10 Comments
User's avatar
Sarabjeet Matharu's avatar

This is the new space race. I wonder if Mythos is the equivalent of the mythical Star Wars ICBM defence system

Abek's avatar

A tech dystopia of AI-gangsterism is a much less likely outcome than simply assuming that as Altman and Amodei have done for years now, that they're just hyping to sell. This is not going to be some world changing model. They're just trying to IPO by peddling fear.

whisperer's avatar

This. They made a marginally better model again together with a bunch of new marketing tricks and suddenly it’s “post mythos” world?

Laughable.

Chinese labs will catch up in time.

Abek's avatar

What are your thought on BABA, Z.AI as an investment? If energy is the bottom line of AI, the Chinese labs will be far better placed to capture much of the global demand. In the same way their phones dominate globally but the Iphones etc do in the west.

whisperer's avatar

Sure. If it comes down to energy Chinese labs will absolutely destroy western offerings, at least for the actual practical usage.

Of course their weakness is less access to chips, but I somehow doubt that’s permanent either.

Abek's avatar

What else can it come down to?

ToeKneeEi's avatar

Mythos - a message directed to the Trump administration: “Hey guys, Anthropic is super important to national security. How about you let us back in the door!”

Yuzu Xu's avatar

Good translation of a nuanced reaction. The 76.6M Xiaohongshu #Claude views are worth putting in context: most of that content is Claude Code tutorials and API workarounds shared by developers accessing the model via VPN, not engagement with Anthropic's safety thesis. Chinese developers follow Claude closely because they use it daily for coding. The 'anti-China AI' label coexists with heavy Claude Code adoption the same way Chinese programmers use American IDEs without tracking the political alignment of the toolmaker. The Project Glasswing consortium list is worth watching from a China angle. Apple, Microsoft, and Google all have meaningful China operations. Whether that creates unintended access vectors is a question Chinese security commentators will ask.

Adam De Gree's avatar

Mythos looks neat but its findings can be reproduced with much (600x) cheaper open-source models...https://aisle.com/blog/system-over-model-zero-day-discovery-at-the-jagged-frontier

Robert C Culwell's avatar

This one is for you Irene, great work! 📝🤖⏰🔐

https://claude.ai/public/artifacts/5849dc77-d80b-4857-aa19-6ee272075d3f

💱🪽 The GlassWing in the Ghost Valley,

Semper Fortis, Semper Supra! 📡🗽💽📲🔍😏

Grace and peace to you sister. 🌏 ⚖️ 🔔