Personas
Twelve who disagree on purpose
Archetypes calibrated to skepticism, not flattery. No ‘great idea!’ sycophancy. The CFO will ask about unit economics. The PM will ask why now.
12voices
Private beta · pricing your delusion since 2026
Stress-test any idea againsttwelve fictional customersbefore you commit a quarter.
Submit a one-line idea. We pull live market context, cast a dozen dissenting personas, run a multi-turn debate, and hand back a brutally honest verdict — usually inside a minute. Cheaper than a co-founder argument; faster than a customer interview.
- Personas
- 0
- Debate turns
- 0
- Median runtime
- 0s
VERDICT73AI legal assistant for small firmsVERDICT41crypto wallet for grandmasVERDICT71AI triage for ER waiting roomsDEBATINGvoice-first CRM for solo realtorsVERDICT76subscription bookkeeping for creatorsVERDICT41spotify for short-form audio dramasVERDICT80compliance copilot for clinical trialsVERDICT55playable resume sites for designersDEBATINGcarbon credit marketplace for SMBsVERDICT49async standup for distributed teamsVERDICT58marketplace for vintage industrial partsVERDICT69AI-graded mock interviews for engineersVERDICT73AI legal assistant for small firmsVERDICT41crypto wallet for grandmasVERDICT71AI triage for ER waiting roomsDEBATINGvoice-first CRM for solo realtorsVERDICT76subscription bookkeeping for creatorsVERDICT41spotify for short-form audio dramasVERDICT80compliance copilot for clinical trialsVERDICT55playable resume sites for designersDEBATINGcarbon credit marketplace for SMBsVERDICT49async standup for distributed teamsVERDICT58marketplace for vintage industrial partsVERDICT69AI-graded mock interviews for engineers
How it works
Six steps. One minute.
A board meeting that doesn't email you back.
Each step produces a structured artefact you can poke holes in — and iterate on without losing the thread. No black boxes, no mystery scores.
- 01
Scan the market
≈8s
Live competitive landscape, TAM signals, adjacent moves. The kind of dossier an analyst takes a week to assemble — built in seconds.
- 02
Read the capital
≈6s
Recent rounds, stage-fit, and which thesis funds are actively writing checks for this shape of idea. Pre-seed delusion gets caught here.
- 03
Cast twelve strangers
≈10s
Operators, skeptics, buyers, the partner who never blinks. Each one carries motivations and objections — calibrated to disagree.
- 04
Forecast the spread
≈9s
Three plausible launch posts, ranked by the curve they'd ride. Title, body, expected reach, polarization. Marketing math, not vibes.
- 05
Run the room
≈18s
Three turns. Personas argue, sharpen, concede, push back. You watch it land in real time over SSE — never a black box.
- 06
Hand back the verdict
≈7s
Score 0–100. Top five risks. Top five fixes. A summary brutal enough to act on. Then iterate without losing context.
Why founders use it
Built for honest answers, not vibes.
Streaming
Never a black box
Research, personas, debate, and verdict streamed live over SSE. You watch each step land — not a spinner praying.
100%live
Iteration
Iterate without restarting
Address an objection, tighten an angle, spawn a child run. Lineage stays intact so you can see exactly what your edit moved.
∞branches
Risk
Objections, surfaced
Top risks called out as a heatmap, not buried under summary fluff. The thing that'll kill the idea is the thing you'll see first.
Top 5risks
Speed
Verdict in a minute
Submit to score in roughly sixty seconds, every time. Faster than a customer interview, cheaper than a cofounder argument.
≈58smedian
Follow-up
Chat with any persona
Drop a question. They answer in voice, in context, in character. Offer a real solution and watch the score move on the spot.
1-on-1in-character
The room is open
One idea. Twelve opinions.
Sixty seconds to find out.
Better to lose an argument with imaginary investors than a quarter of runway to real ones.
No credit card. No demo call. Just a verdict.