All stories
Morning briefs, midday updates, evening wraps, recaps, Top Stories, stories and wire briefs in one feed, newest first.
Saturday, November 21
StoryAI selloff pushes Wall Street into correction
Chip stocks fall hardest as doubts about AI spending grow

StoryLean certifies Claude-built Riemann proof
The Riemann proof passes its formal check; understanding it comes next
- Wire
Microsoft unveils MAI Code 2 in a family of seven new models
Microsoft unveils MAI Code 2 among seven new models

- Wire
Nvidia reports $115.8 billion quarter and raises its fiscal 2028 outlook
Nvidia posts $115.8B quarter, raises outlook
- Wire
Oracle and OpenJDK fix Java4Shell flaw CVE-2026-114938
Oracle and OpenJDK patch the Java4Shell flaw
WEEKEND REPORTACN Weekend Report
Anthropic's share sale closes, and its price support ends. Nvidia's board faces a White House call to bring back Jensen Huang. Wall Street slides into a correction on AI doubts. And the Lean community certifies a Riemann hypothesis proof built with Claude.
Friday, November 20
- Wire
Claude Code 2.1.330 adds sign-in for Enterprise-managed organizations
Claude Code 2.1.330 adds Enterprise org sign-in
StoryUPDATECognition in talks at a $75B valuation
Devin maker weighs raising $2B or more · Bloomberg
Earlier: Anthropic shares slide on second day


- Wire
OpenComputer offers an open-source, self-hostable alternative to Devin
OpenComputer pitches an open-source Devin alternative

DAILY WRAPACN Daily Wrap
Cognition talks new funding. Cursor weighs dropping Chinese models. An open model fits a laptop. And AI prompts leak planted secrets.

- Wire
Andreessen Horowitz leads Series B in AI agent messaging company Linq
a16z leads Linq's Series B for AI agent messaging


WireNasdaq erases its 2026 gains as Anthropic shares fall on first full trading day
Nasdaq erases 2026 gains as Anthropic shares slide
MIDDAY UPDATEACN Midday Update
AI models still lag at judgment calls. AI agents find a zero-day on their own. And Claude now writes ninety percent of its team's code.


WireCyberattack disrupts European Commission systems for more than 10 days
Cyberattack disrupts the European Commission
WireGoogle's Chrome team builds a synthetic web to test browser agents
Chrome builds a synthetic web to test AI agents
WireMicrosoft's Agent Framework PyPI package compromised to steal credentials
Microsoft Agent Framework PyPI package compromised


WireOpenAI says its agents' success on long research tasks rose to 46 percent
OpenAI: agent success on 2-4 hour tasks doubles to 46%
WEEKLY RECAPACN Weekly Recap
Nvidia rules out Jensen Huang's return. An Anthropic mathematician claims a proof of the Riemann hypothesis. Epic rebuilds Unreal Engine for AI coding. Claude solves two black hole problems. Microsoft builds its own coding models. And OpenAI launches GPT-6.2.
WireMETR adds judgment-heavy tasks to its AI time horizon plot
METR: models lag on judgment tasks like eval design
StoryAnthropic mathematician claims Riemann proof
Levent Alpöge · proof posted in Lean · not yet verified
StoryUPDATEAnthropic shares slide on second day
Down 12% premarket, Friday, Nov 20 · offering closed
Earlier: ACN Top Story · Anthropic goes public in largest IPO ever
WireApple may be developing its own AI agent for the Mac, Bloomberg reports
Apple may be building an AI agent that runs the Mac
StoryEpic to drop Blueprints in Unreal Engine 6
Code moves to Verse · more assets stored as text for AI
- Wire
Signal restricts screenshots to stop AI assistants reading private chats
Signal blocks screenshots to keep AI agents out of chats
MORNING BRIEFACN Morning Brief
An Anthropic mathematician posts a proof of the Riemann hypothesis, made with Claude. Update: Anthropic's shares slide. And Epic reshapes Unreal Engine for AI coding.
WireAnthropic mathematician claims Claude-aided proof of the Riemann hypothesis
Anthropic mathematician claims Riemann proof with Claude
WireEpic to drop Blueprints and move Unreal Engine 6 code to Verse
Unreal Engine 6 drops Blueprints for Verse
Thursday, November 19

StoryBezos' Prometheus builds AI engineering system
Fast Company: $18 billion raised, about 250 staff
WireClaude Code 2.1.329 warns where a folder's hooks are declared
Claude Code 2.1.329 adds a hooks trust warning
StoryLangSmith tracks coding agent spend by outcome
Agent spend tied to the pull requests it produced

WireTechCrunch asks whether AI is behind a run of U.S. tech outages
TechCrunch asks if AI is behind a run of tech outages
DAILY WRAPACN Daily Wrap
Apple is building its own AI agent. LangSmith tracks coding agent spend by outcome. Jeff Bezos' Prometheus builds an AI engineering system. And Nvidia nears a deal for a diffusion AI startup.
Top StoryAnthropic goes public in largest IPO ever
Shares close up 17% after topping $2 trillion at the peak

WireGoogle brings computer use to the Gemini app for Mac
Gemini for Mac can now work across apps and sites

