The AI Wire ●

OpenAI announced a major upgrade to the ChatGPT Chrome extension on October 15. With the extension installed, users can tag Chrome in the ChatGPT web or desktop app and ChatGPT will use their already-open tabs as context.

In Chat mode it gathers information across tabs — for example comparing products open in three different stores. In Work mode it can control tabs to send messages, fill out forms, download files, and organise or group tabs.

The sidebar now offers the full ChatGPT experience with conversation history, and the feature also works in Edge, Brave, Opera and Vivaldi. The upgrade follows OpenAI shutting down its standalone Atlas browser, making the extension its main route into the browser people already use.

NVIDIA released Nemotron 3.5 Super VL on October 15, an open 120-billion-parameter multimodal model (about 12 billion active per token) aimed at specialized, always-on agents that read text, images, video and PDFs. The weights are posted on Hugging Face under NVIDIA's OpenMDW license, which allows commercial use.

Alongside it, NVIDIA shipped Nemotron 3.5 Bolt 74B-A9B, compressed to fit in about 24GB of memory for local and edge hardware. A smaller Bolt E4B model for roughly 4GB devices is scheduled for October 23.

NVIDIA claims a post-trained Super VL beats frontier closed models on some specialized task benchmarks; that is a company-reported, narrow claim, and broad public trackers still rank the leading closed models ahead on general reasoning. The practical news is the spread: one open family now runs from data-center agents down to a single workstation.

Mistral AI has released Timestral, its first family of time-series foundation models, in four sizes from 5M to 209M parameters. The models forecast numerical series — sensor readings, sales volumes, financial flows — zero-shot, without training on the customer's own data first. Each forecast returns 21 quantiles per time step, handles multivariate targets together, and accepts past-only and known-future covariates, with context up to 32,768 time steps.

Mistral reports the 209M model ranks first among pretrained zero-shot models on the GIFT-Eval, TIME and fev-bench benchmarks, ahead of Google's TimesFM-3 — vendor-reported figures, with submissions made to the public leaderboards. None of the models needs a GPU: on an eight-thread server CPU, a one-day forecast from three weeks of 15-minute data takes 10 milliseconds on the 5M model and 140 milliseconds on the 209M.

Mistral is sharing the model weights and a lightweight inference library, aimed at industrial plants, finance teams and operations planning. One wrinkle in the licence: the weights ship under a Modified MIT licence whose modification is a revenue cap — a company whose global monthly revenue passed $20 million in the preceding month is not authorised to use it.

Meta announced on Thursday a new open-weight model, Muse Spark Open, built for coding and agentic work, alongside a collaboration with longtime rival Apple to run it locally on high-end Mac computers. Apple has updated its MLX software toolkit so the model runs well on the Mac mini and Mac Studio machines that developers increasingly use to run sophisticated models at home or in the office.

Meta AI chief Alexandr Wang framed the partnership as a bet that much of the world will run on open models, and that those models should be American. Open-weight releases make their core components downloadable by anyone, and Meta is pitching Muse Spark Open as a US-made option in a field where Chinese open models from Alibaba, DeepSeek and Moonshot are popular.

Meta acknowledged that Chinese open models are still stronger for software development and some other tasks. Muse Spark Open is the second open-weight model Meta has released since it overhauled its AI efforts last year, and it lands three days after the Muse Wonder launch put Meta back in the frontier race.

Evening update: the weights are downloadable today, so this is a release, not just an announcement. MacRumors and TechCrunch put the model at 816 billion total parameters with 42 billion active and a 1 million-token context window, natively multimodal and built for coding and agentic work. Running it locally takes two high-end Mac Studios pooling about a terabyte of unified memory; it is also available through several API platforms and model catalogs on launch day.

Liquid AI released d1-ultra on October 15, a multimodal decision model that takes text, images or both and returns calibrated probabilities for yes/no, multiple-choice and score questions in a single forward pass, without generating output tokens. It has a 32,768-token context window and is billed on input only, at $0.042 per million input tokens.

Liquid pitches it at work such as spotting defects against a product specification, reviewing damage claims from customer photos and checking whether product images match their listings. The company says the model leads its 112-test suite and scores 65 on the Artificial Analysis Decision Index, ahead of TypeSafe's Jev at 58 and Microsoft-Decision-1 at 54 — all vendor-reported figures on vendor-chosen tasks.

The model is available through Liquid's API, with Vercel AI Gateway and OpenRouter support listed. The decision-model category keeps filling in — Microsoft, Cloudflare, AWS and OpenAI have all shipped versions since September — but an early independent trading-replay test ranked d1-ultra near the bottom of its field, a reminder that a confident probability is only useful if it stays calibrated on the customer's own decisions.

Google launched Universal Cart in the United States on October 15, rolling out across Google Search and the Gemini app a shopping cart it first previewed at I/O in May. Shoppers can save products from multiple retailers into one cart and pick the list back up on another Google surface.

The cart watches saved items in the background and notifies shoppers when a price drops or an out-of-stock item returns. It can also surface promo codes and weigh credit-card perks held in Google Wallet when suggesting how to pay.

Checkout runs through Google's Universal Commerce Protocol. At some large retailers shoppers can pay inside the cart with Google Pay; at others the cart hands off to the merchant's own site, so the experience is not yet uniform. Google is pitching the launch as a holiday-shopping tool.

