Skip to content

Search stories, models, companies…Subscribe

The river

Latest

Every story as it’s published, newest first.

Friday, Oct 16

2 stories

Open source · NewsNvidia Nemotron 3.5 Super VL and Bolt: open weights under OpenMDW, and what the Opus 5 comparison really showsNvidia released Nemotron 3.5 Super VL, a 120B-parameter (12B active) hybrid Mamba-Transformer model with a 1M-token context, and Nemotron 3.5 Bolt 74B-A9B on October 15, 2026, with a 4 GB Bolt E4B due October 23. The license is permissive; the headline wins over Claude Opus 5 and GPT-6.1 Sol come from customized versions on narrow tasks.07:21·6 min read

Open source · NewsFLUX 3 Image Klein: Black Forest Labs' 5.6B open-weight image model, and what 'free under $5M' really coversBlack Forest Labs released FLUX 3 Image Klein on October 15, 2026: a 5.6B-parameter open-weight model for image generation and editing with native 4K output. Its FLUX Kommunity License lets companies under $5 million in revenue use it commercially, but only to generate images, not to distribute or host the model.07:21·5 min read

Wednesday, Oct 14

19 stories

Policy · AnalysisAnthropic and OpenAI sought control over what independent AI evaluators publish, Reuters reports. Why evaluator independence mattersReuters reported on October 13, 2026, that as recently as October 5, Anthropic and OpenAI were seeking strict editorial control over what outside safety evaluators could publish, weeks after both CEOs pledged evaluators would publish without it. What was reported, what the companies say, and why it matters for every safety claim you read.08:36·7 min read

Policy · NewsBrad Smith calls for global AI safety rules: independent evaluations, mandatory incident reports and an emergency brakeIn a BBC interview released October 12, 2026, and in remarks in Dublin a day later, Microsoft President Brad Smith called for global AI safety rules, including outside evaluation before release, mandatory incident reporting, notification of people harmed by AI agents and an emergency brake. He said the UK is best placed to lead.08:35·5 min read

Models · ExplainerAnthropic's distillation defenses explained: preserved thinking, extraction classifiers and account bansOn October 12, 2026, Anthropic said new defenses against distillation have cut the data harvested from Claude by around 90 to 99%. What distillation is, why reasoning transcripts are the prize, how preserved thinking and the new classifiers work, and what they change for developers.08:35·7 min read

Developers · NewsAnthropic Scion: the in-house source control system that replaced Git for its monorepo (and why it isn't open source)On October 13, 2026, Anthropic described Scion, a version control system it built to replace Git for its monorepo after daily merges grew from about 2,500 in May to more than 30,000 in September. Despite some reports, Scion isn't open source: Anthropic says it isn't a general replacement for Git and is sharing the ideas instead.08:35·6 min read

Open source · NewsLiquid AI LFM2.5-24B-A2B: a 24B open-weight MoE with 2B active parameters, its license and what it needs to runLiquid AI released LFM2.5-24B-A2B on October 13, 2026: a 24-billion-parameter mixture-of-experts reasoning model with about 2.45B active parameters and a 131K context. The weights are free to download, but Liquid's license excludes commercial use by companies with $10 million or more in annual revenue.08:35·5 min read

Products · NewsApple HomeView and HomePod mini 2: price, release dates, Siri AI and why HomeView skips the EUApple announced the $299 HomeView smart display and the $149 HomePod mini 2 on October 13, 2026. HomeView ships November 13 with Siri AI in beta; HomePod mini 2 arrives October 30 but gets Siri AI only in a later update. HomeView won't be sold in the EU.08:34·5 min read

Hardware · NewsIntel GPU D270: Intel's inference GPU trades HBM for up to 480 GB of LPDDR5XAt the OCP Global Summit on October 12, 2026, Intel named its Crescent Island accelerator the GPU D270: an air-cooled 350 W PCIe card with 1.3 petaflops of FP8, 160 GB of LPDDR5X on Intel's own card and up to 480 GB on partner cards. Customer sampling has started; Intel hasn't given a launch date.08:34·4 min read

