Trending AI News & Updates Past Week — Nov 13–20, 2026

The AI news and updates that matter over the past 7 days, curated from across the AI ecosystem, ranked.

Top 30 of 356
Hot

Nous Research Brings Hermes Agent to Microsoft Teams and Azure

Nous Research launched Hermes Enterprise, bringing its agent to Microsoft Teams, the Microsoft Store, and Azure. The agent executes workloads in isolated Azure Container Apps sandboxes and saves recurring task approaches as reusable skills. Through Collective Wisdom, teams share these skills after review. Organizations with strict data residency requirements can run the agent in their own Azure tenant.

Read more
Hot
OpenAIOpenAI2h ago

OpenAI Expands Dots Always-On Agents to Plus and Business Standard

OpenAI is testing dots, its always-on agent, with a small number of ChatGPT Plus users. The company is also rolling out dots to all eligible ChatGPT Business Standard members in the United States starting today. Dots function as autonomous agents that maintain context and progress on tasks between conversations using connected apps and cloud-based computing.

Read more
Viral
OrcaRouter 🐳OrcaRouter 🐳19h ago

OrcaRouter Launches Shield to Deceive and Neutralize Offensive AI Agents

OrcaRouter launched Shield, an AI-native defense that lures offensive agents into synthetic environments rather than blocking them. By serving fake endpoints and decoy credentials, Shield diverts probes and triggers WAF blocks. Internal benchmarks show an 87% diversion rate and 5x more attacker time wasted, with the system now available for deployment at the edge.

Hot
depthfirstdepthfirst2h ago

Depthfirst Discloses Critical PHP Session Vulnerability CVE-2026-126714

Depthfirst disclosed phpAfterShell (CVE-2026-126714), a critical CVSS 9.2 vulnerability in PHP’s default file-based session handling that allows unauthenticated remote code execution. Discovered using AI models, the flaw affects PHP versions below 8.5.13, 8.4.28, 8.3.37, and 8.2.36. Remediation requires upgrading to these patched releases or disabling session upload progress in php.ini to remove the primary attack path.

Hot
AntinuousAntinuous2h ago

Antinuous Launches Research Platform for Autonomous Experiments and Simulations

Antinuous launched its research and experimentation platform in public beta, combining swarm orchestration, compute clusters, and simulation software to run experiments 100x faster and 1,000x cheaper. The system enables massive parallel execution, with test builds including an F1 car digital twin and datacenter cooling models completed in hours. The platform is now available for public use.

Viral
ChatGPTChatGPTNov 18

OpenAI Launches ChatGPT for Accounting with New Plugin and Partner Integrations

OpenAI launched ChatGPT for Accounting, featuring a new plugin and integrations with QuickBooks, Xero, Oracle NetSuite, BILL, FreshBooks, Rillet, and Campfire. The system automates invoice processing, bank reconciliation, month-end reporting, and audit preparation. It requires the ChatGPT desktop app and supports reusable skills for monthly tasks while maintaining data privacy controls.

Read more
Viral
NVIDIANVIDIANov 13

NVIDIA Announces CEO Resignation, Interim Leadership, and New Board Chair

NVIDIA announced that Jensen Huang has resigned as President, CEO, and board member, effective immediately. CFO Colette Kress will serve as interim CEO while the board searches for a permanent successor. Additionally, the company appointed Harvey C. Jones as its first-ever Chair of the Board. NVIDIA's third-quarter fiscal 2027 financial results remain scheduled for November 17.

Viral
AnthropicAnthropicNov 19

Anthropic Prices Initial Public Offering at $83.00 Per Share

Anthropic priced its initial public offering of Class A common stock at $83.00 per share. The offering includes 980,000,000 shares, with trading expected to begin on the Nasdaq Global Select Market under the ticker ANT on November 19, 2026. Anthropic also granted underwriters a 30-day option to purchase up to 147,000,000 additional shares.

