korrents

← Post-training is what turns a model into a product, but it can only do…

On the map

8 connected korrents · 8 moments on record from 19 Jun 2024 to 26 Aug 2026.

Same subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subject Post-training is what turnsa model into a product, butit can only do that on top… AS Aravind Srinivas — holds since 2024-06-19 A benchmark result should be reported under a stated budget, or as a curve against test-time compute — never as a single number. A benchmark resultshould be reported… NB Noam Brown — holds since 2026-06-26 A lab should spend most of its compute on research rather than on building the next model, because research is where the tenfold yearly efficiency gains come from. A lab should spend mostof its compute on… DP Dylan Patel — holds since 2026-03-13 After pre-training, post-training and test-time scaling, the fourth scaling law is agentic: multiplying AI by spawning agents, and the whole loop scales on one thing, compute. After pre-training,post-training and… JH Jensen Huang — holds since 2026-03-23 Bet on a system that is maximally learned and minimally constrained, and add structure only where it improves the scaling laws. Bet on a system that ismaximally learned and… DD Dmitri Dolgov — holds since 2026-08-03 Current AI techniques are four to six orders of magnitude away from optimal in data efficiency and test-time compute efficiency. Current AI techniquesare four to six orders… FC François Chollet — holds since 2026-08-07 Large language models could still plateau, and that possibility should be held open even though no evidence of it has appeared. Large language modelscould still plateau… DH David Heinemeier Hansson — holds since 2026-08-26 Model comparisons understate real progress, because benchmark tables do not control for how much test-time compute each answer used. Model comparisonsunderstate real… NB Noam Brown — holds since 2026-06-26 Models resemble each other because pre-training is the same everywhere; what differentiates labs now is RL and post-training. Models resemble eachother because… IS Ilya Sutskever — holds since 2025-11-25
same subject