Tokens are just readouts: why LLMs aren't fundamentally discrete systems
Sauers_ · x · 2026-11-11
Pushing back on the claim that LLMs are fundamentally discontinuous because they emit discrete tokens rather than a smooth continuous flow like humans, the author argues tokens are merely intermediate readouts from an underlying continuous process — tokens carry only 0.001% of the available past information, with the real computation living in the continuous internal state.
More from Research
- Why mini-batch k-means converges fast but clusters worse: the MacQueen update — thomasahle · 2026-11-11
- UCLA's TriMI: a $800 tri-modal robot hand for demos, haptic teleop, and policy execution — DennisHongRobot · 2026-11-11
- Glen Berseth ranked all 687 CoRL 2026 papers into a prioritized 101-paper reading list — GlenBerseth · 2026-11-11
- Jason Wei maps which benchmarks frontier labs actually reported in 2026 — _jasonwei · 2026-11-11
- Apple's HALO hits 91% success in real homes with zero real-world training — emmanuelvivier · 2026-11-11
- Why AI can't cure cancer yet: Precigenetics pitches live-cell dynamic data for virtual cell models — iScienceLuvr · 2026-11-11