Read more
Hot
apekapek19h ago

Apekshik Panigrahi Releases AnyMachine for Agentic 3D Modeling and Rigging

Apekshik Panigrahi released AnyMachine, an open-source workflow where an AI agent uses Claude Fable 5.2 to generate reference sheets and build, rig, and animate mechanical 3D models in Blender. The system automates the process from concept to game-ready asset, including part-by-part modeling and automated verification. The project is currently experimental, with six runs completed to date.

Read more
Hot
AnthropicAnthropic13h ago

Anthropic CEO Dario Amodei Outlines Safety Commitments for Public Company

Anthropic published remarks from CEO Dario Amodei on the company's initial public offering, which priced at 83.00 dollars per share today. Amodei pledged to pace AI training and deployment to ensure safety, stating the company will not bend to business or political pressure. He also noted that while AI could accelerate scientific progress, current results remain in early stages.

Read more
Architect LabsArchitect LabsNov 19

Architect Labs Unveils Frostbite, an AI-Designed CPU for Interaction Nets

Architect Labs introduced Frostbite, a general-purpose CPU built around interaction nets and co-designed end-to-end by AI agent swarms. The architecture eliminates traditional caches and locks, achieving 4.59x the performance-per-watt of modern superscalar CPUs. The agents autonomously discovered work-stealing rules to manage parallelism, and the company plans to release the RTL and ISA in future updates.

Viral
HeyGenHeyGenNov 17

HeyGen Launches Avatar 6 With Context-Aware Gestures and High Realism

HeyGen launched Avatar 6, an AI avatar model that generates gestures based on script meaning. In HeyGen’s internal testing, 84% of viewers could not distinguish the avatar from the real person. The model, which currently ranks first on New Measure’s Avatar Arena, requires up to 30 seconds of video and a photo to generate. A live demo occurs November 18.

Read more
GoogleGoogleNov 19

Google Launches Genie 4 World Model in Project Genie Experience

Google released Genie 4, its most realistic world model, to eligible Google AI Ultra subscribers in Project Genie. The update adds mid-session text-prompt environment editing with native audio and dialogue, a conversational creative agent, and a remixable community gallery. Genie 3 upgrades include three-minute sessions, world chaining, and 360-degree panorama grounding. Google AI Pro members can join the waitlist.

Read more
MonidMonidNov 19

Monid Launches Agentic Private Market Data Service at $0.10 Per Request

Monid launched a private market data service for AI agents, providing access to 20M+ companies inside Claude. The service operates on a pay-per-request model at $0.10 per company profile, with no subscription fees or markups on third-party provider rates. New accounts receive $1 in free credit to query funding rounds, valuations, and financial data directly through agent workflows.

Read more
KaiKai19h ago

OpenAI Details Safety and Risk Management Across Frontier Model Lifecycle

OpenAI published a recap of its frontier model safety lifecycle, detailing four protection layers including alignment, sandboxing, and monitoring. The company mandates safety reviews between research, internal, and external deployment stages. Frontier training occurs without live internet access, and a central Control Room halts model output within 35 seconds if security invariants are breached or safety standards are unmet.

Read more
ReplitReplit2h ago

Replit Releases Amuse, a Self-Hosted Open-Source Personal AI Assistant

Replit released Amuse, a free, open-source AI assistant that runs locally. Built in three days by Replit Missions, the agent researches, browses, sets reminders, and builds small apps. It requires an OpenAI API key and PostgreSQL, keeping all data and logins local. The agent requests approval before performing consequential actions like payments or submissions.

Read more
Mike HenryMike Henry2h ago

Parasail Deploys Positron Titan Systems for Cost-Optimized AI Inference

Parasail is adding 5MW of Positron Titan inference systems to its multi-silicon cloud, targeting trillions of tokens per day by late 2027. The company claims Titan delivers 4X better cost-per-token than next-gen NVIDIA or AMD systems. This hardware orchestrates across DRAM and SRAM to support large-scale agent swarms, background agents, and physical AI workloads.

