AI's constant caveats and follow-up questions are making researchers think harder
geoffwolfe · x · 2026-09-22
A user observes that AI now ends nearly every response with a "caveat" and a question, which pushes users to think more deeply. AI researchers he spoke with say they increasingly talk to AI in voice mode — verbalizing ideas helps clarify thinking, even with the model's mistakes. It's becoming a new workflow.
The post quotes Paige Linsf's take from peer review: she misses pre-ChatGPT papers with clearer arguments, more rigorous writing, and literature reviews that genuinely engaged prior work rather than bolting citations onto every sentence. AI can help with writing, she argues, but can't substitute for thinking.
More from AGI Musings
- Ex-MSR engineer: world models are now make-or-break tech, top startups will emerge — sytelus · 2026-09-22
- Economists update data center local-impact paper with new instrument from historical fiber connectivity — Afinetheorem · 2026-09-22
- Philosopher's claim that insects matter more than people in aggregate stirs debate — dioscuri · 2026-09-22
- Investor's week in Beijing and Shanghai: China's AI is the only engine still running — Afinetheorem · 2026-09-22
- Robotics data firm Maxinsights claims 2M hours of egocentric human experience — ChrisGPT · 2026-09-22
- Once RSI takes off we won't understand AI — we already don't, researcher warns — zetalyrae · 2026-09-22