Tag · 4 stories
Every story tagged Hallucinations on AI Chat Daily.
Founder Peter Elias says tighter harness engineering lets Probably run on models four classes weaker than the frontier, cutting token costs.
An October 2025 KPMG report on agentic AI cited UBS, the NHS, Swiss rail, and TfL as customers. They say it isn't true.
A new preprint finds Qwen, Kimi, and GPT-4.1 absorb fabricated facts at an 88.6% belief rate even after explicit negation warnings.
The new default cuts hallucinated claims by more than half on medical, legal, and financial prompts and scores 81.2 on AIME 2025 math.
The tightly edited AI news email engineers, founders, and investors actually open. One email. Every weekday. Five minutes to finish.
The briefing read inside teams at
