HypeCheck — verify AI product claims

Turn hyped AI product tweets into verifiable claims and test plans

Problem

A V2EX thread dissected a heavily promoted 'decision model' and concluded it was a well-packaged stack of a small model, classification/ranking, constrained output, and calibration — with most money going to marketing. Developers can't easily tell whether such claims hold up, and verifying each one by hand is slow.

Solution

A lightweight claim-extraction tool: paste a launch post or tweet, and it pulls out testable technical claims ('matches a frontier model', 'autonomous decisions'), sorts them into testable vs. vague, and suggests a minimal verification or control experiment for each. The output is a structured checklist, not another marketing summary.

Users

Developers bombarded by AI hype, tech-selection leads, indie hackers, and due-diligence analysts.

MVP

Paste text or a tweet link; extract 5–10 core claims; flag testability and suggest a check; score 'hype density'; export results as markdown.

Revenue

Free tier with limits; subscription for batch analysis and history; team plan for tech media and VCs.

Why now

AI launch volume keeps rising and the marketing language is highly templated, leaving developers without cheap verification. An existing LLM is enough to build the core.