Who watches the watchers?
Every model is trained on somebody's priors. Who's auditing the auditor when the auditor is opaque?
A deliberative platform where 127,000+ citizens vote on AI's power to slow research, deny loans, or weigh in on parole. Not a poll. A verdict.
If an AI can't reliably tell what's dangerous, should it have the power to decide what research gets slowed down?
Every model is trained on somebody's priors. Who's auditing the auditor when the auditor is opaque?
California convenience, Manhattan throughput, Lagos patience — only one of these ships in v1.0.
A "review hold" on the wrong preprint can bury a cure. A flag on the wrong grant can end a career. The cost of a false positive isn't abstract.
Open deliberation closes in 2d 14h 22m 47s. Every verified citizen counts once.
Move the sliders. Watch the public's trust in the algorithm react in real time. Compare your stance to the crowd median of 127,000+.
Every closed deliberation. Tagged by domain. Read the 200-word summary before you cast your next vote.
| Date↓ | Question↕ | Domain↕ | Result↕ | Margin↕ | Status |
|---|---|---|---|---|---|
| Oct 28, 2025 | Should AI gate dual-use research papers? Read the 200-word deliberation summary → | Research | Human oversight affirmed | 82–18 | Closed |
| Oct 25, 2025 | Can AI triage emergency room admissions? Read the 200-word deliberation summary → | Healthcare | AI authority expanded | 54–46 | Closed |
| Oct 22, 2025 | Should AI recommend criminal sentences? Read the 200-word deliberation summary → | Justice | Human oversight affirmed | 71–29 | Closed |
| Oct 20, 2025 | May AI deny mortgage applications? Read the 200-word deliberation summary → | Finance | Human oversight affirmed | 68–32 | Closed |
| Oct 18, 2025 | Should AI flag defense research for review? Read the 200-word deliberation summary → | Defense | Split decision | 49–51 | Closed |
| Oct 15, 2025 | Can AI screen peer-review submissions? Read the 200-word deliberation summary → | Research | Human oversight affirmed | 77–23 | Closed |
Anonymous by default. The role tag is what they chose to share. The vote count is what they actually did.
I came in thinking AI could triage risk better than tired reviewers at 2 a.m. After twenty-three votes, I now think the question isn't capability — it's consent. Who gave the machine that authority, and to whom does it answer when it answers wrong?
My grandfather trusted the village council, not the village computer. I want my AI the same way: advised, never sovereign. Voting here makes me feel like the council still meets — and that the minutes are public.
I build these systems. I see how confident a model looks on the dashboard and how brittle it is in production. Voting "no" on autonomous sentencing isn't Luddism — it's a Monday morning postmortem with a pager in your pocket.
When a child comes in coding, the algorithm can't see the mother's eyes. It can't hear the wheeze change between the waiting room and the bay. I vote for tools, not judges.
The same platforms that deny a campesino a loan will gate a dual-use paper. Power learns to dress itself in code. The Pulse is the one place I see the dressing being named out loud, in public, by everyone.
I'm seventeen and I outvoted three Fortune 500 CEOs on a defense question. That shouldn't be remarkable. It is. The Pulse is the first place my vote has the same weight as my parents', which is the only way a republic survives the algorithm.
One email. Weekly digest of the votes that mattered, the verdicts that shifted, and the questions nobody else is asking.
No spam. No surveillance. No algorithmic profiling of you — we promise. Read the Privacy Pledge.