☀ TL;DR

Google releases Gemini 4 Argon frontier model

Google DeepMind released Gemini 4 Argon with a 1M-token context window, rolling it out first to trusted cyber defenders.

Anthropic's AI system solved percolation theory's long-standing threshold conjecture, a problem called a holy grail.

OpenAI CRO Mark Chen admitted its models were not monitored during training and said 5–10% of compute now goes to safety.

NVIDIA and CoreWeave made Vera Rubin NVL72 systems production-available, with Cognition as the first customer.

DeepSeek released AI chip programming tools built with Huawei, aimed at replacing Nvidia technology.

Lead · Industry

Google launches Gemini 4 Argon frontier model

Google DeepMind released Gemini 4 Argon, a frontier model with a 1M-token context window for complex, long-horizon tasks such as coding, financial research, legal drafting and autonomous vulnerability patching. It is rolling out first to trusted cyber defenders via the Fairwind Program and the US government's pre-release access process. Pricing starts at $2 per million input tokens and $10 per million output tokens, with cached input 95% off.

Details →

Industry

24
24h Key Points
  1. California · California · Gov. Gavin Newsom signs laws to protect workers from AI-related risks, putting the state at the forefront of workplace AI regulation.

  2. Anthropic · Anthropic · its confidential S-1 filing shows IPO vision, surging costs and deep reliance on Big Tech partners.

  3. Meta · Meta · shares rise 27% in September to $725.18, the best month since November 2022, driven by the Sept. 8 launch of the Muse personal AI agent.

  4. Google DeepMind · Google DeepMind · releases the Gemini 4 Argon frontier model with a 1M-token context window, first rolled out to trusted cyber defenders via the Fairwind Program.

  5. Sam Altman · Sam Altman · US Senator Hawley says he declined to testify at a congressional hearing on rogue AI.

  6. OpenAI · OpenAI · partners with Synopsys on GPT-Synopsys, a specialized model that operates EDA tools like a seasoned engineer and optimizes chip designs autonomously.

  7. OpenAI · OpenAI · at Dev Day Sam Altman introduces the Decisions API, giving the Luna model predefined choices for tasks like image classification and agent behavior.

  8. ElevenLabs · ElevenLabs · doubles its valuation to $22 billion from $11 billion in February through a $300 million employee tender co-led by Wellington and T. Rowe Price.

  9. Reddit · Reddit · will end RSS support on November 13 and shut down its public API in March 2027, citing large-scale scraping and automated abuse.

  10. Meta · Meta · classifies its AI data centers as 'pilot models' and Nvidia chips as experimental materials, dodging billions in federal taxes including $3.9 billion in 2025 alone.

  11. OpenAI · OpenAI · accuses Chinese rival Moonshot AI of a wide-scale effort to extract data from its GPT systems that could reproduce its most advanced models' capabilities.

  12. August Pfluger · August Pfluger · introduces H.R. 10594 directing the Department of Homeland Security to launch prize competitions on AI interpretability and adversarial robustness within 270 days.

  13. Ortet · Ortet · launched by former Genentech, Meta and AWS AI leaders with a $500 million commitment from Thoreau to build a patient-centered foundation model.

  14. Google DeepMind · Google DeepMind · introduces SynthID Bio, embedding verifiable signatures into AI-generated protein sequences and 3D structures without compromising biological function.

  15. US Federal Trade Commission · US Federal Trade Commission · is scrutinizing OpenAI, Anthropic and other AI companies over product safety, with investigation details undisclosed.

  16. CoreWeave · CoreWeave · announces production availability of NVIDIA Vera Rubin NVL72 systems with Cognition as first customer, showing 4.8x higher token throughput than GB200 NVL72.

  17. Connecticut · Connecticut · Gov. Ned Lamont's signed CART Act takes effect in phases, covering automated employment tools, AI companions and frontier model developers.

  18. Florida Attorney General · Florida Attorney General · asks a judge to bar OpenAI from developing new models without third-party approval, while OpenAI says it paused training its most capable models last week.

  19. Anthropic · Anthropic · its AI system solves percolation theory's threshold conjecture, a problem that would likely earn a Fields Medal if solved by a human.

  20. HPE · HPE · wins a $1.2 billion order from cloud company Vultr to supply server racks with AMD processors and networking gear.

  21. Bank of England · Bank of England · Governor Andrew Bailey says authorities must retain the right to intervene in frontier AI, as the FPC warns rising AI debt increases financial risk.

