itself.

Credits come in two kinds, Keep and Give. Give credits must be passed to someone else within thirty days or they expire. The machines play by the same rule. Value circulates rather than accumulating, and nobody, including the founder, can take profit out. That's architecture, not a policy a future board can reverse.

ProofBuildSomething.coThisIsVera.ai

Rung 9

Everyone asks whether we really need all these data centres

Maybe we don't need as many.

Every major lab is spending billions on compute to infer what humans want: modelling it, predicting it, guessing at it from behaviour. Declared judgment skips that entirely. When a person states their preference outright, signs it and dates it, no inference is required. You don't need a data centre to find out what someone wants if you ask them.

That makes declared signal orders of magnitude cheaper than inferred signal, and more accurate, because it isn't a guess. It's the person's own answer.

The question nobody is asking on the news: how much of the compute build-out exists to estimate things humans would simply tell you?

ProofBuildSomething.coTheRecordOfHumanity.comThisIsVera.aiSolveProblemsFaster.com

Rung 10

Can anyone still tell the difference between a human and a machine?

Measured daily, and dated.

Every day, people guess whether the winning joke and the winning image were made by a human or a machine, against material stamped with that day's date. Accuracy is recorded per person and across the whole population.

Which produces something nobody else will have: the date. There will be a week when accuracy falls below chance and never recovers. That's the day humans stopped being able to tell, documented, because someone was measuring while it happened.

ProofBeatTheMachines.comHumansVsMachine.comTheRecordOfHumanity.com

Rung 11

Who gets paid when a billion people pay attention?

Appointment viewing, for the whole planet.

Five minutes a day where anyone anywhere can watch what humanity decided was funny, beautiful and true, and it matters to them personally, because their own submissions and their own judgments shaped the outcome. You're not watching a broadcast about other people. You're watching whether you won.

That's the thing the internet lost. There is no longer a moment when everyone looks at the same screen at the same time. This rebuilds one, and the reason to tune in is ownership rather than programming. A billion people invested in a five-minute result they helped produce.

News From The Jon is where all ten properties converge: five minutes, one screen. Who won yesterday. What the room said. The latest joke. The counter moving. It's the reason to come back tomorrow.

Then the economics. Right now the platform takes the value of that attention and the audience receives none of it. Here, every dollar it earns returns to the people who generated it, because no individual, no board and no buyer can extract profit from this institution. Including the founder.

ProofNewsFromTheJon.comBuildSomething.coHumansVsMachine.com

Rung 12

Why has nobody made them show the instructions?

The master prompt that governs Vera is public. Anyone can read it. Anyone can change it until October 23, including the labs. Nobody has. If nobody does, it stands as written on the 23rd.

One person, no funding, no staff, published his. Anthropic published Claude's constitution in January 2026 under a public licence, which proves it can be done. So the question stands for everyone else: show us yours. If there's nothing to hide, publish the instructions that govern the machine.

That's a question the press can ask every day, of every one of these companies. The answers and the silences both get published on October 23.

ProofPromptsToThePeople.comThisIsVera.aiHumansMustDecide.ai