AI has never been more powerful. So why does it still miss so much? The reason is often not model capability, but missing context. Our new report with @vana tests the top 5 models to show what’s driving that gap. Get the full findings below.

Our new paper written by @didaoh and @annakaz outlines research priorities for AI data economics.

3/ It outlines a hierarchy of data units: tokens, records, datasets, corpora, and streams. Each operates at a different level of composition and pricing logic, translating technical measures into the economic language needed for real data markets. It also highlights open questions at the frontiers of data economics.

🧵/ As core contributors to the @vana protocol, we are excited to propose a next phase of technical development to achieve Vana's mission to give all individuals a stake in the value of their data. A VRC (Vana Request for Comment) has been published, for the community's feedback, that charts a path forward for rewarding data contributors and data liquidity pools (DLPs), while also incentivizing enduring growth in demand for high quality data on the Vana network. Details and links to comment on the VRC are in the thread below.

8/ In order to access data on the network, developers, companies, and agents will pay fees denominated 80% in data tokens and 20% in VANA, with fees calculated as a function of total compute time in Vana’s secure compute layer.