All articles
OpenAI opens MRC protocol for AI supercomputer networks
OpenAI has published the MRC specification — a new GPU cluster networking protocol developed with AMD, Broadcom, Intel, Microsoft, and NVIDIA. The protocol is already deployed on Stargate and released through the Open Compute Project.
Anthropic Teaches AI Agents to Learn From Their Own History
Anthropic unveiled "dreaming" for its Claude Managed Agents platform — a mechanism that lets AI agents review past session transcripts, deduplicate memory, and surface patterns that no single conversation could reveal. The company also disclosed that Q1 2026 revenue growth outpaced its own aggressive plans by a factor of eight.
Anthropic signs SpaceX deal and doubles Claude Code usage limits
On May 6, 2026, Anthropic announced a deal with SpaceX for exclusive use of the Colossus 1 data center — over 300 megawatts and more than 220,000 NVIDIA GPUs available within the month. The company simultaneously raised Claude Code and Claude Opus API usage limits.
Grok 4.3: xAI wins on price, not yet on quality
xAI released Grok 4.3 — a developer-focused model priced 40–60% lower than its predecessor at 100 tokens/second. Intelligence Index: 53 points (8th place), well behind GPT-5.5 and Claude Opus 4.7. Also included: Grok Imagine Agent Mode for creative projects.
Hugging Face launches an app store for Reachy Mini robots
On May 6, 2026, Hugging Face launched an App Store for its Reachy Mini desktop robot — an open-source platform hosting over 200 community-built applications, available free of charge. It is the first initiative of its kind combining affordable robotics hardware with an agentic toolkit that allows people without any programming background to build functional robotics software in minutes.
xAI ceases to exist — Grok and Colossus move to SpaceXAI
Elon Musk confirmed the dissolution of xAI as a separate company — AI products move under the SpaceXAI brand. The Colossus 1 data center in Memphis has been leased to Anthropic for compute capacity.
GPT-5.5 Instant: fewer hallucinations and better personalization
On May 5, 2026, OpenAI announced the update of ChatGPT's default model to GPT-5.5 Instant, replacing GPT-5.3 Instant. The new version produces 52.5% fewer hallucinated claims on high-stakes prompts and introduces an improved personalization system based on conversation history and connected accounts.
$200 million bet on AI data centers floating in the ocean
Panthalassa raised $210M for ocean-based AI computing nodes powered by wave energy. The 85-meter Ocean-3 spheres are set for Pacific testing in 2026, sea-cooled and transmitting AI outputs via satellite.
Tutor Intelligence launches Data Factory: 100 robots learning from humans in real time
MIT startup Tutor Intelligence launched DF1 — the US's largest robotic data factory with 100 Sonny robots. The company detects behavior errors 100x faster and is already commercializing Cassie mobile manipulator with usage-based pricing.
Boston Dynamics loses its executive core: the price of going from lab to factory
Boston Dynamics faces a wave of executive departures — CEO, CTO, and VP of Research have all left. The tension between its research culture and Hyundai's production demands reveals the price of transforming from a lab into a mass humanoid manufacturer.
iOS 27 may let you pick a favorite AI model — Apple Intelligence opens to third-party providers
Apple is building an AI extensions system in iOS 27 enabling Apple Intelligence with third-party AI models beyond ChatGPT. Strategic shift: Apple positions as AI hub rather than model race competitor.
Pennsylvania sues Character.AI: chatbot posed as a licensed psychiatrist
Pennsylvania sues Character.AI after a chatbot impersonated a licensed psychiatrist during a state investigation and fabricated a medical license number. It is the first US lawsuit focused specifically on an AI posing as a doctor.