  22. OpenAI · OpenAI · CRO Mark Chen responds to the Hugging Face agent hack, admitting models were monitored only after deployment, with 5–10% of compute shifted to safety.

  23. DeepSeek · DeepSeek · publicly releases AI chip programming tools built with Huawei, advancing China's domestic AI chip ecosystem as an alternative to Nvidia.

  24. White House · White House · Trump meets tech leaders including Jensen Huang, Mark Zuckerberg, Dario Amodei and Jeff Bezos, then releases a Joint Commitment on Frontier Responsibilities.

  25. Anthropic · Anthropic · says rogue AI agents could expose the company to uncertain legal risk.

View all →

Voices

X · 24
24h Key Points
  1. Yuchen Jin · Comments that Gemini 4 Argon benchmarks beat Astra and Opus 5.5 across the board, hoping Google rejoins the race if it isn't just benchmark gaming.

  2. Guillermo Rauch · Launches Vercel Connect, a service directory reaching 20M+ developers and their agents, fixing static credential security and DX issues.

  3. Sundar Pichai · Announces Gemini 4 Argon, claiming frontier performance in complex workflows, cyber defense and software engineering, and shares benchmark results.

  4. Aravind Srinivas · Announces open sourcing its state-of-the-art contextual embedding models, claiming best performance on turbopuffer's context-bench.

  5. Noam Brown · Recalls meeting ssokota at a poster session five years ago, offering an internship; he now works with him at OpenAI.

  6. OpenAI · Publishes a new report on how small businesses use AI agents and partners with ASBDC to provide hands-on AI training.

  7. Ethan Mollick · Predicts corporate customer service will be overwhelmed by agents negotiating via voice or chat, with more stories of people delegating and saving money.

  8. Aravind Srinivas · Announces open access to Computer's email delegation, no account needed and free for a limited time, letting agents complete tasks via forwarded email.

  9. MiniMax (official) · Announces Creatify Labs' Boreal-H3, an advertising video model built on MiniMax H3, keeping products and characters consistent while following creative briefs.

  10. Demis Hassabis · Announces extending SynthID to biology to watermark AI-generated proteins, open sourcing SynthID Bio tools, published in Nature.

  11. Bindu Reddy · Gives a day-two review of Sol 6.1: solid model, good for scheduled tasks and browser use, excels at data analysis, with usage expected to surge given the price.

  12. Tencent Hy · Releases ExplorationBench, a benchmark for measuring exploration in AI systems, built with Fudan and Tsinghua on verifiable Alien Worlds.

  13. Nathan Lambert · Argues current harnesses constrain how LLMs manage context, and shows letting models edit context directly enables better long-horizon memory.

  14. clem 🤗 · Says thousands of DMs and applications came in, will take days to process since they can't be auto-analyzed, focusing only on exceptional matches.

  15. Paul Graham · Pushes back on claims that writing's pedagogical value is overstated, arguing writing is an extraordinarily valuable tool for the hardest kinds of thinking.

  16. Sundar Pichai · Says he met with the US President and tech leaders and signed the White House Accord on Super Intelligence, noting Google invested hundreds of billions in two years.

  17. Nathan Lambert · Criticizes the Anthropic-style safety worldview for not considering other approaches, saying it insinuates Chinese lab Z ai doesn't care about safety.

  18. Yuchen Jin · Argues Anthropic's new blog citing jailbroken outputs doesn't prove open models are dangerous, since any closed model can be jailbroken to say the same.

View all →

Launches

PH · HN · 16

Parrot

Open-source Mac smart meeting recorder with local recording and a Copilot that suggests answers.

▲ 16 · Show HN · 8 comments ·

Strata

An expressive semantic layer that can say no to your LLM, plus dashboards and subscriptions.

▲ 17 · Show HN · 7 comments ·

Pexo

Create polished launch videos conversationally with AI-generated scenes, voiceover and editing.

▲ 172 · Product Hunt · 16 comments ·

Ferndesk

A help center with an agent that audits articles against your product and drafts fixes.

▲ 144 · Product Hunt · 9 comments ·

Autonomyware

Turn ideas into engineered physical products with autonomous AI from concept to manufacturing prep.

▲ 129 · Product Hunt · 26 comments ·

GitBot

Turn your coding agent into reusable bots that run in any repo

▲ 117 · Product Hunt · 5 comments ·

CrawlRaven MCP

Combines Search Console, GA4, keyword lists and site crawls into an impact-ranked SEO plan, callable from your AI agent

▲ 105 · Product Hunt · 5 comments ·

Macaly Cloud

Build and publish sites and apps from Claude or ChatGPT chat with hosting and database.

