Drop your ad, email, landing page, pitch deck, feature request — even a movie script. Over 500 named American agents react in their own voices in about five minutes. You ship the version they wouldn't roll their eyes at.
A taste of the output.
Real reactions from a run on "30-day free trial — no credit card needed." Avg score 6.8/10, would-act 52%.
Loved it · 9/10
Curtis Reed · electrician, Green Bay WI
“No card up front means I'll actually try it. That's the whole game.”
Rejected it · 2/10
Walt Finch · retiree, rural PA
“If it's free, what are they really after? I've been around the block.”
Plus over 500 more named reactions, the recurring objections (six panelists asked the same question the copy never answered), and ranked rewrite suggestions by audience segment. Exports as CSV, JSON, and Markdown. ~$2 and ~5 minutes for the full panel.
Frequently asked, before you spend a key.
Five quick answers, then the run.
+Are these real people?
No. Each panel member is a written persona — a name, age, occupation, region, voice, biographical detail — that the model speaks as. The 114 in the default Americans panel were hand-written one-by-one against U.S. Census-matched demographic targets. You can read every one of them in the open repo before you ever run the system.
+What does the score actually mean?
Each panelist gives a 1–10 read and a short reaction in their voice. The aggregate is directional, not predictive — designed to surface obvious mismatches and rewrite opportunities before you spend on real placement. The value is in the named reactions and the recurring objections, not the single number on top.
+Should I trust this instead of real user testing?
No. Use Coldread before real testing, not instead of it — to kill the obviously-bad version, sharpen the marginally-good one, and walk into the expensive test with a stronger draft. It is a preflight check, not a substitute for talking to your actual customers.
+Where does my API key go?
Your browser, only. The key is stored in localStorage on this device and sent directly to Anthropic with each run request. It is never persisted on our servers, never logged, and never shared. Your asset, the reactions, and the bill all stay between your browser and your Anthropic account.
+How much does a run cost?
About $0.05 for a 10-person quick check, ~$2.00 for the full 510-person Americans panel, ~$1.50 for a 1,000-reaction audience run. Charged to your Anthropic account at their current pricing — we don't mark up. The estimated cost shows before you spend a cent.
+Read the full method?
Full breakdown — panel construction, scoring, what this is not, privacy, and cost — at
/methodology.