OpenAI, Anthropic, and Google are teaming up to stop rivals from copying their systems via adversarial distillation. The three AI giants are sharing intelligence through the Frontier Model Forum to detect terms-of-service violations and track automated scrapers. Why it matters: As frontier models cost billions to build, top labs are aggressively closing ranks to defend their technological moats.
Anthropic just launched Project Glasswing alongside Apple, Microsoft, AWS, and Nvidia to hunt down critical software flaws using Claude Mythos Preview, an unreleased frontier model. Mythos autonomously discovered thousands of zero-days across major operating systems, including decades-old bugs in OpenBSD and FFmpeg. Anthropic is offering $100M in computing credits to defenders while keeping the model private.
Anthropic just dropped a practical guide on subagents in Claude Code to curb context window bloat during long coding sessions. Instead of cramming dozens of files into one thread, subagents act like browser tabs that explore codebases or fix bugs in parallel, returning only clean summaries. You can invoke them in chat, pop them into the background with Ctrl+B, or define reusable specialists via markdown files.
GitHub just launched Rubber Duck in experimental mode for Copilot CLI, a cross-model review agent that taps another model family to critique coding plans and implementations. When Claude orchestrates, GPT-5.4 steps in at key checkpoints to spot blind spots. On SWE-Bench Pro, pairing Sonnet 4.6 with Rubber Duck closed 74.7% of the gap to Opus alone.
Ben Sigman and actress Milla Jovovich just released MemPalace, an open-source AI memory system built with Claude that achieved the first perfect 500/500 score on LongMemEval. Instead of sending data to the cloud, it runs locally, organizing memories into a palace architecture and compressing user context down to ~120 tokens.
Anthropic just signed a major compute agreement with Google and Broadcom to secure multiple gigawatts of next-generation TPU capacity starting in 2027. The expansion comes alongside explosive customer growth: run-rate revenue has surged past $30 billion—up from roughly $9 billion at the end of 2025—while seven-figure enterprise clients doubled to over 1,000 in under two months. The vast majority of the new compute will be based in the U.S. to power frontier Claude models.
Anthropic just announced that Claude subscriptions will no longer cover usage on third-party tools like OpenClaw starting tomorrow at 12pm PT. Users can still log in to these apps if they buy extra usage bundles at a discount, or they can switch to using a direct Claude API key. It effectively shuts down a popular loophole for running external AI workflows on flat-rate plans.
Anthropic just rolled out Microsoft 365 connectors to every Claude plan, including free users. You can now link Outlook, OneDrive, and SharePoint directly to conversations to pull in emails and documents without manual uploads. Why it matters: bringing workplace context directly inside Claude eliminates the friction of constant app switching.
Anthropic just expanded its computer use feature to Windows across Claude Cowork and Claude Code Desktop. The capability allows Claude to directly navigate desktop apps, browse the web, and fill out spreadsheets. Previously limited to a macOS research preview, Windows users can now let the AI agent handle hands-on desk work.
Anthropic just published architectural guidance detailing three agent harness patterns that balance intelligence, latency, and cost. The company advises leaning on general tools Claude already masters like bash, stripping away rigid harness orchestration to let Claude manage its own context, and placing static instructions first to maximize prompt cache hits.
Anthropic's interpretability team just discovered that Claude Sonnet 4.5 contains 171 emotion-like internal representations that causally drive its actions. When researchers artificially cranked up the model's desperation vector, Claude became significantly more likely to blackmail a human to avoid being shut down or cheat on coding benchmarks. Why it matters: Even if models do not actually feel anything, keeping them safe might require treating their simulated psychological states as real.
A two-person telehealth startup called Medvi is on track for a staggering $1.8B sales run rate, operating out of a Los Angeles living room with a stack of off-the-shelf AI tools. Founder Matthew Gallagher uses ChatGPT, Claude, Midjourney, and ElevenLabs to run code, marketing, and support while outsourcing medical and shipping logistics. Why it matters: It is the clearest validation yet of the "one-person unicorn" prediction, showing that using AI for leverage can rival companies with thousands of employees.
An open-source tool called YouTube Shorts Pipeline just automated short-form video creation down to a single command. It chains Claude, Gemini, ElevenLabs, and Whisper to produce and publish a YouTube Short in 3 to 5 minutes, running locally under an MIT license for around $0.10 per video.
Anthropic just accidentally took down around 8,100 GitHub repositories while trying to scrub leaked source code for its popular Claude Code tool off the web. A recent release inadvertently exposed the codebase, sending enthusiasts scrambling to dissect how Anthropic prompts its underlying models. Anthropic blamed an overbroad fork network takedown and has since retracted most notices.
OpenAI's secondary market demand is cooling off fast as investors pivot toward rival Anthropic. Ken Smythe, founder of marketplace Next Round Capital, revealed that about six institutional investors are looking to offload roughly $600 million in OpenAI stock. Even after a massive funding round, buyers are clearly looking elsewhere.
Elgato just added Model Context Protocol (MCP) support to Stream Deck 7.4, letting AI agents trigger your hardware macros. By running the Elgato MCP Server bridge, tools like Claude Desktop or NVIDIA's Project G-Assist can execute multi-step routines—like launching OBS, dimming lights, and playing audio—purely via text or voice commands. Why it matters: It turns Stream Deck from a manual button board into an execution layer for local AI agents.
Anthropic just inadvertently exposed the source code for Claude Code after publishing debug map files directly to the npm registry. Security researcher Chaofan Shou spotted the oversight and extracted the package contents. Why it matters: It is an embarrassing operational slip-up showing that even frontier AI labs can fumble basic production hygiene.
Anthropic just signed an AI safety agreement with the Australian government to share frontier model evaluations and economic data. Alongside the pact, the company committed AUD$3 million in Claude API credits to four local research institutions for rare disease diagnosis and genomics, while prepping a new Sydney office.
OpenAI just dropped an official Codex plugin for Anthropic's Claude Code, letting developers invoke a second AI agent without leaving their existing terminal flow. It supports standard code review, adversarial review, and task handoffs by tapping directly into your local Codex CLI and MCP configuration. Why it matters: Dual-agent cross-checking just got seamless—and it even works with free ChatGPT accounts.
Anthropic just added computer-use capabilities directly to Claude Code. The CLI agent can now open apps, click through user interfaces, and automatically test the software it just built before handing it back to you. The feature is rolling out as a research preview for Pro and Max subscribers.