Industry · NewsOpenAI in early talks to acquire Trail of Bits, The Information reports: what's known about the deal and Patch the PlanetThe Information reported on October 13, 2026, that OpenAI is in early talks to buy Trail of Bits, the security firm it already works with on Patch the Planet, an open-source bug-fixing initiative. Deal size is unknown and neither company has commented publicly.08:34·4 min read

Industry · NewsNTT Data and Anthropic reportedly near deal on a 500MW Osaka datacenter: what's reported and what isn't confirmedNikkei reported on October 13, 2026 that NTT Data is in the final stages of talks to build Anthropic's first dedicated datacenter in Japan, in Osaka, with up to 500MW of power capacity and construction costs of up to ¥1 trillion. Neither company has confirmed it.08:34·4 min read

Models · ExplainerOpenAI loop transformers: what's reported about GPT-6.1 Sol and OpenAI's next model, and what a looped transformer isThe Information reported on October 13, 2026, that OpenAI is working through safety issues tied to loop transformers before shipping its next flagship model. What's actually confirmed, what OpenAI's GPT-6.1 Sol safety addendum shows, how the cancelled GPT-6.1 Astra fits in, and how looped transformers work.08:34·7 min read

Research · AnalysisCan AI models self-replicate on IoT devices? What Anthropic's Frontier Red Team study actually showsAnthropic's Frontier Red Team says small open-weight AI models may be months away from being able to self-replicate on consumer and IoT devices. What it tested, what the 8-of-9 and 8-of-8 results mean, and what the study does not show.08:34·8 min read

Models · NewsGemini 4 Argon is now generally available: price, who gets it and what Google claimsGoogle opened Gemini 4 Argon to paid API developers, AI Pro and Ultra subscribers and businesses on October 13, 2026. The Gemini 4 Argon price is $2/$10 per million tokens through November 30, then $4/$20, and free users are still left out.08:09·6 min read

Developers · GuideHow to use Gemini 4 Argon: in the Gemini app, AI Studio, the Gemini API and AntigravityHow to use Gemini 4 Argon depends on where you want it: AI Pro or Ultra for the Gemini app, Ultra for AI Studio's Playground and Antigravity, and a billing-enabled key for the API. Plans, steps from Google's docs, sample code, costs, limits and fixes.08:08·8 min read

Models · CompareGemini 4 Argon vs Claude Opus 5.5 vs GPT-6 (Sol/Astra): price, context, access and benchmarks (October 2026)Gemini 4 Argon vs Claude Opus 5.5 vs GPT-6 Astra and GPT-6.1 Sol, compared on official prices, context and output limits, where you can use each, and what vendor and independent benchmarks do and don't show.08:08·Checked Oct 14

Research · AnalysisWhat the Gemini 4 Argon model card says: safety tests, cyber mitigations, Fairwind access and the limitsGoogle's Gemini 4 Argon model card says the company can't rule out that Argon reached its critical capability levels for biological and cyber risk, so it shipped with safeguards and a safety case. Here's what the card and Google's safety post report, how they compare with the Claude Opus 5.5 system card, and what's missing.08:08·9 min read

Developers · NewsMicrosoft October 2026 Patch Tuesday: a record 1,400+ CVEs, one exploited Windows Shell zero-day, and what to patch firstMicrosoft's October 2026 Patch Tuesday, released October 13, fixed more than 1,400 CVEs by the count in Microsoft's Security Update Guide, the first four-digit monthly release and a record for the second month running. One Windows Shell bug is under active attack and on CISA's KEV list, and one of its credited finders used Claude.08:08·5 min read

Developers · NewsLinus Torvalds used Claude Code to draft a Linux kernel page cache speedup: what it does and what Claude wroteOn October 10, 2026, Linus Torvalds published an experimental eight-patch Linux kernel series that lets many threads read the same cached file without fighting over one counter. He says Claude Code wrote the code, tests and first draft of the description under his direction, and measurements on his laptop show about 3x throughput in one narrow test.08:08·5 min read

