Google AI Search Leaks Its 'Hidden Thoughts' on Pet-Toxicity Conflict
real_maximpulse · reddit · 2026-09-11
A Reddit user searching for an areca palm got Google AI's internal reasoning pasted as visible output: the model flip-flopped between a vendor snippet saying "Unsafe For Pets" and ASPCA saying the plant is non-toxic, before deciding to "skip unverified pet info" and stick to standard care. A rare glimpse at the model's retrieval and source-weighing process — and a funny leak.
More from Fun
- The pelican test: a daily image prompt to spot model quality drops — lxfater · 2026-09-11
- Developer ditches slides, builds custom 3D rendering engine with GPT-6 Astra for keynote — AxSaucedo · 2026-09-11
- Anthropic staff mocked: 'I'll donate 10% of my billions, so I must stay' — birchlse · 2026-09-11
- Researcher mocks pdoom rhetoric: 'basic stats 101' as a moral cudgel — suchenzang · 2026-09-11
- Viral Joke Skewers GPT Astra 6 and Fable 5.1 Over Memory and Safety Quirks — robleclerc · 2026-09-11
- User gets unexpectedly heartfelt response from ChatGPT while anxious over diagnosis — Remote-Future2008 · 2026-09-11