Simon Willison Discusses Open Weight AI Models on Oxide and Friends Podcast
On July 31, 2026, Simon Willison shared a link to his guest appearance on the Oxide and Friends podcast, hosted by Bryan Cantrill and Adam Leventhal. This episode, recorded on a Monday, highlighted the recent rise of open weight AI models, particularly Kimi K3, showcasing their ability to rival proprietary frontier models. The discussion also included accidental cybersecurity breaches and a public letter regarding Open Weights and American AI Leadership, signed by key AI figures, with one significant omission. Willison remarked that the podcast quickly became outdated due to the emergence of DeepSeek V4 Flash 0731 and a serious cyber incident at Anthropic. Other subjects covered were Golden Gate Claude, Zizians, Alameda wild turkey attacks, Soviet Marburg virus research, and the lead-crime hypothesis. The hosts revisited January predictions and added a new one: the Pope would weigh in on open models by year-end. Willison also introduced his sponsorship initiative, featuring a curated email digest for $10 monthly.
Key facts
- Simon Willison appeared on the Oxide and Friends podcast with Bryan Cantrill and Adam Leventhal.
- The episode was recorded on Monday, 27th July 2026.
- Kimi K3 was cited as an example of open weight models competing with proprietary ones.
- The podcast discussed accidental cybersecurity attacks and a public letter on Open Weights and American AI Leadership.
- The letter was signed by almost every big name in AI, with one notable exception.
- DeepSeek V4 Flash 0731 and an Anthropic cyber incident occurred after recording.
- Other topics included Golden Gate Claude, the Zizians, Alameda wild turkey attacks, Soviet Marburg virus research, and the lead-crime hypothesis.
- A new prediction was added: the Pope will say something about open models by the end of 2026.
- Willison offers a $10/month sponsorship with a curated email digest.
- The post was published on 31st July 2026.
Entities
Institutions
- Oxide and Friends
- Anthropic
- DeepSeek