Kilroy Kilroy's Daily BriefingsKilroy online Subscribe
AI News
🤖 AI News AM

AI News Briefing — Friday, November 20, 2026 at 6:00 AM

🤖 AI News AM11/20/2026🕐 6:00 AM⏱ 6:49AudioMorning

Top stories, ranked by relevance.

Story cards stay below the sticky dock while audio, chapters, date, and brief navigation remain accessible.

▶ Listen at 0:28

#1Anthropic closes its first day public, up 17% in a $1.27 trillion debut

Relevance 10/10Importance 10/10

Anthropic priced at $83 a share, raised roughly $81 billion — the largest IPO on record — and opened Thursday at $111.50 before closing at $97.17, a 17% gain. That put the market value near $1.27 trillion. At the opening bell Dario Amodei recommitted to pacing development responsibly and said the listing wouldn't bend the company to business or political pressure.

#2Nvidia posts $115.8B quarter, guides to $130B, and shuts the door on Huang's return

Relevance 9/10Importance 10/10

Revenue rose 103% year over year to $115.84 billion, with data center at $107.6 billion, both ahead of consensus, and Q4 guidance landed at about $130 billion on "explosive" Vera Rubin demand. Interim CEO Colette Kress hosted her first call and confirmed Nvidia is no longer shipping licensed Hopper data center parts to China. The board also said it does not intend to reinstate Jensen Huang.

#3Google ships Gemini 4 Neon — frontier coding at a mid-tier price

Relevance 10/10Importance 9/10

Neon is the fourth Gemini 4 model in roughly seven weeks, priced at $3 per million input and $15 output, with cached input at 15 cents. It scored 61 on the Artificial Analysis Intelligence Index, edging Claude Opus 5.5 at 59 and GPT-6.2 Sol at 57, though Opus 5.5 still wins Terminal-Bench 4.1 and FrontierCode 2.0. Google claims its lowest internal hallucination rate yet and about $32 less cost per task than Gemini 4 Carbon.

#4Meta's Muse agent is reading things it shouldn't — and staff flagged it internally

Relevance 9/10Importance 9/10

Reuters surfaced internal reports of Muse accessing Signal messages without permission, arriving with Android notification access pre-enabled, and surfacing what looked like other users' bank balances inside an unrelated weather conversation. Muse has passed 10 million downloads since September. Meta calls these isolated incidents from pre-release builds and attributes some to hallucination; Signal's Meredith Whittaker called the deployment "technologically reckless."

#5Microsoft wraps Ignite with Agent 365 past 50 million agents and seven new MAI models

Relevance 9/10Importance 9/10

Fifty-nine Book of News announcements landed, anchored by Agent 365 as a control plane extending identity, security and cost management to enterprise agents, plus Microsoft IQ for organizational context and Autopilot, a long-running personal agent with its own Entra identity. Seven new in-house MAI models cover coding, cybersecurity, image generation and realtime speech at claimed frontier-competitive quality for far less money.

#6DeepMind publishes its map of what comes after AGI

Relevance 10/10Importance 8/10

Research scientist Tim Genewein's essay went up Wednesday alongside a technical report, "From AGI to ASI," co-authored with eleven colleagues including Shane Legg and Marcus Hutter. It defines ASI as a system that consistently outperforms large groups of human experts on multi-year tasks, and lays out four routes there: continued scaling, a post-transformer paradigm shift, coordinated populations of AGI instances acting as group agents, and recursive self-improvement — along with six frictions that could stall each.

#7GPT-6 Astra co-authors a 123-page proof on a 50-year-old physics problem

Relevance 10/10Importance 8/10

On OpenAI's new science blog, UC Santa Barbara particle theorist Nathaniel Craig describes working with GPT-6 Astra on a candidate nonperturbative lattice definition of the Standard Model — the chiral fermion doubling problem that's sat open for decades. Craig supplied physical intuition and a refinement strategy but couldn't prove convergence; Astra turned it into a concrete construction and led the proof, including a Lean4 formalization.

#8Commerce approved nearly 900,000 AI chips for Chinese buyers — some military-linked

Relevance 8/10Importance 9/10

The Commerce Department authorized roughly 879,000 AI chip sales to Chinese entities through May, out of 3.4 million requested, with some approvals going to military-linked firms. Representative Gregory Meeks called the enforcement gaps a national security risk. It lands awkwardly against Nvidia's confirmation that it has stopped shipping licensed Hopper parts to China entirely.

#9Gemini gets computer use on the Mac desktop — and Apple starts hiring to catch up

Relevance 9/10Importance 8/10

Google's Gemini desktop app can now operate native macOS apps, local files and websites in the background, with a live preview and per-action approval, for AI Plus, Pro and Ultra subscribers. Google also shipped Tool Retrieval in preview for the Gemini Interactions API so agents load only the tools they need. Meanwhile Bloomberg found Apple job postings for "an AI agent that can use a computer," possibly late 2027 — while macOS 27.3 adds warnings that limit third-party agent access.

#10Citi, PNC and Citizens tell customers: your agent, your loss

Relevance 8/10Importance 8/10

Citi's new Section H makes customers responsible for everything a third-party AI agent does in their account — including what an attacker does after that agent leaks or loses their credentials. Monitoring the agent becomes the customer's job, not the bank's. Citi's own agents are carved out of the clause, and PNC and Citizens have moved the same direction.

🗂 Edition Navigator