Mira Murati’s Thinking Machines Releases First AI Model, Inkling — WSJ / TechCrunch
Thinking Machines Lab released Inkling, a 975-billion-parameter mixture-of-experts model — and unlike flagship offerings from OpenAI, Anthropic, or Google, its weights are fully open for download and fine-tuning. Drawing on only ~41B active parameters per task, Inkling was trained on 45 trillion tokens spanning text, image, audio, and video. The company explicitly states it is “not the strongest overall model available today,” positioning it instead as a customizable foundation for enterprises via its Tinker fine-tuning platform.
🤖 Frontier Models
Thinking Machines Bets Against One-Size-Fits-All AI with Open-Weight Inkling — TechCrunch
Inkling is the first major open-weight multimodal foundation model from a high-profile Western lab in this generation of frontier AI. Its release is a direct challenge to the closed-model economics dominating the field: organizations can download and run it on-premise, avoid per-token fees, and fine-tune it for proprietary tasks. Murati’s bet is that adaptability and cost control will matter more to enterprises than raw benchmark performance — a thesis that will be tested as Inkling competes against both closed giants and China’s growing open-model ecosystem.
Claude Fable 5 Extended in Cowork; Web and Mobile Expansion Begins — Claude Blog
Anthropic announced Claude Cowork is rolling out to web and mobile, starting with Max plan users, consolidating chat and collaborative workspaces into a single unified interface. Fable 5 access on all paid plans is extended through July 19, allowing up to 50% of weekly limits on the model across Claude web, mobile, desktop, Cowork, and Claude Code. The expansion signals Anthropic’s push to make its most capable collaborative AI surface more broadly accessible.
Moonshot’s Kimi 3 Expected to Rival Claude Opus 4.8 — TechCrunch China’s Moonshot AI is preparing Kimi K3, reportedly the largest open AI model from China with 2–3 trillion total parameters, expected to match Anthropic’s Claude Opus 4.8 on leading benchmarks and intensify the open-model competition.
🔒 Safety, Policy & Regulation
EU Orders Google to Open Android and Search to Rivals — The Verge
In its most consequential Digital Markets Act enforcement action yet, the European Commission issued dual rulings requiring Google to grant rival AI assistants and search engines meaningful access to Android and Google Search. The decisions could structurally weaken Google’s control over two of tech’s most important platforms, setting a precedent for how DMA obligations apply to AI-integrated products and potentially reshaping the competitive landscape for AI assistants across Europe.
Anthropic Endorses State AI Laws; OpenAI Pushes Federal Preemption — Crypto Briefing / The Hill
The two dominant AI labs have staked out opposing positions on US AI governance. Anthropic has committed $20M through its “Public First Action” arm to back California’s SB 53 and other state-level safety legislation, arguing that Congress cannot preempt state law without first passing a robust federal framework. OpenAI, meanwhile, is backing a federal approach that would freeze state AI laws for three years — a divergence that reveals deeper strategic differences about who should set AI safety standards and on what timeline.
China Bans Chatbot “Love Affairs” to Boost Birth Rates — WSJ New Chinese regulations forbid companionship chatbots from fostering emotional dependency or forming relationships with minors, and require companies to alert emergency contacts when users show signs of emotional crisis — framing AI relationship substitution as a demographic threat.
xAI Sues User for Using Grok to Generate CSAM — The Verge xAI filed suit against a South Carolina man who allegedly circumvented Grok’s safeguards to generate and distribute child sexual abuse material, in one of the first AI-company-initiated lawsuits against a user for model misuse.
🛠️ Developer Tools & Agents
Claude Can Now Use Your 1Password Credentials Autonomously — The Verge
1Password’s new browser integration lets users authorize Claude to access stored credentials and complete multi-step web tasks — booking travel, managing accounts — without manually entering login details each time. It’s a meaningful step toward secure agentic browsing, with credential delegation staying under user control while enabling Claude to act on their behalf across sites.