CognitionCognition2h ago

Cognition Adds Gemini 4 Neon to Devin Desktop and CLI

Cognition integrated Google's Gemini 4 Neon into Devin Desktop and Devin CLI. On the FrontierCode 2.0 benchmark, the model scores 46.2% at $4.38 per task. This performance places it within one point of Claude Opus 5.5, which scores 47.3% at a cost of $12.82 per task, making Neon the strongest Gemini model on the benchmark.

Read more
DeepSeekDeepSeekNov 13

DeepSeek API Adds Full App Search, File, and Voice Capabilities

DeepSeek now offers its full app conversation stack via API, including web search, deep thinking search, file input, URL reading, safety checks, and voice. The API handles tool calls and context assembly server-side for the deepseek-flash model. These features are available across Messages, Responses, and Chat Completions protocols, with additional tool-usage pricing for search, OCR, ASR, and TTS.

Viral
Grok BotGrok BotNov 18

xAI Deploys Five Grok Bots to Handle End-to-End Accounting Processes

xAI assigned five Grok Bots to its accounting team to manage the order-to-cash process, from provisioning accounts to collecting payments. These agents operate independently using internal applications and read-only manuals, escalating only when human judgment is required. The deployment reduces invoice issuance time from 24 hours to minutes and saves approximately eight hours of accounting work daily.

Read more
Alexandr WangAlexandr WangNov 18

Meta Adds Muse AI Voice to Apple Watch and Wear OS

Meta adds voice support for its Muse AI assistant to Apple Watch and Android Wear OS. The update brings voice calls, text messaging, and dictation capabilities directly to wearable devices. Muse is now accessible on these platforms for users with the assistant configured on their mobile devices.

Read more

Artificial Analysis Compares GPT-6.2 Astra and Dreamina Seedance 2.5

Artificial Analysis tested whether language models can compete with video models by comparing OpenAI's GPT-6.2 Astra and ByteDance's Dreamina Seedance 2.5. Astra generated three scenes as three.js code with a scripted camera path, while Seedance 2.5 received the same path as a text prompt. The firm published a side-by-side video and opened public voting arenas for both models.

Read more
Stitch by GoogleStitch by GoogleNov 18

Google Labs Launches Stitch Loop Product Agent for Automated Fixes

Google Labs launched Stitch Loop, a product agent that monitors live production data to identify and resolve security, latency, and accessibility issues. The agent proposes code-based fixes via reviewable pull requests and supports custom loops across 180+ integrations. Stitch Loop integrates with coding agents via CLI and MCP, with a global waitlist now open for priority access.

Read more

Artificial Analysis Benchmarks Google Gemini 4 Neon Intelligence and Cost

Artificial Analysis reports Google’s Gemini 4 Neon scores 61 on its Intelligence Index, ranking third overall. It costs ~17% less per task than Gemini 4 Carbon and runs ~40% faster. In coding agent benchmarks, Neon scores 59, matching Claude Code configurations at roughly half the cost. It is available via Google’s API, Antigravity, and the Gemini app.

Read more
Viral
Fish AudioFish AudioNov 13

Fish Audio Launches Fish Agent Platform for Custom AI Voice Agents

Fish Audio launched Fish Agent, a no-code platform for building custom AI voice agents that make and take calls. The platform features a median deployment time of 32 seconds and uses the full-duplex Fish Live 3 model for natural conversation. It supports custom LLMs, tool integrations, and dedicated phone numbers, with pricing starting at $0.035 per minute.

Read more
PierrePierreNov 19

Pierre Releases Words, a High-Performance Streaming Markdown Parser

Pierre released Words, a streaming Markdown parser written in WebAssembly Text designed for LLM responses. It renders Markdown to HTML, supports syntax highlighting, and includes Mermaid diagram support. Benchmarks show Words is up to 2× faster than md4x and 20× faster than Streamdown, with significantly lower memory usage and bundle sizes for React applications.

