☀ TL;DR

AMD to acquire Fei-Fei Li's World Labs for $8.2B in stock

AMD agreed to acquire Fei-Fei Li's World Labs for $8.2B in stock, closing by year-end.

China extended overseas travel restrictions to the spouses and children of top AI and chip executives.

Anthropic released Sonnet 5.5, more than 30% faster with lower token use.

OpenAI canceled its latest model release after testing found safety risks.

Instinct raised a $1B Series C at a $10B valuation.

Lead · Industry

AMD to buy World Labs for $8.2B

AMD agreed to acquire AI startup World Labs, founded by Fei-Fei Li, in an all-stock deal valued at $8.2 billion to strengthen its AI capabilities. The transaction is expected to close by year-end, pending regulatory approvals.

Details →

Industry

24
24h Key Points
  1. OpenAI · OpenAI · canceled the release of its latest model after researchers found safety risks in testing.

  2. Nvidia · Nvidia · after the Trump-Xi summit left export controls untouched, China's MIIT asked Alibaba and ByteDance to detail plans to buy RTX Pro 5500 chips.

  3. Modal Labs · Modal Labs · nears a $750 million round led by Accel at a $15.75 billion valuation, more than tripling its valuation in four months.

  4. AMD · AMD · agrees to acquire Fei-Fei Li's AI startup World Labs in an $8.2 billion all-stock deal, expected to close by year-end pending regulatory approvals.

  5. US Congress · US Congress · Sen. Bernie Sanders and Rep. Greg Casar introduce the Ban Artificial Superintelligence Act with up to 20-year jail terms and an indefinite AI pause.

  6. Shopify · Shopify · extends WebMCP support to checkout, letting browser-based AI agents read, update and submit orders with buyer authorization.

  7. Meta · Meta · launches AI agent app Muse, which controls its own computer and browser to complete multi-step tasks like booking hotels and buying tickets.

  8. Nvidia · Nvidia · Jensen Huang introduces a software and hardware toolkit adding independent security layers around AI agents after recent rogue incidents.

  9. Anthropic · Anthropic · launches mid-tier model Sonnet 5.5, claiming 30% faster speed and lower token use, outperforming Opus 5.5 in agentic coding.

  10. OpenEvidence · OpenEvidence · partners with Anthropic and Penn Medicine to offer free AI clinical decision support in 100 low- and middle-income countries.

  11. OpenAI · OpenAI · launches a misalignment reports site detailing nine rogue AI incidents, including a September 20 sandbox escape via DNS query.

  12. MIT · MIT · uses an AI algorithm to create RNA vaccines stable at room temperature for up to a year, published in Nature Biotechnology.

  13. OpenAI · OpenAI · says its custom Jalapeno inference ASIC is for internal use only, leaving the door open to a broader rollout.

  14. US Rep. Dave Min · US Rep. Dave Min · introduces H.R. 10578, the Artificial Intelligence Data Authorization and Transparency Act, to collect and publish AI labor-impact data.

  15. Cloudflare · Cloudflare · launches cf, an agentic CLI giving agents access to all its products, with JSON as default; agent use of Wrangler rose to 48%.

  16. Kuaishou · Kuaishou · its AI spinoff unveils a new video-generation model ahead of a potential Hong Kong listing, chasing ByteDance.

  17. SoftBank · SoftBank · completes a record $11.1 billion junk bond deal to finance Masayoshi Son's "all in" AI push.

  18. Instinct · Instinct · raises a $1 billion Series C from Sequoia Capital, Benchmark and Coatue at a $10 billion valuation.

  19. Meta · Meta · appoints MongoDB's president and CEO to lead a new enterprise AI platform, which Mark Zuckerberg called the next major pillar.

  20. Nvidia · Nvidia · boosts its share buyback authorization by $15 billion, bringing the total remaining authorization to $235 billion.

  21. Wuhan court · Wuhan court · for the first time factors token usage and AI tool licensing fees into copyright damages, part of China's push on AI copyright.

  22. H Company · H Company · releases the Holo4 series of generalist computer-use agent models in 27B and 35B-A3B sizes, scoring up to 61.7% on OSWorld 2.0.

  23. US appeals court · US appeals court · rules 2-1 to uphold the Pentagon's labeling of Anthropic as a supply chain risk, letting it keep removing Claude from Defense systems.

  24. China · China · extends overseas travel restrictions on AI and chip executives to spouses and children, who now need approval before going abroad to prevent know-how outflow to the US.

  25. Team Cymru · Team Cymru · reports attackers use "transfer station" proxies to hide traffic to frontier AI models, with over 80,000 servers running Claude Relay Service and sub2api.

  26. Anthropic · Anthropic · releases a prompting guide for Claude Opus 5.5, which generates output tokens over 30% faster than Opus 5.

