1. Jensen Huang Says We've Reached AGI. Young Workers Are Learning Plumbing. "I think we've achieved AGI," Jensen Huang told Lex Fridman on Monday. The Nvidia CEO delivered the line as settled fact, not provocation.

2. On-Device AI Hits Its Inflection Point A 400-billion-parameter language model ran on an iPhone 17 Pro this week with no cloud connection.

3. The Verge Confronted Superhuman's CEO Over AI Impersonation Shishir Mehrotra booked the interview to discuss his company's roadmap. He got an interrogation.

In Brief

• Senator Warren Calls Pentagon's Anthropic Blacklisting "Retaliation" Warren sent a letter to Defense Secretary Pete Hegseth arguing the DOD could have simply ended its Anthropic contract instead of labeling the company a supply-chain risk. The senator frames the designation as punitive, escalating congressional scrutiny of the Pentagon's AI vendor decisions. TechCrunch

• Sam Altman Exits Helion Board as Fusion Startup Negotiates Power Deal with OpenAI Altman is stepping down as board chair of Helion, the fusion energy company he backed, to clear a conflict of interest. The two companies are negotiating a deal for Helion to sell 12.5% of its power output to OpenAI. TechCrunch

• Gimlet Labs Raises $80M to Run AI Inference Across Competing Chip Architectures The startup built software that lets AI models run simultaneously on hardware from NVIDIA, AMD, Intel, ARM, Cerebras, and d-Matrix. The $80 million Series A targets enterprises locked into single-vendor GPU stacks. TechCrunch

• Vibe-Coding Startup Lovable Starts Shopping for Acquisitions Lovable's founder said the company is actively looking for startups and teams to acquire. The move signals consolidation in the crowded AI-assisted code generation market. TechCrunch

• Littlebird Raises $11M for Always-On Screen-Reading AI Assistant The startup captures context from your computer screen in real time without relying on periodic screenshots. Littlebird uses continuous screen analysis to answer questions and automate tasks based on what you're actively viewing. TechCrunch

• OpenAI Details Safety Architecture for Sora 2 Video Generator OpenAI published its safety framework for Sora 2 and the accompanying social creation platform. The post describes content filtering, provenance tracking, and abuse prevention built into the model and app layers. OpenAI Blog

• Starlette 1.0 Ships After Eight Years of Development The Python ASGI framework that underpins FastAPI hit its 1.0 milestone. Starlette powers a large share of Python async web services but has received far less attention than the frameworks built on top of it. Simon Willison

• AI "Personality of the Year" Contest Formalizes Influencer Economy for Virtual Characters An awards program now recognizes AI-generated personalities competing for audience and brand deals. The contest follows earlier AI beauty pageants and music competitions as synthetic influencers push for mainstream commercial legitimacy. The Verge

• Bay Area Animal Welfare Groups Begin Recruiting AI Researchers Advocates and AI researchers met in San Francisco to explore applying machine learning to animal welfare problems. The effort pairs domain experts in wildlife protection with technical talent from the city's AI community. MIT Technology Review

Read the full edition →