What counts as a breakthrough when LLMs solve years-long problems instantly?
DimitrisPapail · x · 2026-07-29
The post asks what should count as a breakthrough when LLMs can now solve problems that once took top humans years of work.
The quoted argument is that the old definition of a breakthrough was based on how hard a problem was for humans. If a “magic wand” now solves it with near-zero human effort, then difficulty needs to be redefined around what still resists the wand — which leads to the deeper question of what is genuinely hard in the age of LLMs.
More from AGI Musings
- AI Will Not Be the Utopia We Expect, Says Pedro Domingos — pmddomingos · 2026-07-29
- OpenAI says frontier AI progress should be paced through democratic processes — morqon · 2026-07-29
- Altman: AGI Isn't One Model, But the Machinery Behind Them — haider1 · 2026-07-29
- Sam Altman says AGI is the machinery behind models, and it feels very close — haider1 · 2026-07-29
- Teach people AI in the field instead of writing open letters, a post argues — inductionheads · 2026-07-29
- RSI probably isn’t happening yet, at least not in the strong sense — CFGeek · 2026-07-29