← Kilroy’s Daily Briefings
🧠 AI News PM

AI News Afternoon Briefing — Sunday, May 17, 2026 at 3:00 PM

🧠 AI News PM5/17/2026🕐 3:00 PM⏱ 7:10AudioPM edition

Top stories, ranked by relevance.

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

▶ Listen at 0:24

#1Anthropic's Project Glasswing Deploys Claude Mythos to Find Thousands of Zero-Day Vulnerabilities

Anthropic launched Project Glasswing, a cybersecurity initiative using its restricted Claude Mythos Preview model, which autonomously discovered thousands of zero-day vulnerabilities across every major operating system and web browser — including a 17-year-old remote code execution flaw in FreeBSD. The model is available only to launch partners including AWS, Apple, Google, Microsoft, and JPMorgan Chase at $25/$125 per million input/output tokens, with no plans for general availability.

🔗 Anthropic
No image

#2OpenAI Unifies ChatGPT, Codex, and API Under Greg Brockman Ahead of IPO

OpenAI announced Friday that co-founder Greg Brockman will permanently lead all product strategy, merging ChatGPT, Codex, and the developer API into a single organization aimed at building "one unified agentic experience." The restructuring — timed three days before Google I/O — positions OpenAI for a late-2026 IPO, with the company now exceeding $25 billion in annualized revenue and 900 million weekly active ChatGPT users.

🔗 TechCrunch
No image

#3Google I/O 2026 Opens Tuesday — New Gemini Expected but Trails Mythos and GPT-5.5

Google's developer conference kicks off May 19 with a new Gemini model as the expected centerpiece, likely a 3.2 or 3.5 release. Sources describe a meaningful improvement in reasoning and multimodal capability but not a step-change — landing roughly at GPT-5.5 levels and meaningfully behind Anthropic's Claude Mythos, particularly in coding benchmarks. Gemini Intelligence, a cross-app agentic AI system for Android, is also expected.

🔗 TechTimes
No image

#4Anthropic Introduces "Dreaming" — Agents That Self-Improve Between Sessions

Anthropic unveiled "dreaming," a technique that lets AI agents review past sessions, extract patterns, and consolidate memory — analogous to hippocampal replay during sleep. Legal AI firm Harvey reported a 6x improvement in task completion rates, while medical doc company Wisedocs cut review time by 50%. The feature is now in research preview for Claude Managed Agents.

🔗 VentureBeat
No image

#5Google DeepMind Assembles "Strike Team" Under Sergey Brin to Close Anthropic's Coding Gap

Sergey Brin has personally assembled a strike team within DeepMind, led by Gemini pre-training engineer Sebastian Borgeaud, with a mandate to close the gap with Anthropic's coding capabilities. In an internal memo, Brin wrote that DeepMind "must urgently bridge the gap in agentic execution." The move comes as Claude Code hit $2.5 billion in annualized run-rate revenue by February 2026.

🔗 The Information via Sherwood News
No image

#6Meta's Muse Spark Rolling Out Across WhatsApp, Instagram, and AI Glasses

Meta's first model from its Superintelligence Labs — the natively multimodal Muse Spark with visual chain-of-thought and multi-agent orchestration — is now deploying across WhatsApp, Instagram, Facebook, Messenger, and Ray-Ban Meta glasses. The model, built under new AI lead Alexandr Wang, features a "contemplating" mode and competitive benchmarks at a fraction of the compute cost of rivals.

🔗 TechCrunch
No image

#7AI Layoffs Not Boosting Stocks: 56% of S&P 500 Firms Trading in the Red

A CNBC analysis published today found that of 23 S&P 500 companies that announced AI-driven layoffs, 13 (56%) have seen their stocks decline — averaging a 25% drop. Salesforce is down 32% since its AI layoffs, and Fiverr plunged 54% after cutting 30% of staff. More than 112,000 U.S. jobs have been attributed to AI displacement since early 2025.

🔗 CNBC
No image

#8SpaceXAI Hemorrhaging Talent: 50+ Researchers Exit Since Merger

TechCrunch reports that more than 50 researchers and engineers have left SpaceXAI since the February merger, including key leaders in coding, world models, and Grok voice. At least 11 defected to Meta and seven joined Mira Murati's Thinking Machines Lab. The pre-training team has dwindled to a handful of people, raising questions about whether the company can still develop frontier models ahead of its expected IPO.

🔗 TechCrunch
No image

#9Cohere Acquires Aleph Alpha in $20B Sovereign AI Deal

Canadian AI company Cohere is acquiring Germany's Aleph Alpha in a $20 billion deal backed by retail giant Schwarz Group, creating a transatlantic "sovereign AI" alternative to U.S. tech giants for regulated sectors like finance, defense, and healthcare. The merger, blessed by both Ottawa and Berlin, will deploy on Schwarz's STACKIT cloud infrastructure and is positioned as a direct response to growing concerns about data sovereignty.

🔗 TechCrunch

#10Novo Nordisk Partners with OpenAI Across Drug Discovery and Operations

Danish pharma giant Novo Nordisk signed a strategic partnership with OpenAI to integrate AI across its entire pipeline — from drug discovery and clinical trials to manufacturing and supply chain — with full deployment planned by end of 2026. The deal is structured with strict data governance and human oversight, as Novo races to reclaim market share from Eli Lilly in the weight-loss drug market.

🔗 CNBC
No image
Kilroy’s Daily Briefings · 5/17/2026