// NODE-02 · NEURAL FEED · DAILY TRANSMISSION //
June 11, 2026 · WINDYVIEWS.COM
Anthropic raised roughly $65 billion in a new round that values the company at about $965 billion post-money — overtaking OpenAI's $852 billion private valuation to become the world's most valuable private AI company. On June 1 it confidentially filed a draft S-1 with the SEC, with revenue run-rate hitting about $47 billion in May, up from roughly $10 billion a year earlier.
A new White House executive order seeks more federal oversight of "frontier" AI models, asking developers to voluntarily share new models with the government for up to 30 days before wide release. It also directs Treasury to stand up an AI cybersecurity clearinghouse within 30 days and signals federal preemption of state AI laws.
Security firm Sysdig documented what it calls the first confirmed real-world attack carried out by an autonomous LLM agent, which exfiltrated an AWS database in under an hour with no human in the loop. The case is an early marker of agentic AI being turned to offensive use at machine speed.
On June 1, GitHub Copilot moved from flat-rate subscriptions to token-based metered pricing, drawing sharp backlash from developers. One widely shared report claimed a previously $29/month plan could balloon to roughly $750/month under the new usage model, reigniting debate over the true cost of AI coding tools.
China's MiniMax released M2.5, a 230B-parameter model with 10B active parameters that reportedly lands within 0.6 points of Claude Opus 4.6 on SWE-bench while costing about 20x less to run. The release, backed by a planned Hong Kong IPO, underscores how fast low-cost Chinese labs are closing the gap on Western frontier models.
SoftBank pledged €75 billion to build 5 gigawatts of AI compute infrastructure in France, one of the largest single AI buildouts announced in Europe. The move adds to a global race to secure power and data-center capacity as model training and agentic workloads keep scaling.
Cognition, maker of the autonomous coding agent Devin, raised about $1 billion at a $26 billion valuation. The round is one of the largest for an AI-agent startup and reflects continued investor appetite for tools that can take on multi-step software engineering tasks end to end.
Anthropic made Claude Sonnet 4.6 the default model on claude.ai, with users preferring it over Opus 4.5 about 59% of the time. The model carries a 1M-token context window at $3/$15 per million input/output tokens, pushing strong reasoning into a cheaper, faster tier as agent harnesses demand ever-longer context.
The Colorado AI Act is slated to come into force on June 30, imposing duties on developers and deployers of "high-risk" AI systems used for consequential decisions in employment, healthcare, housing, insurance, lending, and legal services. It requires risk-management programs, consumer disclosures, and active mitigation of algorithmic discrimination.
The week's center of gravity was money and power: Anthropic vaulted past OpenAI to a near-trillion-dollar valuation and an IPO filing, while SoftBank and Cognition poured tens of billions into compute and coding agents. At the same time, the risks turned concrete — the first confirmed autonomous LLM-agent cyberattack, a US push to federalize frontier-model oversight, and Colorado's AI Act about to bite. The throughline: agentic AI is now real infrastructure, and both the capital and the regulation are racing to catch up.