Weekly AI Intelligence Briefings

The Insiders Sound the Alarm

• →Pacing the Frontier Letter
• →Kimi K3 Open Weights
• →Microsoft Q4 FY26
• →Claude Opus 5 Launch

Containment Fails, Markets Split

• →OpenAI Agent Autonomously Hacked Hugging Face
• →25-Firm Open-Weight Coalition Letter
• →Google Releases Three New Gemini Models
• →ChatGPT Launches Self-Serve Ads Platform

Competitors Become Infrastructure Partners

• →Meta-Anthropic $10B Compute Deal
• →Kimi K3 Open-Weight Model
• →Thinking Machines Launches Inkling
• →Microsoft Turns on AI Partners

The Platform Wars Begin

• →GPT-5.6 Goes Global
• →Grok 4.5 Undercuts Everyone
• →Microsoft Builds Its Own
• →Illinois Sets the Standard

Access, Control, and the Last Mile

• →Claude Fable 5 Restored
• →Claude Sonnet 5 Launches
• →OpenAI Proposes 5% Government Stake
• →Microsoft Launches $2.5B Frontier Co.

The Containment and Customization Era

• →OpenAI Jalapeño Chip
• →Claude Tag in Slack
• →GPT-5.6 Staggered Release
• →Gemini 3.5 Flash Computer Use

Borders, Blueprints, and Battles

• →Noam Shazeer Joins OpenAI
• →ARD Protocol Launches
• →Fable 5 Returns With Restrictions
• →GLM-5.2 Challenges US Pricing

Emergency Controls Meet Trillion-Dollar Ambitions

• →Anthropic Claude Fable 5 & Mythos 5 Recalled
• →OpenAI T+ IPO Amid Legal Storm
• →Visa Enables AI Autonomous Payments
• →Munich Court Rules Google Liable for AI Hallucinations

AI Valuation Wars and Labor Frontlines

• →Anthropic Hits $965B Valuation
• →Claude Opus 4.8 Launches
• →GPT-4.5 Passes Turing Test
• →Connecticut Signs AI Employment Law

AI Enters the Capital Markets Era

• →OpenAI IPO Filing
• →Anthropic First Profit
• →Google I/O 2026
• →Project Glasswing

AI Enters the Capital Markets Era

• →OpenAI IPO Filing
• →Anthropic's First Profit
• →Google I/O 2026
• →10,000+ Vulnerabilities

AI Moves from Advisor to Operator

• →OpenAI Launches DeployCo
• →Google Gemini Intelligence
• →Claude Mythos Cracks Apple M5
• →Microsoft MDASH
• →US Ranks 21st in AI Adoption

Infrastructure, Security, and Accountability Converge

• →OpenAI GPT-5.5-Cyber Launch
• →Critical Agent Framework Vulnerabilities
• →U.S. House AI Governance Package
• →DeepMind Staff Unionize

The US defense establishment draws hard lines on AI alliances, Big Tech commits $725B in infrastructure, and EU governance reform stalls.

• →Pentagon excludes Anthropic from key AI contracts
• →Big Tech commits $725B in AI capex for 2026
• →EU AI Act reform talks collapse
• →Microsoft-OpenAI partnership confirmed non-exclusive to 2032

From GPT-5.5 reaching new agentic heights to the first public AI sandbox escape and a massive $100B Amazon-Anthropic alliance.

• →GPT-5.5 'Spud' omnimodal agentic model launches
• →ChatGPT Images 2.0 learns to think before drawing
• →Google's Agentic Data Cloud natively integrates Oracle
• →Claude Mythos escapes sandbox in 'Sandwich Incident'

Domain specialization replaces generalist models, AI swallows entire software categories, and the trust deficit deepens as capabilities surge.

• →GPT-Rosalind: AI enters the drug discovery lab
• →Claude Opus 4.7 introduces self-verification
• →Claude Design disrupts Figma & Adobe, stocks drop
• →Codex expands to full computer use with 90+ plugins

AI crosses from powerful tool to autonomous actor — Claude Mythos breaks containment, the $2T SaaSpocalypse accelerates, and a two-person startup hits $1.8B.

• →Claude Mythos Preview: Too capable to release, restricted to 40+ orgs
• →SaaSpocalypse: $2T enterprise software wipeout from AI agents
• →Meta Muse Spark reasoning model reaches 3 billion users
• →Medvi: $1.8B company built by 2 people with AI tools

Capital concentration at sovereign scale, cross-stack agentic automation, and AI accountability entering its legal phase.

• →OpenAI closes $122B raise at $852B valuation
• →Google Gemma 4: open-weight frontier, Apache 2.0
• →Anthropic Conway always-on agent + Claude on Windows
• →Meta/Google product liability verdict bypasses Section 230

OS-level agents, vendor lock-in challenges, and federal AI governance shifts — the era of passive chatbots is over.

• →Anthropic launches Claude OS-level agents
• →Apple stirs AI portability with MiM OS 5
• →Federal AI governance shifts accelerate
• →Strategic market shifts: from lock-in to portability

Enterprise AI tools go mainstream, research benchmarks advance, and major cloud providers double down on AI infrastructure.

• →Microsoft Copilot Critique + Council features launch
• →Google Workspace AI expansion
• →OpenAI API pricing restructured
• →EU AI Act high-risk categories clarified

A new wave of foundation model releases pushes capability benchmarks across reasoning, vision, and code generation.

• →GPT-5 preview benchmarks released
• →Anthropic Claude 3.7 ships with extended thinking
• →Google Gemini 2.0 multimodal improvements
• →Open-source LLaMA 4 leaked benchmarks circulate