Top Stories
Anthropic Accidentally Reveals Claude Mythos Model
A content management mishap at Anthropic exposed the existence of “Claude Mythos,” an unreleased model the company describes as “a step change” in capability. Around 3,000 assets from Anthropic’s internal blog became publicly accessible, including draft announcements revealing that Mythos outperforms Opus 4.6 in cybersecurity, software programming, and academic reasoning.
Internal documents warn the model could “significantly heighten cybersecurity risks” by rapidly finding and exploiting software vulnerabilities. The leak also revealed a planned “Capybara” tier that would exceed Opus capabilities.
Why it matters: The accidental disclosure comes as Bloomberg reports that Anthropic is considering going public as early as October 2026, racing OpenAI to IPO.
Source: SiliconANGLE
Meta Pours $10 Billion Into West Texas AI Data Center
Meta increased its commitment to an El Paso data center by more than sixfold, bringing total investment to $10 billion. The facility aims to reach 1 gigawatt of capacity by 2028 and will create 300 permanent jobs with over 4,000 construction workers at peak.
This adds to Meta’s $135 billion planned AI capital expenditure for 2026—nearly double last year’s spending. The company continues cutting staff to fund the AI buildout, with roughly 1,500 Reality Labs jobs already eliminated and reports of further reductions affecting up to 20% of the workforce.
Source: CNBC
Meta Unveils AI Glasses for Prescription Wearers
Meta plans to debut two new Ray-Ban smart glasses models next week designed specifically for prescription lenses. While previous models supported prescription lenses as an add-on, these mark the first purpose-built frames for that market.
Source: Bloomberg
Google DeepMind Expands Robotics Push
Google DeepMind partnered with Munich-based Agile Robots to integrate its Gemini Robotics foundation models with Agile’s hardware. The companies will test and deploy robots using Gemini models across electronics manufacturing, automotive, data centers, and logistics. Agile Robots already has more than 20,000 deployed robotic systems globally.
Source: TechCrunch
Quick Hits
-
OpenAI Shopping Expands: ChatGPT gets richer product discovery with side-by-side comparisons powered by the Agentic Commerce Protocol, plus a Walmart in-app experience rolling out to all users. OpenAI
-
Codex Plugin System: OpenAI ships a broad Codex release with first-class plugins, clearer multi-agent workflows, and new TUI features including prompt recall and remote websocket connections. OpenAI
-
Perplexity Personal Computer: Available to Max subscribers at $200/month, the always-on Mac mini system monitors triggers and executes proactive tasks 24/7, requiring explicit approval for sensitive actions. 9to5Mac
-
Claude Service Disruptions: Users reported elevated errors on Friday particularly affecting Opus 4.6 and Sonnet 4.6 models. IBTimes
-
AI Layoffs Continue: Tech companies have cut 59,121 workers in 2026, with over 9,200 layoffs directly attributed to AI adoption. Amazon leads with 16,000 cuts, followed by Block at 4,000 (40% of workforce). IBTimes UK
Worth Watching
Anthropic IPO Timeline: With the Mythos leak and Bloomberg’s October IPO report, watch for official announcements on both the new model launch and public offering plans. The accidental reveal may accelerate Anthropic’s disclosure timeline.
Meta’s Prescription AR Play: The dedicated prescription glasses suggest Meta sees corrective lens wearers as an underserved market for AI wearables. Next week’s announcement should clarify pricing and availability.
Gemini Robotics Partnerships: DeepMind’s Agile Robots deal follows a pattern of expanding Gemini beyond software into physical systems. More robotics partnerships likely incoming as Google pushes embodied AI.