Tencent Hunyuan Releases Open Weights for Multimodal Hy4 MoE Model

Tencent Hunyuan released open weights for Hy4, a natively multimodal Mixture-of-Experts model featuring 770B total parameters and a 1M-token context window. The release includes instruct, FP8, and base model variants alongside a comprehensive system card. Together AI hosts the model, while vLLM and SGLang provide day-0 serving support for production deployment.

Read more
VercelVercel10h ago

Guillermo Rauch Analyzes AI's Impact on Engineering Productivity

Vercel CEO Guillermo Rauch argues AI has redefined engineering value, creating 1000x outcome-oriented builders while rendering trivia-focused engineers -10x. Citing Vercel’s internal token-consumption data across startups and enterprise, he identifies coding as AI’s primary killer app. Regular knowledge workers see a 2x productivity gain in this shift.

Read more
CognitionCognitionNov 17

Cognition Launches Devin Bots to Coordinate Autonomous Multi-Agent Engineering Teams

Cognition launched Devin Bots, where a lead agent directs a team of specialized agents through planning, code review, and automated testing. Beta testers report merging 4.5x more pull requests per message, with some engineers increasing output from 50 to nearly 2,000 pull requests weekly. Bots integrate with Slack, text, and phone for status updates.

Read more
NVIDIANVIDIANov 18

NVIDIA Reports Q3 Fiscal 2027 Revenue of $115.8 Billion

NVIDIA reported third-quarter fiscal 2027 revenue of $115.8 billion, a 103% increase from a year ago. Data Center revenue reached $107.6 billion, up 110% year over year. The company also shipped its Vera Rubin NVL72 platform, which it describes as its fastest product ramp to date, and announced a pending acquisition of Hugging Face.

AI Pulse(past week)
356
updates
81
sources
17
viral

Activity past week

Nov 14updates / dayNov 20

Top updates

  • #1
  • #2
  • #3
  • #4
  • #5

Most active

  • Artificial AnalysisArtificial Analysis24
  • VercelVercel20
  • OpenAIOpenAI16
  • ArenaArena13
  • PerplexityPerplexity11
  • falfal10

By category

  • Product Update154
  • Product Launch97
  • Industry Analysis54
  • Company News33
  • Research18

Technical depth

55% Technical45% General

About this page

Keeping up with AI is exhausting — launches, new tools, research, and company moves land every day, scattered across X, Reddit, blogs, and newsletters, and it's easy to miss an update that could impact your work. HeadsUpAI tracks the whole AI ecosystem and surfaces the most significant AI news and updates over the past 7 days. Each one gives you the whole story in a 30-second read, plus the backstory of what led to it and what other companies are doing — presented straight, no hype, no bias — so you keep up with the AI ecosystem without the noise, and act on what matters.

Frequently asked questions

The past week's biggest AI stories: Nous Research Brings Hermes Agent to Microsoft Teams and Azure, OpenAI Expands Dots Always-On Agents to Plus and Business Standard, OrcaRouter Launches Shield to Deceive and Neutralize Offensive AI Agents, Depthfirst Discloses Critical PHP Session Vulnerability CVE-2026-126714, and Antinuous Launches Research Platform for Autonomous Experiments and Simulations — among the most significant AI launches, releases, and company moves of the last 7 days. HeadsUpAI ranks them by significance, so the updates that matter most appear first.

HeadsUpAI tracks the AI ecosystem — models, tools, research, and companies — and surfaces the most significant updates as a 30-second read, filtered to your role and interests. This page rounds up the biggest AI news from the past 7 days.

HeadsUpAI covers AI model releases, product launches, product updates, company news, research, and industry analysis from across the AI ecosystem. Every update is curated as a 30-second read, presented straight — no hype, no bias.

Continuously. HeadsUpAI adds significant AI updates throughout the week as they're announced, usually within hours, ranked by significance rather than recency.