Featured · October 8, 2026
Research
How much gameplay data does a world model need?
All data is equal, but some data is more equal than others
Read the featured postThoughts on Artificial World Intelligence™, rights-cleared data, and building the infrastructure for what comes next.
All data is equal, but some data is more equal than others
Read the featured postThree datasets and one model are now available on Hugging Face. See what we tested, what we found, and what we would like the community to do with them.
Data quality is not a data-team problem. It is a capital allocation problem. The executive case for why scraped data is the most costly line item in modern AI.
Eniac, FPV, Seven Stars, SV Angel, and angels from robotics, AI, gaming, and Google are joining. The capital builds the channel between game studios and the AI labs training world models, video generators, and embodied agents.
The scraping problem. The provenance crisis. Licensed data is not just a legal precaution. It is better data. And the market is already shifting.
AI learned language. Now it needs to learn worlds. What AWI means, why it matters, and why the data requirements behind it are fundamentally different from anything the AI industry has built for so far.
Better data, not just bigger models, will define the next era of AI. The story behind Origin Lab and the thesis driving what we build.
New posts on world data, AI training infrastructure, and the licensed data economy. No spam, just signal.