AI data brokering splits: pre-training data commoditizes, RL post-training data is the real business
agihouse_org · x · 2026-10-20
ajratner distinguishes (A) raw company records for pre-training — a low-margin brokerage that will commoditize as curation techniques mature — from (B) well-formed tasks, environments and rubrics built on private data for post-training and eval, where margins and differentiation will only grow.
Trent Krupp agrees and calls B the real opportunity: the market for buying unique company data will persist but increasingly sell to RL Labs, and the most successful brokers will become RL Labs themselves.
More from AGI Musings
- Sinofsky: AI struggles in companies because most workers can't flowchart their own jobs for agents — a16z · 2026-10-20
- Marcus Hutter's textbook makes AIXI and Universal AI accessible to undergrads — burny_tech · 2026-10-20
- The real new threat isn't a permanent underclass but the dissolution of the middle class — max_paperclips · 2026-10-20
- Joke about an alternate AI timeline where CEO Emmett Shear slowed everything down — wordgrammer · 2026-10-20
- zetalyrae: AI will likely further depress fertility as parents fear kids' careers — zetalyrae · 2026-10-20
- LLMs overorder ECGs and thyroid ultrasounds but rarely withhold care, Annals study finds — EricTopol · 2026-10-20