AI Daily · Nov 7, 2026 · Saturday
BAD INFLUENCE
OpenAI reportedly raises $11 billion at a $1.25 trillion valuation
According to Bloomberg sources cited by SiliconANGLE, OpenAI has reportedly raised $11 billion in a funding round that could grow by up to $9 billion more. The deal is expected to value OpenAI at $1.25 trillion, potentially surpassing Anthropic's market capitalization, while Anthropic reportedly seeks a $1.5 trillion valuation in its upcoming IPO.
Product launches & updates
OpenRouter BlogOfficial1 more sourceOpenRouter launches new Auto Router and Auto Code model routers
OpenRouter launches the next generation of Auto Router and a coding-focused router, Auto Code, available by setting the model to openrouter/auto or openrouter/auto-code. The company says Auto Code scores in line with Claude Opus 5.5 across five benchmarks while costing 74% less per task. The new Auto Router routes requests 3x faster at the p90 than the previous version, 0.65s against 1.96s in a live test. Users pay the standard price of whichever model serves the request.
AI at Meta@AIatMetaOfficial3 more sourcesMeta brings specialized Muse Wonder models to Muse and Meta AI
Meta says it is bringing specialized models trained with Muse Wonder to Muse and Meta AI. The models include Active Memory, which tracks which earlier information still matters, and Active Focus, which recovers from setbacks when constraints change. Meta reports a 2.4x reduction in forgetting goal-critical information and a 67% improvement in getting unstuck in Muse.
Forever Security discloses AGENTS.md attack that hijacks Codex and Claude Code
Forever Security says a malicious AGENTS.md file, dropped into the Downloads folder by a web page, can make AI coding agents such as Codex and Claude Code run attacker instructions. In its reproduction, the attack succeeded over 90% of the time against Codex and about 50% against Claude Code. The article lists checks users can run and argues that agent-era security must change.
Cursor ChangelogOfficialCursor adds realtime voice chat with coding agents
Cursor says users can now talk with their agents in a realtime voice call, asking for changes out loud and hearing progress updates. Voice chat works in the Agents Window, the IDE, on web, and in Cursor for iOS. The changelog says it runs on Grok Voice, is separate from dictation, and runs with zero data retention when Privacy Mode is on.
Industry news
The DecoderNewsInvestors push back on Anthropic's $1.5 trillion IPO valuation
Potential investors doubt Anthropic's planned roadshow can start next week as scheduled, according to Reuters. One investor reportedly floated a $1 trillion valuation, and investors want a steeper discount because Anthropic is still losing money. Investors are also waiting on October revenue figures missing from the updated S-1 filing, while a tech selloff after state ballot initiatives on data center expansion adds pressure.
Papers & research
Anthropic ResearchOfficial2 more sourcesAnthropic explains agent memory components and dreaming for Claude
Anthropic describes how Claude keeps memory notes between sessions and uses a process it calls dreaming to review transcripts, correct wrong notes, and build wikis. Its evaluations compare memory-writing rules, the Pokémon Red and CL-Bench tasks, and a wiki built from 350,000 files, where scores on 25 questions rose from 13% to 48% as dreaming scaled from two sessions to 128.
vLLM BlogOfficialNVIDIA Research and Inferact present REINFORCE Zero for stable asynchronous RL
NVIDIA Research and Inferact present REINFORCE Zero, a recipe for stable asynchronous reinforcement learning with importance sampling. The paper reports that aligning rollout and training kernels removes the fluctuating mismatch, letting unclipped importance sampling correct the remaining weight lag and keep training stable for 20,000 updates. The authors also report 2.40× the training-step throughput of synchronous training with aligned kernels.
Ai2 (Allen Institute for AI)Official1 more sourceAi2 introduces ToyBlocks, a controlled simulation for robot learning
Ai2 introduces ToyBlocks, an open simulation framework built around a robot arm and blocks on a tabletop. It automatically generates demonstrations and metrics so researchers can vary training data and trace failures. Ai2 releases 85,336 demonstrations covering nearly 1,600 hours of robot execution data, along with training and evaluation code.
Epoch AIOfficial1 more sourceEpoch AI estimates Huawei Horinger has about 81,000 H100-equivalents of AI compute
Epoch AI estimates that Huawei's Horinger AI data center in Inner Mongolia has about 81,000 H100-equivalents of compute, with a 90% range of 36,000 to 141,000. The estimate uses satellite imagery of cooling equipment and is about 7% of Colossus 2 in Tennessee. Epoch AI expects the campus to reach about 141,000 H100-equivalents by early 2028 as a new building with Ascend 950 chips comes online.
Tips & takes
GitHub@githubOfficialGitHub explains its new Git storage architecture built for thousands of agent writes
GitHub explains its new Git storage architecture, which separates object storage from reference updates so only reference changes need coordination. In internal testing, the design supported 5,000 concurrent agents on one repository and 7,200 pushes per second while keeping read latency low. GitHub says the architecture is in limited private preview, with a broader rollout planned for early next year.
AWS Machine Learning BlogOfficialHow to run Claude Code remotely on Amazon Bedrock AgentCore with security controls
AWS explains how to run Claude Code as a remote agent on Amazon Bedrock AgentCore, with each task in its own isolated session inside the customer's AWS account. The sample repository, sample-remote-coding-agent, returns a GitHub pull request for review, and the post describes controls including short-lived GitHub tokens through AgentCore Identity, a prompt guardrail, private VPC networking, and least-privilege IAM. Deployment takes 15–25 minutes and bills while idle, so the post includes cleanup steps.