O'Reilly: Big AI Labs Are Building Wrong Future
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Tim O'Reilly argues the big AI labs are 'reading the future wrong' by pouring resources into the biggest, best frontier models rather than what ordinary people want, saying breakthroughs are 'pushing us further away' from what users need.
- O'Reilly says open-source AI must go beyond open-weight models to encompass the full architecture — a clean separation between model, harness, and application — so users can 'paint outside the lines' rather than be tracked and locked in.
- O'Reilly's nonprofit, the AI Disclosures Project, is building an 'open-memory consortium' explicitly aimed at countering Mark Zuckerberg's thesis that Meta will own users by offering the AI that 'knows you best.'
- O'Reilly claims all the cybersecurity incidents on record have come from frontier models, and frames that risk profile as an argument to slow frontier development — not restrict open-weight releases.
- O'Reilly compares today's AI capital concentration to the Uber/Lyft template of VC-subsidized market capture, calling it a 'failed model' that chokes out alternative paths the way the web emerged outside VC funding in the 1990s.
- O'Reilly says his own company's book business has fallen from a roughly $70 million peak to about $30 million, and is now exploring AI tools to help experts monetize knowledge as publishing revenue declines.
- O'Reilly describes AI as a 'thought partner' he uses to brainstorm and convert hour-long interviews into usable drafts, while insisting human writing and editing remain the final step.
Why it matters: O'Reilly's open-memory consortium gives institutional weight to the open-source counter-movement against frontier-lab consolidation, and his historical analogy — frontier models ending up like mainframes while diffused innovation happens elsewhere on smaller open models — gives open-source advocates a concrete frame to redirect both regulatory attention and developer mindshare away from the biggest labs.
Ask SkimNews