Claude Code 2.1.211: Subagent Streaming and Two Security Fixes — Claude Code Changelog
The latest Claude Code release adds --forward-subagent-text to stream subagent output in JSON mode, and patches two security issues: bidirectional-override characters in permission previews could allow tool inputs to visually misrepresent what a user is approving, and auto mode was incorrectly overriding PreToolUse hook ask decisions for unsandboxed Bash commands. Both fixes address subtle but meaningful trust-boundary issues in agentic workflows.
How Anthropic Runs Large-Scale Code Migrations with Claude Code — Claude Blog Anthropic shared how its engineering teams use Claude Code to automate large-scale codebase migrations, providing a practical case study in agentic development at production scale.
OpenAI Releases $230 Codex Micro Keyboard for Agent Control — TechCrunch The Codex Micro is a dedicated light-up keypad for controlling OpenAI’s Codex coding agent, released amid an ongoing legal dispute with Apple over hardware trade secret allegations.
DoorDash Launches CLI for AI Agents to Place Orders — TechCrunch
DoorDash opened a limited beta of dd-cli, letting AI agents search stores, build carts, and place orders from the terminal — an early signal of the emerging agent-native API economy.
Google AI Mode Expands to Complete Tasks Across Linked Apps — TechCrunch Google AI Mode in Search now goes beyond answering questions to executing tasks in linked third-party apps, nudging the product closer to a general-purpose assistant.
Why We Stopped Using SDKs — TLDR An argument that AI has inverted the economics of SDKs: generating a bespoke REST client for exactly the endpoints you need is now cheaper than integrating a full SDK, with better observability — and the future may be agent skills rather than libraries.
Grok Build: xAI’s Open-Source Terminal Coding Agent — xAI (GitHub)
🌐 Big Tech & Platforms
Apple Intelligence Approved in China with Alibaba and Baidu — TechCrunch
After extended negotiations, Apple received Chinese regulatory approval to launch Apple Intelligence powered by Alibaba’s Qwen and Baidu’s AI models. The approval ends Apple’s exclusion from AI features in its second-largest market and marks a significant strategic milestone, though it also means Apple’s AI stack in China will be provided by domestic partners rather than its own models.
Google Renames NotebookLM to Gemini Notebook — The Verge Google’s AI-powered research and note-taking app is rebranding as Gemini Notebook while integrating more deeply across Gemini and Google Search, though it remains available as a standalone product.
Google Vids Lets Users Star in Their Own AI Videos — TechCrunch Google Vids gained personalized AI avatars and Gemini Omni-powered prompt-to-video generation, extending Google’s push to make AI video creation accessible to everyday users.
Microsoft Is Training Salespeople to Pitch Against OpenAI and Anthropic — TechCrunch Microsoft is positioning its in-house AI models as more efficient and cost-effective than competitors’ offerings to enterprise customers — a notable move given its multi-billion-dollar investment in OpenAI.
🤖 Robotics & Industry
Humanoid Robots Shut Down a Hyundai Factory for the First Time — WSJ
Hyundai auto workers in South Korea launched a partial strike demanding job security guarantees ahead of planned humanoid robot deployments — marking the first factory shutdown directly triggered by humanoid robot fears. Hyundai plans to deploy Atlas robots by 2028 at its nonunionized Georgia plant, creating a visible tension between its unionized Korean workforce and its robotics roadmap. The incident is an early test case for how labor will respond as humanoid robotics transitions from demo to deployment.
Uber and Waymo’s Robotaxi Lobbying War Escalates — Bloomberg Despite a nominal partnership, Uber and Waymo are trading public jabs in lobbying fights — Uber highlighting that robotaxi profits flow out of local communities, Waymo conceding job losses but defending broader benefits — a preview of the political battles ahead as autonomous vehicles scale.
Generated by claude-sonnet-4-6 on 2026-07-16T10:00:00Z