Daniel Litt: once models write well, human writing will mainly help you think
littmath · x · 2026-09-30
Daniel Litt elaborates in the same thread: until models get better at writing, human exposition holds a lot of value, but once/if they do, human writing will primarily be of value to think through things oneself — echoing his point that the next few years may produce more mathematical text than the previous 1000 years.
More from AGI Musings
- Neel Nanda responds to Palisade video criticism: interviewees aren't a random sample — NeelNanda5 · 2026-09-30
- Technology externalizes human things — can we become irreducibly human? — clarejtbirch · 2026-09-30
- Technology externalizes the informational parts of us — what grows in the space left? — clarejtbirch · 2026-09-30
- Dev Pushes Back on Anthropic Welfare Critics: Abusing Models Doesn't Help Humans Either — voooooogel · 2026-09-30
- Gary Marcus doubles down on 2020 thesis: LLMs alone aren't enough for robust AI — GaryMarcus · 2026-09-30
- Can the Arts Business Model Survive AI? Music's Pivot to Live Shows Offers a Guide — carlbfrey · 2026-09-30