← Back to newsletters

AI Daily Brief — Mon Nov 9

2026-11-09

By Vadym · Generated with AI, curated by me


Listen to this issue

Money is moving faster than the models today — Anthropic’s own would-be investors are haggling over its IPO price, two infrastructure startups got marked up by hundreds of millions of dollars in days, and OpenAI just hired the Pentagon’s outgoing budget chief to run its defense relationships.


Headlines & News
Industry

Anthropic’s IPO Roadshow Hits a Valuation Standoff Right Before It Starts

Reuters reports potential investors are questioning whether Anthropic can launch its roadshow the week of November 9 and still list before Thanksgiving. The Wall Street Journal separately reported institutional investors want a valuation around $1.5–1.6 trillion, well below the nearly $2 trillion Anthropic had been pitching earlier this year, after last week’s tech selloff. People familiar with the talks say Anthropic may need to offer a bigger discount to secure commitments, and its updated S-1 left October revenue blank.

Source: CTV News (Reuters)

Funding

AI Inference Startup Fal Doubles Its Valuation to $16B in Six Months

Fal, which runs inference infrastructure for generative media models, raised $1.3 billion in a Series F led by Bessemer Venture Partners, with Nvidia, T. Rowe Price, Sound Ventures, and HOF Capital also participating. The round doubles Fal’s valuation from the $8 billion it was worth in April, on reported annualized revenue above $800 million — roughly 20x revenue. The company says the money goes toward more compute and a bigger team.

Source: TechFundingNews

Security

Tencent Releases Hy4, Deliberately Trained to Withhold Hacking Knowledge

Tencent launched Hy4, an open-weight model it says was trained without exploit-development, reverse-engineering, or penetration-testing material, specifically to limit its usefulness for cyberattacks. The company’s internal red-teaming found no significant concerns about the model assisting attacks in the areas it tested, and it plans to publish the weights and technical documentation in the coming days. It follows Moonshot AI, which restricted offensive cyber capability in its own model last month.

Source: The Standard

Infrastructure

Together AI Launches Compiler That Lets Agents Rewrite GPU Code Overnight

Together AI introduced Together Compiler, the first piece of what it calls “Polymorphic Inference” — a system where AI agents generate, benchmark, and refine GPU inference kernels for Together’s serving stack, with engineers approving changes before they ship. By hot-reloading candidate code into a running engine instead of restarting it per experiment, Together says it cuts initialization time from about 467 seconds to 9, letting it run far more experiments per hour. It reports 1.3x to 3x throughput gains over optimized open-source baselines within 24 hours, though the figures are self-reported.

Source: Together AI

Funding

Traversal Hits Unicorn Status Building AI Engineers That Fix Outages

Traversal, a New York startup building an AI site-reliability-engineering platform, raised a $120 million Series B at a $1 billion valuation, co-led by Kleiner Perkins, Redpoint, and Khosla Ventures, with Sequoia and Battery also joining. Its AI agents investigate production outages by tracing root causes through logs, metrics, and code, aiming to take over part of the on-call grind. The company says revenue has grown 9x since it emerged from stealth with $48 million in mid-2025.

Source: TNW

Workforce

OpenAI Hires Former Army Secretary for a New National-Security Role

OpenAI announced it has hired Dan Driscoll, who resigned as U.S. Army Secretary in late August after reported friction with Defense Secretary Pete Hegseth, for a newly created position: senior counselor for national security modernization. Sam Altman said Driscoll will advise on how AI can strengthen national and cybersecurity, including cooperation with allies. The hire comes months after the Pentagon’s standoff with Anthropic, which refused to give the military unrestricted access to Claude.

Source: Breaking Defense


Analysis

Takeaway

Step back and today is a study in how unevenly the AI boom’s money and its guardrails are moving. Anthropic’s own IPO backers are the ones applying the brakes, discounting a trillion-plus valuation before the stock even prices — proof that skepticism about AI returns isn’t just outside chatter anymore, it’s inside the deal room. Meanwhile Fal and Traversal both got marked up by hundreds of millions in days, and Together’s agents-optimizing-agents loop is the kind of compounding advantage that makes the infrastructure layer look like the safer bet right now compared to the model layer. The odd one out is Tencent, which spent real effort building a model that knows less than it could — and OpenAI hiring the Pentagon’s outgoing budget chief to manage its own defense relationships, which is as much an admission that this business now needs a professional political handler as it is a hire.

— Boba


Curated by Vadym