RL data startups get harder: insider says don't start one without lab-level alpha, as profit-sharing data shops skip VC
Scobleizer · x · 2026-10-19
A widely shared thread on the RL data/environment industry, amplified by Scobleizer as "the best take on RL data on X."
Key points:
- On "should I start a data company": probably not, unless you have real alpha — deep industry knowledge, firsthand exposure to frontier RL environments via a lab or another data company, or strong connections with lab researchers.
- Data work is much harder than 6 months ago and quality standards keep rising; but if it's your passion, "now is the best time and it'll get increasingly worse."
- Emerging trend: many smaller data shops are skipping venture backing, instead offering employees very generous profit-sharing arrangements.
More from AGI Musings
- "AI is the new OS" inverts UX logic: more typing, less cognitive load — firasd · 2026-10-19
- FT chart ignites debate: is 20% GDP growth from AI actually plausible? — ChrisGPT · 2026-10-19
- OpenAI staffer says distillation matters more than he thought, drawing 'double game' criticism — stanfordnlp · 2026-10-19
- Mathematician claims LLMs only interpolate, can't discover new math like Riemann — burny_tech · 2026-10-19
- Scott Alexander: AI Safety Is Not A Corporate Plot — SammaVaco · 2026-10-19
- AI circle debates post-singularity: technofeudal inequality vs abolishing status — zetalyrae · 2026-10-19