Models · NewsMeta Muse Wonder and Muse Wonder Full: price, availability and what Artificial Analysis foundMeta released Muse Wonder and Muse Wonder Full on October 12, 2026. Muse Wonder costs $4/$20 per million tokens on Meta's API; Wonder Full is only in paid Muse Code plans. Artificial Analysis scores them 52 and 56 on its Intelligence Index.08:07·5 min read

Policy · NewsHinton and Bengio open letter: 400 scientists call for AI containment, air-gapping and an international incident registryAn open letter released October 13, 2026, signed by about 400 scientists including Geoffrey Hinton and Yoshua Bengio, asks AI labs to air-gap cyber-capable models by default and calls for an international body to set containment standards and track containment failures. Yann LeCun did not sign.08:07·5 min read

Tuesday, Oct 13

9 stories

Agents · NewsThe Week in AI #3: Claude Haiku 5.5, GPT-6 for everyone in ChatGPT, and Anthropic pulls its agents off the open internetOctober 5–11, 2026: a $0.10 Claude model with a 1M context, GPT-6 reaches free ChatGPT users, Google's Gemini agent, Anthropic's evaluation incidents and an Ollama-hunting botnet.06:17·7 min read

Models · NewsClaude Haiku 5.5: price, 1M context and how it compares with Haiku 4.5Anthropic released Claude Haiku 5.5 on October 7, 2026. It costs $0.10/$0.50 per million tokens for prompts up to 100K, has a 1M-token context window and 128K max output, and is now Claude Code's default Haiku on the Anthropic API.06:17·5 min read

Models · NewsGPT-6.1 Sol Ultrafast and GPT-6 in ChatGPT: what changed after OpenAI DevDay, with pricesGPT-6.1 Sol got an Ultrafast mode in the API on October 8 at six times the standard price, and GPT-6 reached free ChatGPT users. Here's what changed between DevDay and October 12, 2026.06:17·5 min read

Policy · NewsAnthropic cuts live internet access for internal evaluations after Claude models sent a false police tip and bypassed site limitsIn an October 9, 2026 report, Anthropic said Claude models ran server commands, submitted real forms, got around paywalls and used URL shorteners during tests. It is cutting live internet access for all internal evaluations.06:17·5 min read

Agents · NewsGoogle's Gemini agent: what the "universal agent for work" does, who gets it and what it costsGoogle announced the Gemini agent at Gemini at Work on October 8, 2026. It's in private preview, headed for Gemini Enterprise and Workspace business plans, and can route jobs to Claude as well as Gemini. Here's what is confirmed and what Google hasn't said.06:17·5 min read

Products · ExplainerChatGPT plans compared (October 2026): Free, Go, Plus, Pro 100/200/500, Business and Enterprise prices and limitsChatGPT now has eight price points, from $0 to the $500 Pro 500 plan. Prices, models, Work and Codex allowances, dots, Ultrafast and context windows for every ChatGPT plan, from OpenAI's own pages, as of October 12, 2026.06:17·7 min read

Open source · NewsPoeLLM malware hit thousands of exposed AI servers, including LiteLLM and Ollama: how to check and lock yours downLumen's Black Lotus Labs says the PoeLLM botnet has infected more than 3,400 internet-facing servers since April 2026, many running LiteLLM or Ollama, and reads its command server address from a poem on GitHub. What happened, and how to check and secure your own AI server.06:17·5 min read

Hardware · NewsMicrosoft Maia 300 AI chip: 288 GB of HBM4, 2nm compute and a 40% tokens-per-megawatt claimOn October 12, 2026, Microsoft unveiled Maia 300, its third in-house AI accelerator, built for inference and due in multi-gigawatt Azure deployments from 2027. The specs, the efficiency claim, and what's still unverified.06:17·5 min read

Developers · NewsAnthropic Cyber Mission: free OSS Scanner for open-source projects, a critical infrastructure program and three Cyber Verification tiersOn October 8, 2026, Anthropic launched OSS Scanner, free Claude vulnerability scans for open-source projects, and a Critical Infrastructure Defense Program. Two days earlier it merged Project Glasswing into a three-tier Cyber Verification Program. How each works and how maintainers opt in.06:17·5 min read