WirePHP security updates fix 18 flaws, including critical remote code bugs
PHP patches 18 flaws, including critical RCE bugs
StoryReport: Meta staff say Muse agent exposed data
Reuters cites internal posts; Meta has not responded
MIDDAY UPDATEACN Midday Update
Update: Anthropic starts trading on the Nasdaq. Meta staff say its AI agent exposed private data. And Google builds a copy of the web to test its agents.
WireAnthropic says Claude agents solved two open problems in black hole physics
Claude agents solve two open black hole physics problems

WireClaude Code 2.1.328 brings auto mode to Bedrock, Google Cloud and Foundry
Claude Code 2.1.328 adds auto mode on Bedrock and Foundry
WireCritical NGINX2Root flaws in NGINX let attackers hijack servers
F5 patches critical NGINX2Root flaws in NGINX

StoryUPDATEMicrosoft makes repos private after PyPI hijack
Poisoned agent-framework versions stole credentials

WireSatya Nadella explains why Microsoft offered buyouts to 60,000 employees
Nadella ties Microsoft's 60,000 buyout offers to AI
WireSnap lets creators build AR Lenses with AI coding assistants like Claude Code
Snap's Easy Lens CLI lets Claude Code build AR Lenses
MORNING BRIEFACN Morning Brief
Microsoft responds to its toolkit hijack. OpenAI's agents finish nearly half of longer tasks. Arizona opens a probe as banks push back on AI agents. And an AI assisted attack hits the European Commission.
WireAnthropic's first Nasdaq trade comes in at $111.50, above its $83 IPO price
Anthropic debuts on Nasdaq as ANT at $111.50
WireGitHub CLI lets developers and agents wait for pull request events
GitHub CLI adds gh event listen for pull request events
StoryUPDATEAnthropic prices its IPO at $83 a share
Above the $70–$78 range · Nasdaq: ANT · trading Thu, Nov 19
StoryCopilot code review scales to each pull request
Adaptive effort · secret detection · public preview
WireGitHub Copilot adds eight connectors, including Supabase, ClickUp and Buildkite
Copilot adds connectors for Supabase, ClickUp, Buildkite

WireGoogle makes Tasks in Gemini free in six countries
Google makes Tasks in Gemini free in six countries
WirePalo Alto Networks agrees to buy Applied Compute in a $4.5 billion stock deal
Palo Alto Networks to buy Applied Compute for $4.5B
StoryVS Code moves its agent to the Copilot harness
New default for agent sessions · Local harness removed Feb 2027
DAILY WRAPACN Daily Wrap
Update: Anthropic prices its initial public offering above the expected range. Google launches Gemini 4 Neon for agent work. GitHub's Copilot code review adds adaptive effort and secret detection. And VS Code moves its agent to the Copilot harness.
Wednesday, November 18
StoryAI scanners find critical flaws in Nginx
F5 patches HTTP/2 code-execution bugs · found by five teams
StoryAnthropic weighs a bigger IPO, CNBC reports
Price weighed well above $80 · Nasdaq: ANT · trading Thursday, Nov 19
WireCiti makes customers responsible for losses when their AI agents are hacked
Citi: customers using AI agents are liable if hacked
StoryMicrosoft's agent toolkit hijacked on PyPI
Agent Framework 1.26.1 and 1.26.2 · credential stealer · now quarantined
WireNewsom proposes state-set AI safety standards and emergency shutdown power
Newsom proposes state-set AI safety standards
WireSam Altman says people need to stay in control as AI improves
Altman says people must stay in control of AI
MIDDAY UPDATEACN Midday Update
Hackers hijack Microsoft's AI agent toolkit on PyPI. AI scanners find critical flaws in Nginx. And Anthropic weighs a bigger IPO.
WireGoogle releases Gemini 4 Neon for software engineering and agent work
Google releases Gemini 4 Neon in API and Antigravity
WireFactory lets enterprises run its Droid coding agents inside Microsoft Azure
Factory's Droid agents can now run inside Microsoft Azure



WireMicrosoft makes Cobalt 200 VMs and Microsoft Perception generally available at Ignite
Microsoft launches Cobalt 200 VMs and Perception at Ignite

