To Build an RL Envs and Human Data Startup - Sean Cai
This is an early version of V0.1 of a living document guide to building a human data startup, a category of business that is increasingly bootstrappable
This is an early version of V0.1 of a living document guide to building a human data startup, a category of business that is increasingly bootstrappable. Feb 04, 2026 ∙ Paid *Early draft for early paid supporters. Please add comments/redactions as you see fit. Escape the underclass by joining the newest gold rush. Its inevitable - don’t listen to the crabs in the bucket that always want to have another brother. Some people have asked me why I choose to write and disseminate this knowledge, which is seemingly alpha generating. Keep reading with a 7-day free trial Subscribe to Sean Cai…
saved by
related reading
- To Build an RL Envs and Human Data Startupsubstack.com
- Pavlov's List: A List of RL Environment Startupspavlovslist.com
- What if RL Environments Aren't Mispriced?benanderson.work
- RL Environments and RL for Science: Data Foundries and Multi-Agent Architecturesnewsletter.semianalysis.com
- The Bitter Lesson - RL Environments Versionseancai.com
- RLHF: Reinforcement Learning from Human Feedbackhuyenchip.com
- Mechanize, Inc.mechanize.work
- The founder's playbook: Building an AI-native startup | Claudeclaude.com
- Convictionconviction.com
- A World of Verifiable Domainsseancai.com
- New Age Commoditiessubstack.com
- Sweatshop data is overmechanize.work