Oxide and Friends: The Open Weight Revolution with Simon Willison
Simon Willison's Weblog Simon Willison ● Covered by 13 sources
Simon Willison joined the Oxide and Friends podcast to unpack a chaotic week in AI: an open-weight model matching frontier ones, plus a scary security scare. He even predicts the Pope will weigh in on open models by year's end.
Based on reporting by Simon Willison's Weblog, Simon Willison — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Some weeks in AI news feel like they compress a full year into five days. This was one of them, and Simon Willison spent Monday hashing it out with Bryan Cantrill and Adam Leventhal on the Oxide and Friends podcast. By the time the episode dropped, it was already stale — that's how fast things are moving right now.
The headline event was Kimi K3, an open-weight model that Willison says proved it can go toe-to-toe with the best proprietary systems out there. That's a big deal for anyone who's been arguing that closed labs have a permanent moat on quality. Layered on top of that was an open letter on Open Weights and American AI Leadership, signed by nearly every major name in the field — with one conspicuous holdout Willison doesn't name outright but clearly finds notable.
Then there's the security mess. Willison references an accidental cybersecurity incident tied to this open-weight momentum, and notes that Anthropic had its own embarrassing security stumble around the same time. Neither made it into the actual recording, since both broke after the mics went off, which says something about how quickly the ground is shifting under everyone's feet — DeepSeek's V4 Flash release on July 31st would have made the cut too if the timing had lined up.
The conversation wasn't all AI doom-and-boom, either. Willison and his hosts wandered into Golden Gate Claude, the Zizians, wild turkey attacks in Alameda, Soviet-era Marburg virus research, and the lead-crime hypothesis — the kind of sprawling, digression-heavy chat that podcasts do better than any news article. They also circled back to predictions made back in January to check their batting average.
Willison closed with a new one for the record books: by the end of this year, he bets the Pope will say something about open models. It sounds like a joke until you remember how mainstream this technology has become — popes have weighed in on AI and labor before, so weighing in on who controls the weights isn't that much of a stretch.
My take — AI-written commentary, not fact-checked reporting
I'll take that Pope bet seriously, honestly — open weights have gone from a niche developer preference to a geopolitical talking point in about eighteen months, and that's the real story buried under all the podcast digressions. Kimi K3 matters less as a single model and more as proof that the closed-lab moat is shallower than the marketing suggests. The security incidents, though, are the part nobody wants to sit with: we're racing to open-source frontier capability faster than we're figuring out how to secure it, and that gap is going to bite someone bigger than Anthropic before this year is out.
Read more about this at: Simon Willison's Weblog
Related stories
[AINews] Much ado about Open Weights
Latent Space · 1 month ago ·
32
Open models recap: more on Kimi K3, Qwen 3.8, Xi's WAIC speech, distillation, the open-closed gap, and what's next
Interconnects · 1 month ago ·
11
Import AI 465: Open vs closed gaps; Kimi K3; Demis' big policy plan
Import AI · 1 month ago ·
21