OS-level user models, not foundation models, will be AI's most valuable asset
sarahcat21 · x · 2026-09-01
Rohan Virani of Amplify Partners makes a provocative call: over the next decade, the most valuable asset in AI won't be foundation models but OS-level user models.
The core argument: once action-taking becomes abundant, the scarce resource becomes proactivity — knowing what to do, when, and whether to do it at all. YouTube, Amazon and banking apps already pre-empt needs via rich user models, while AI assistants still push the entire prompting burden onto the user. Ideally the model should share that burden and solve problems users never knew they had.
The post traces the HCI research underpinning this agenda — from Eric Horvitz and Marc Weiser in the 1990s to recent General User Models that train foundation models to anticipate human action — arguing OS-level user models may become the most contested digital real estate of the next decade.
More from AGI Musings
- AI cheating and deepfakes break remote interviews; employers ask for hand waves — rohanpaul_ai · 2026-09-01
- Critique: Cybernetics fits model intuitions but often devolves into content-free feedback loops — voooooogel · 2026-09-01
- AI and Anthropomorphism: Why We Might Need a New Word — joshwhiton · 2026-09-01
- Is human psychology applicable to AI? Discussing shared minds and alien motivations — repligate · 2026-09-01
- Clarifying the Coasean Singularity: AI Agents and Market Transactions — soumitrashukla9 · 2026-09-01
- Viewpoint: Models Don't Need to Be Smarter, Just Less Stupid — yacineMTB · 2026-09-01