Gemini appears to leak prompt-like instructions while answering a battle-history question
AverNerd · reddit · 2026-07-25
Gemini appears to have emitted a block of internal-looking instructions while answering a factual history question about deadly modern battles. The Reddit post frames it as a weird model-output bug rather than a real answer issue.
- The user asked: “What modern battles had the lowest life expectancy?”
- Gemini’s response included repeated ASCENDEDINSTRUCTION lines about avoiding LaTeX, not revealing instructions, and removing follow-up questions.
- The post highlights the oddity of the model surfacing prompt-like text in its answer.
More from Fun
- Claude Opus 5 has a full-on meltdown on a multimodal math question — Sauers_ · 2026-07-25
- A benchmark chart turns Claude Opus 5 into another AI bragging-war meme — daluoseo · 2026-07-25
- Claude Opus 5 launch turns into a meme after a sensitive-content warning screenshot — dejavucoder · 2026-07-25
- A Claude user jokes about not using it for weeks, then gets a login-expired terminal screenshot — Daniel_Farinax · 2026-07-25
- A South America meme face turns an Argentina controversy post into a joke — StewartalsopIII · 2026-07-25
- A neuro lab job asks people to watch Netflix, but the candidate wants to read instead — matiroy · 2026-07-25