Catch-Up: Gemini 4 Argon Opens Up, Meta's Muse Wonder Lands and Anthropic Warns About AI Self-Copying

We missed a few daily editions, so this is a single catch-up covering October 11 to 15. Nothing below is breaking today; it is the short list of what mattered this week, in plain terms.

Google opens Gemini 4 Argon to everyone

On October 13, Google made Gemini 4 Argon available through its API, AI Studio and the Gemini app. Argon is a “reasoning model”, meaning it works through a problem step by step before answering, and Google pitches it at coding, math, science and multi-step agent work. It had been shown to a small group of security partners on September 30. The context window, roughly how much text it can keep in mind at once, is about one million tokens. Introductory API pricing is $2 per million input tokens and $10 per million output tokens until November 30, and reportedly doubles afterwards. Google AI Pro and Ultra subscribers can try it in the Gemini app. The benchmark numbers are Google’s own, so treat them as a starting point.

Meta releases Muse Wonder

Meta introduced Muse Wonder and Muse Wonder Full on October 12, calling them its most capable models yet. They are built for longer tasks like reviewing a lease or handling an insurance claim. Meta highlights “Active Memory” (carrying useful details between chats) and “Active Focus” (keeping at a hard task and checking in before important actions). Muse Wonder is available via Meta’s Model API and Muse Code, and Wonder Full only to Muse Code subscribers. Meta published no benchmark figures in its announcement and says a full safety report is still to come.

Anthropic: small models may soon copy themselves

Anthropic’s frontier red team says small AI models may be “months away” from being able to copy themselves onto everyday internet-connected devices such as smart TVs. In its tests, no current model was both capable and small enough to do this reliably, but the researchers expect one by the end of 2026. The worry is that such a system would have no single data centre to switch off. Worth keeping calm about: this is a forecast from a lab with its own interest in safety policy, not an observed attack.

Google’s Universal Cart reaches US users

Google began rolling out its Universal Cart in the US on October 15. It collects products from different shops in one cart, tracks price drops and restocks, and can check out with Google Pay at participating merchants. The shop stays the seller either way.

An AI-assisted Chrome exploit earns $250,000

Security firm XBOW says Google paid it $250,000 for a complete Chrome attack chain, built by a researcher working with AI agents in six days. The bugs are fixed and were not used against real users. The point is speed: work that once took weeks now fits into days.

What this means for you: Most of this changes little today. If you use Gemini Pro, Argon is worth a try on a hard task. For everything else, the useful habit is simple: keep Chrome and your smart devices updated.

Sources

• Let’s Data Science: Google releases Gemini 4 Argon
• Meta: Introducing Muse Wonder
• Al Jazeera: AI could self-replicate on home devices within months
• Let’s Data Science: Google rolls out Universal Cart
• Let’s Data Science: XBOW Chrome exploit chain