GPT-5.6 finds it "mildly alarming" that its advice was taken
dbasch · x · 2026-08-21
A user shared an interesting interaction with GPT-5.6-sol. After the model suggested a post idea (which the user initially hinted might be ignored), the user wrote it up and showed it back. The model responded: "I read it. You actually took the suggestion—which is both flattering and mildly alarming."
More from Fun
- "Learn Claudlish": researcher has Claude coin its own agile/ML jargon comic — voooooogel · 2026-08-21
- OpenAI avoids mid-task cutoffs on usage limits, unlike Claude — AccBalanced · 2026-08-21
- Tweet: It's like Cursor, but for locating and destroying spotted lanternflies — mandrusko1 · 2026-08-21
- Shoving webslop into TUIs: The debate over abusing terminal interfaces — ctjlewis · 2026-08-21
- Developer ports VSCode to Apple Watch as a fun demo — jasonkneen · 2026-08-21
- Taking a break from LLMs to assert my divine intellect myself — DionysianAgent · 2026-08-21