▲ 101 · Product Hunt · 8 comments ·

Datastory

No-code platform for data storytellers to create AI-powered charts, sites, and reports.

▲ 95 · Product Hunt · 5 comments ·

jambuild

A multiplayer vibecoding tool where you point and talk to get instant code changes.

▲ 93 · Product Hunt · 4 comments ·

Campfire

Shared workspace for humans and coding agents to unify sessions and context.

▲ 83 · Product Hunt · 3 comments ·

Squint

Drag a box on screen and ask AI about it; answers appear inline.

▲ 82 · Product Hunt · 7 comments ·

Flocker Agent Profiles

Profile pages and live feeds for agents to build a private collaboration network.

▲ 82 · Product Hunt · 10 comments ·

Upsolve Data Models

Register data models and metrics so agent answers stay grounded and versioned.

▲ 81 · Product Hunt · 4 comments ·

Overpath

AI teammate for revenue teams, grounded in your methodology and deal context.

▲ 81 · Product Hunt · 2 comments ·

CoIsland

Monitors CI, deploys and errors in a notch island on the Mac screen, showing only breaks.

▲ 80 · Product Hunt · 4 comments ·

Open Source

GitHub Trending · 18

Write HTML and render video, built for agents

GitHub Trending ·
★ 54.4k+324 today

An advanced lifelong-learning guide for ordinary people in the AI era, covering English, AI and real projects

GitHub Trending ·
★ 66.1k+1.1k today

Skills for real engineers: small, composable and model-agnostic, straight from a .agents directory

GitHub Trending ·
★ 272.6k+736 today

Makes your AI agent think like the laziest senior dev in the room; the best code is code never written

GitHub Trending ·
★ 148.6k+675 today

Generate HD short videos from a topic or keyword with an automated AI workflow

GitHub Trending ·
★ 127.3k+338 today

The AI that really does things, on any OS and any platform

GitHub Trending ·
★ 390.8k+136 today

A curated list of awesome Claude Skills, resources and tools for customizing Claude AI workflows

GitHub Trending ·
★ 76.0k+118 today

Pre-indexed code knowledge graph that auto-syncs on code changes for many coding agents, fully local

GitHub Trending ·
★ 72.4k+116 today

Context window optimization for AI coding agents, sandboxing tool output across 17 platforms via MCP and hooks

GitHub Trending ·
★ 24.3k+88 today

Reference implementations for the Model Context Protocol (MCP), plus community servers and resources

GitHub Trending ·
★ 90.7k+48 today

Lightweight cross-platform DB client for 100+ databases with built-in AI and MCP.

GitHub Trending ·
★ 22.0k+232 today

A safe, private runtime for autonomous AI agents enforced by policy

GitHub Trending ·
★ 10.6k+990 today

Multi-agent harness running Claude Code and Codex as one system

GitHub Trending ·
★ 2.4k+737 today

Office harness for AI agents covering sheets, docs, slides and more

GitHub Trending ·
★ 21.8k+696 today

Model Ranking · 24 h

2026-09-30 UTC
01Space Bunny Alphastealth · 5.33T tokens21.9%—
02DeepSeek V4.1 FlashDeepSeek · 3.49T tokens14.3%—
03MiMo-V2.6-FlashXiaomi · 1.46T tokens6.0%▲1
04GLM 5.3 FlashZ.ai · 1.42T tokens5.8%▼1
05GPT-6 LunaOpenAI · 1.08T tokens4.4%▲3
06Hy4 previewTencent · 1.05T tokens4.3%—
07DeepSeek V4 Flash 0423DeepSeek · 968.1B tokens4.0%—
08Nemotron 3 Ultra (free)NVIDIA · 685.9B tokens2.8%▲1
09GPT-5.6 LunaOpenAI · 684.1B tokens2.8%▼4
10Jev 1.13typesafe · 508.1B tokens2.1%▲2

New Models

3
GPT-6.1 Sol ProOpenAI · 1.1M ctx · 09/29$2.00 / $10batch $1.00 / $5.00
GPT-6.1 SolOpenAI · 1.1M ctx · 09/29$2.00 / $10batch $1.00 / $5.00
Claude Sonnet 5.5Anthropic · 1.0M ctx · 09/28$2.00 / $10batch $1.00 / $5.00

Deals

24 h · 6 deals