AssimilatedHuman/LLM-Inquisitor

Evaluating AI behaviour under real‑world work conditions to surface issues before they become problems. LLM INQUISITOR identifies failures (drift, instability etc) by observing AI during normal tasks — a tool the industry desperately needs to stem the 85% failure rate. Includes Quick Start, Practitioner’s Guide and Methodology.

2 best points · 0 comments · 2 Hacker News posts · 2 GitHub stars · View on GitHub

Hacker News discussions

• New method for testing AI using real work-flows (2 points, 0 comments, by ballista2026, 2026-05-27)
• LLM INQUISITOR: Evaluating how AI models handle long, realistic tasks (1 points, 0 comments, by ballista2026, 2026-05-20)