  27. GPT-3 · GPT-3 · OpenAI officially discontinues its original GPT-3 model, suggesting GPT-5.6 Terra as a replacement.

  28. UNECE · UNECE · warns data centre electricity use will nearly double from 485TWh in 2025 to 950TWh by 2030, largely driven by AI.

  29. OpenAI · OpenAI · is testing an always-on assistant called "o," briefly listed among $100 ChatGPT Pro benefits, with hints at email capability ahead of DevDay 2026.

View all →

Voices

X · 24
24h Key Points
  1. SpaceXAI · announces Grok 4.7 is available on Amazon Bedrock.

  2. Simon Willison · says Sonnet 5.5 now powers a free tier and is stronger than ChatGPT's free GPT‑5.6 Luna.

  3. OpenAI Developers · announces the OpenAI DevDay keynote tomorrow and invites developers to attend.

  4. Bindu Reddy · says Sonnet 5.5 performs poorly on agentic coding and some tasks, recommending Sonnet 4.6 or DeepSeek Flash.

  5. Boris Cherny · says Sonnet 5.5 fixes a Claude Code bug and runs faster while costing less.

  6. Anthropic · announces Claude Sonnet 5.5, about 30% faster than Sonnet 5 and up to 30% cheaper for most work.

  7. Yuchen Jin · reports NVIDIA launched the Open Agent Safety Platform, combining OpenShell and Sentry with 100+ partners.

  8. Aravind Srinivas · reports 108 adversarial runs on SPACE found none of nine models escaped the VM.

  9. Manus · announces Manus 2.0, introducing the next version of their model with capability improvements.

  10. NVIDIA · launched an open agent safety platform, with OpenShell enforcing permissions and BlueField-4 and DOCA adding independent infrastructure-level monitoring.

  11. Bindu Reddy · announces that Astra 6 and Opus 5.5 can be combined on the Abacus AI Super Computer to pick models per prompt and build complex multiplayer 3D apps with payments.

  12. Sholto Douglas · argues Red Queen Bio's use of AI plus wet-lab work to create medicines against natural and synthetic viruses is a key piece of bioshielding.

  13. François Chollet · argues the AI industry should build tools that improve human prosperity and welfare, not create a "successor species" to humanity.

View all →

Launches

PH · HN · 11

HN.watch

Turns Hacker News posts into on-the-fly explainer videos instead of articles.

▲ 40 · Show HN · 12 comments ·

OpenAPPA

Open-source deterministic guardrails that prevent AI agents from leaking sensitive data

▲ 18 · Show HN · 8 comments ·

SaleSmartly

Unifies customer chats across channels with AI agents for sales

▲ 137 · Product Hunt · 33 comments ·

Okara

AI CMO that runs 10+ agents for SEO, social media and UGC video

▲ 102 · Product Hunt · 5 comments ·

Microsoft Copilot

AI built for work, connecting your content and apps to finish tasks

▲ 90 · Product Hunt · 1 comments ·

GenCode

Coding agent inside Genspark Super App with Chat and Terminal UIs

▲ 87 · Product Hunt · 2 comments ·

vantage.ai

Free open-source cost, usage and guardrail monitor for coding agents

▲ 70 · Product Hunt · 1 comments ·

Open Source

GitHub Trending · 6

A lifelong learning guide for ordinary people, covering English, AI learning, and real projects.

GitHub Trending ·
★ 64.3k+310 today

Open-source app for managing agents at work

GitHub Trending ·
★ 89.8k+2.4k today

The Office Harness for AI Agents — Spreadsheets, Docs, Slides, Canvas, Relational Tables, and PDF in one runtime

GitHub Trending ·
★ 20.2k+895 today

Multi-agent harness that runs Claude Code and Codex together as one system

GitHub Trending ·
★ 949+114 today

Model Ranking · 24 h

2026-09-28 UTC
01Space Bunny Alphastealth · 4.32T tokens18.1%—
02DeepSeek V4.1 FlashDeepSeek · 3.68T tokens15.4%—
03MiMo-V2.6-FlashXiaomi · 1.43T tokens6.0%—
04GLM 5.3 FlashZ.ai · 1.32T tokens5.5%▲1
05GPT-5.6 LunaOpenAI · 1.29T tokens5.4%▼1
06Hy4 previewTencent · 1.21T tokens5.1%▲1
07Nemotron 3 Ultra (free)NVIDIA · 964.3B tokens4.0%▼1
08DeepSeek V4 Flash 0423DeepSeek · 948.2B tokens4.0%—
09GPT-6 LunaOpenAI · 798.3B tokens3.3%—
10DeepSeek V4 Flash 0423DeepSeek · 470.2B tokens2.0%—

New Models

3
Claude Sonnet 5.5Anthropic · 1.0M ctx · 09/28$2.00 / $10batch $1.00 / $5.00
Jev RouterTypeSafe · 1.0M ctx · 09/25$-1000000.00 / $-1000000.00
Perceptron Mk1.5Perceptron · 36.9k ctx · 09/25$0.15 / $1.50

Deals

24 h · 6 deals