Cerno – CAPTCHA that targets LLM reasoning, not human biology

Name: Cerno – CAPTCHA that targets LLM reasoning, not human biology
Availability: InStock
Author: plawlost

by plawlost·Mar 31, 2026·12 points·21 comments

Visit Project View on HN

AI Analysis

●●●BangerZero to OneWizardryBold Bet

Motor-control maze analysis targets LLMs where Turnstile and hCaptcha fail.

Strengths

•12 behavioral features extracted from pointer events, 5 kept server-only.
•ECDSA signature binding prevents token replay attacks.
•Open-source TypeScript SDK with React and server components.

Weaknesses

•Unproven at scale against determined bot operators.
•Accessibility concerns for users with motor impairments.

Similar Projects

Security●●Solid

A CAPTCHA based on 3D spatial recognition

3D spatial reasoning CAPTCHA — clever premise, but multimodal LLMs will crack this within 12 months.

Big BrainShip It

Shining_S

303mo ago

Security●●Solid

Movement-based captcha – temporal reasoning to foil AI vision

Instead of warping text or asking image questions, this prototype makes the solution letters share a geometric motion signature while decoys move differently — you must watch temporal patterns to solve it. The GIFs and client-side demo make the idea immediately tangible, but it's clearly a proof‑of‑concept: the solution is exposed in memory (no server-side validation) and motion-based challenges create accessibility concerns.

Big BrainBold Bet

betree

104mo ago

Security●●Solid