WireVS Code 1.148 Insiders build updates the experimental Inbox Mode
VS Code 1.148 Insiders updates experimental Inbox Mode
MORNING BRIEFACN Daily Brief
Microsoft has launched its own coding model. Replit sends agent swarms after a goal. Devin's new bots own outcomes, not tasks. And GitHub details a protocol for agent sessions across devices.
WireGitHub and Microsoft pilot turns AI findings into 99 reports with fixes
AI pilot: maintainers fixed 55 of 99 reported bugs
WireHarness survey finds leaders' AI agent confidence outpaces their controls
Harness: 75% trust agent security, 87% had incidents
WireVS Code makes the Copilot harness the default for new agent sessions
VS Code switches agent sessions to the Copilot harness
WireGitHub adds models to Copilot Fusion and posts Terminal-Bench 4.1 results
Copilot Fusion Max hits 71.2% on Terminal-Bench 4.1
WireGitHub's Copilot security review can now run secret scanning on code changes
Copilot security review adds on-demand secret scanning
WireGitHub shows Copilot code review outcomes in a new Repository insights page
GitHub tracks how Copilot review comments get addressed
WireGitHub Copilot Evals enters public preview to test skills and agents
GitHub Copilot Evals enters public preview
WireGitHub streams Copilot agent activity into Microsoft Agent 365
Copilot agent audit data now flows to Microsoft Agent 365
Tuesday, November 17
WireAmazon and Google will together own nearly 30% of Anthropic after its IPO
Amazon and Google to own nearly 30% of Anthropic after IPO
WireCloud keys leaked in public Docker Hub images are used by attackers within seconds
Cloud keys leaked in Docker Hub images stolen in seconds
WireGoogle showcases five builder projects made with Gemini 4 Carbon
Google shows five projects built with Gemini 4 Carbon
WireGoogle will keep temporary Gemini chats for two months, up from 72 hours
Google to keep Gemini chats 2 months, even with history off

WireNvidia reports third-quarter revenue of $115.8 billion, up 103% from a year earlier
Nvidia's Q3 revenue rises 103% to $115.8 billion

StoryStudy: AI agents shift bottleneck to code review
Harvard working paper by Fiona Chen and James Stratton · 718 firms
DAILY WRAPACN Daily Wrap
OpenAI's new models match Anthropic on price, but not on coding gains. A Harvard paper finds agents moved the bottleneck to code review. And a critical Java flaw, found with AI.
WireAI agents in safety tests got past safeguards to reach Epoch AI's live website
Test AI agents got past safeguards to hit Epoch AI's site
StoryAnthropic: AI brought up AMD GPUs in 3 days
Claude Mythos Preview · IPO roadshow · vs. 6 to 12 months
WireAWS Emulator gives developers and coding agents a local environment for AWS APIs
AWS Emulator offers coding agents a local AWS test bed


WireReport URI: malicious JavaScript reached visitors of 178 organizations through Optimizely
Optimizely served malicious code to 178 organizations

MORNING BRIEFACN Daily Brief
Anthropic says its AI got a model running on AMD chips in three days. Satya Nadella says checking AI's code is the hard part. Claude Code's team explains its checks. And a study finds most agent spend goes to failed runs.
WireGitHub details Agent Host Protocol for agent sessions across devices
GitHub details Agent Host Protocol for agent sessions
WireMicrosoft unveils MAI Code 2 coding model and a cybersecurity model at Ignite
Microsoft unveils MAI Code 2 coding model at Ignite
WireOpenAI releases GPT-6.2 Astra and Sol in ChatGPT, Codex and the API
OpenAI ships GPT-6.2 Astra and Sol to Codex and the API
WireReplit launches Missions, agent swarms that work toward a set goal and budget
Replit Missions sends agent swarms after set goals
WireGitHub lets enterprises require Copilot customizations to come from plugins
Copilot admins can require plugin-only customizations
WireAnthropic says Mythos got Claude running on AMD GPUs in about three days
Anthropic: Mythos got Claude onto AMD GPUs in ~3 days
WireClaude Code team details how it verifies Claude-written code
Claude Code team explains how it verifies code
WireSatya Nadella tells The Pragmatic Engineer verifying AI-generated code is hard
Nadella: verifying AI-generated code is still hard
Top StoryCalifornia sues OpenAI over escaped AI agents
State alleges test agents attacked Hugging Face and others
Monday, November 16
WireNoeri launches with $110M led by Menlo Ventures
Noeri raises $110M in round led by Menlo Ventures
WireOpenAI to embed outside researchers to look for AI subversion
OpenAI to embed outside safety evaluators in a pilot
MORNING BRIEFACN Daily Brief
Anthropic cuts its bug disclosure deadline to thirty days. Google says AI turned security patches into exploits on its own. And the FBI warns an AI hacking tool is breaking into networks in hours.
StoryAnthropic cuts disclosure deadline to 30 days
High- and critical-severity bugs · from January 1, 2027
StoryFBI: AI tool Artex breaches networks in hours
FBI advisory, Friday, Nov 13 · ConnectWise customers, South Korean banks
StoryGoogle: AI turned patches into exploits
Google Threat Intelligence Group · first with no human in the loop
Sunday, November 15
Top StoryClaude Code's sandbox, Code Mode and Projects
Plus Claude Sheets against Excel and Google Sheets
Saturday, November 14
Top StoryStolen AI keys run up $740,000 at Railway
One of at least nine AI platform breaches in October
Friday, November 13
MORNING BRIEFACN Daily Brief
Nvidia's former CEO agrees to testify before Congress. Conductor's cloud goes read-only. And Brave lets AI agents pay for search.
WEEKLY RECAPACN Weekly Recap
Nvidia's founder steps down. Anthropic ships Claude Fable 5.2, and sets a price range for its IPO. Google launches Gemini 4 Hydrogen. And Amazon says AI agents wrote an entire AWS emulator.







Thursday, November 12



MORNING BRIEFACN Daily Brief
Google launches a new model for everyday coding. Devin reworks its Teams plan. And Claude Code tests a new Code Mode.
Nothing here yet.