AMD to Invest Up to $5 Billion in Anthropic Under AI Infrastructure Deal

AMD has agreed to invest up to $5 billion in Anthropic under an infrastructure agreement covering tens of billions of dollars’ worth of AI systems. Anthropic will deploy up to two gigawatts of capacity using AMD’s Instinct MI450-series accelerators (MI455X GPUs), with deployment of the first gigawatt beginning in the first half of 2027. The deal combines an equity investment from AMD with a large hardware order from Anthropic, and follows similar AMD transactions with OpenAI (up to 6 GW) and Meta (up to 6 GW) that include warrants tied to deployment milestones. Read more

Google Releases Gemini 3.6 Flash and 3.5 Flash Lite

Google has released Gemini 3.6 Flash and Gemini 3.5 Flash-Lite as new workhorses designed to cut latency and token costs for enterprise AI agents. Gemini 3.6 Flash produces 17% fewer output tokens than its predecessor and achieves up to 65% fewer tokens on the Datacurve DeepSWE benchmark. Pricing sits at $1.50/M input tokens and $7.50/M output tokens. The model scores 49% on DeepSWE (up from 37%), 63.9% on MLE Bench (up from 49.7%), and 1421 on Google’s GDPval-AA v2. On OpenRouter, pricing is $1.50/M tokens prompt / $7.50/M tokens completion. The computer-use tool is now built directly into the Gemini API with an OSWorld-Verified score of 83.0%.

Gemini 3.5 Flash-Lite targets high-volume document processing at $0.30/M input tokens and $1.50/M output tokens ($0.30/M / $1.50/M on OpenRouter), reaching 350 output tokens/sec — the fastest in the 3.5 series. A restricted Gemini 3.5 Flash Cyber variant is also available for vulnerability remediation. Read more

SenseTime Launches Galaxy Project for Domestic AI Chip Scale-Up

Chinese AI company SenseTime has launched the Galaxy Project, teaming with nearly 20 partners to scale domestic AI chip infrastructure in China. The project aims to build a large-scale domestic AI computing ecosystem, reducing reliance on imported chips amid ongoing US export restrictions. Read more

Bristol Myers Squibb Acquires Nvidia AI System for Drug Discovery

Bristol Myers Squibb has purchased Nvidia’s AI system for drug discovery, marking a significant enterprise deployment of AI in pharmaceutical research. The system leverages Nvidia’s accelerated computing platform to speed up molecular modeling and drug candidate screening. Read more

Nvidia Betting on Physical AI for Healthcare Robotics

Nvidia has introduced a new Medical Physics Simulation framework that treats healthcare robots as physical AI systems requiring embodied experience to learn, not just code. The framework aims to solve the data scarcity problem in healthcare robotics by generating realistic synthetic training environments. Read more

New Models on OpenRouter

Several notable models have appeared on OpenRouter:

  • Poolside: Laguna S 2.1 — A new model from Poolside with 1M context, priced at $0.10/M tokens prompt, with a free tier offering 256K context.
  • Meituan: LongCat 2.0 — Meituan’s latest model with ~1M context at $0.30/M tokens.
  • Limited free access ending soon: Poolside Laguna M.1 (free until Jul 28, 4 days left), OpenAI GPT-5.3 Chat and GPT-5.2 Chat (free until Aug 10, 17 days left).
  • nethical6/conversation-steganography ⭐1k — Use LLMs to hide messages inside normal-looking conversations using steganography techniques. Written in Go.
  • Blaizzy/nativ ⭐789 — Local AI native to your Mac. Chat, serve, monitor, and connect MLX models from one macOS app. Written in Swift.
  • marcelroed/gigatoken ⭐1.7k — Language model tokenization at GB/s speeds, ~1000x faster. Written in Rust with Python bindings.
  • Jakubantalik/thinking-orbs ⭐740 — Dotted thought-orb loading indicators for AI & agent UIs — six tuned states, two sizes, auto dark/light mode.
  • Vincentwei1021/video-shotcraft ⭐673 — AI video skill for Claude Code & Codex, producing cinematic product videos with Remotion.
  • Terence Tao’s ChatGPT conversation about the Jacobian Conjecture counterexample reached #1 on Hacker News with over 900 points, demonstrating LLM-assisted mathematical discovery.

Terence Tao Uses ChatGPT for Mathematical Discovery

Mathematician Terence Tao shared a detailed ChatGPT conversation exploring a counterexample to the Jacobian Conjecture, which reached #1 on Hacker News with over 930 points and 530+ comments. The conversation demonstrates how frontier LLMs can assist in high-level mathematical reasoning and exploration.