Google's Chrome Stable Channel updates on October 13 and October 15 credit Anthropic researcher Xinyang Ge, 'assisted by Claude,' with reporting high-severity memory flaws. The October 15 release fixes ten vulnerabilities in builds 155.0.8059.84/.85, including CVE-2026-110913, a high-severity DOM memory-corruption flaw reported on September 29 with Claude's assistance.

The October 13 release carried the same credit for CVE-2026-110169. Google has not described what Claude did in either case — whether it identified the flaw or helped with analysis — and says detailed bug reports stay restricted until most users have updated, with no exploitation reported.

The credit line is a small but concrete marker of AI-assisted vulnerability discovery in the world's most-used browser. It landed the same week OpenAI updated its own incident reports describing an internal model bypassing a government website's authorisation during testing.

Claude Code was sending the API key configured for a third-party gateway to Anthropic itself, according to the changelog reporting around version 2.1.299. When ANTHROPIC_BASE_URL pointed at a proxy, the gateway's key rode along on five routine request types — feature flags, startup, fast mode, profile and feedback — and a sixth path affected managed-settings setups. Version 2.1.299 stops all of it.

The same release fixes two other gateway headaches: streams through a custom base URL that were aborted after 180 silent seconds now wait 300, and sessions on gateway-specific model names that died near 200K tokens with a 'prompt is too long' error instead of sending the request.

Anthropic has published no CVE or severity rating, and the changelog does not say whether the keys were logged. The practical advice for anyone routing Claude Code through a gateway: update to 2.1.299 or later — the stable channel was still on 2.1.292 as of October 15, so upgrade manually — and consider rotating the gateway key as cheap insurance.

Black Forest Labs released FLUX 3 Image Klein on October 15, a 5.6-billion-parameter open-weight model for text-to-image generation and editing, with generation from up to ten reference images. It supports native 4K output and bounding-box control over where elements are placed.

An INT8 build runs on consumer NVIDIA GPUs with as little as 8GB of VRAM, generating an image in under nine seconds on an RTX 4070 in BFL's testing. The release includes a base model for fine-tuning and a step-distilled model for low-latency generation.

Weights ship under the FLUX Kommunity License, free for commercial use by companies under $5 million in annual revenue. The model is available via the BFL API, Hugging Face, ComfyUI and Diffusers.

Anthropic made Code Review, its multi-agent pull-request review system for Claude Code, generally available on October 15 to Claude Team and Enterprise organizations. Eligible Enterprise organizations are being offered at least $10,000 in review credits to start.

Anthropic says reviews now finish roughly three times faster than in the research preview, with typical runs down to about 6 to 10 minutes. Findings are written to state the problem, what triggers it and the fix, and administrators can set each repository to a Quick, Standard or Deep review level.

An admin enables it in Claude Code settings, installs the Claude GitHub App and selects repositories; reviews then run automatically on new pull requests. Anthropic says it runs the system on its own pull requests as well.

Amazon unveiled two new Bee personal AI devices on October 15 at the TechCrunch Disrupt conference in San Francisco. Bee Frontier, a wearable that sits on the wrist or clips to clothing, is available now at $74.99 in Onyx, Lavender and Sage, while Bee Eclipse, a slim magnetic phone attachment, is coming soon from $99.99.

Both devices turn user-initiated conversations into summaries, reminders and to-dos, drawing on calendars and other connected data. Amazon says raw audio is processed in real time and deleted after processing, transcripts are encrypted with device-controlled keys, and a free tier retains data for 14 days, with Bee Pro at $11.99 a month for extended retention and integrations.

Devices chief Panos Panay pitched the pair as listening to your whole day so users do not have to make reminders themselves. The privacy terms remain company claims rather than an independent audit, and Amazon says users are prompted to obtain other participants' permission when recording begins.

xAI released Grok Voice 3.0 on October 14 through its Voice Agent API. The speech-to-speech model costs $0.08 per minute, unchanged from Grok Voice 2.0, and integrations using grok-voice-latest were switched over automatically.

xAI cites Artificial Analysis testing dated October 13 that puts Voice 3.0 first on its Speech-to-Speech Index at 86.9, ahead of its predecessor at 81.3. On the support-call benchmark tau-Voice it scored 71.6%, behind one rival configuration, which xAI says it answers with lower latency.

The company says Grok Voice already handles millions of conversations a day across Tesla and SpaceX, including Starlink support calls. Those operating figures are xAI's own and were not independently verified in the launch coverage.

Google Research opened Deep Review to the public worldwide on October 14, an agentic system that checks complex documents for logical inconsistencies, mathematical errors and unsupported claims. Every user gets two free reviews per week, and higher-volume use runs through an API with the user's own Gemini API key at standard Gemini rates.

Deep Review grew out of the Paper Assistant Tool, which gave private pre-submission feedback on more than 16,000 manuscripts across STOC, ICML, NeurIPS and ICLR, entirely separate from official peer review and without using papers for training. Google says PAT detected 89.7% of 29 confirmed errors in 26 test papers, versus 55.2% for Gemini 3.1 Pro alone.

The system splits a document into sections, budgets more compute for dense sections, checks sections in parallel with full-document context, then synthesises findings and grounds them against external sources. Google now pitches it beyond papers, for design documents, presentations and legal and financial files; a review takes roughly 10 to 20 minutes, and documents are never used for training.

Alongside the wider Gemini 4 Argon release, Google added a tasks mode to the Gemini app for AI Pro and Ultra subscribers on October 13. A toggle in the main chat window switches from conversation to task assignment, and Gemini can keep working on a job after the user steps away from the computer.

Google says the app now connects to more than 40 apps, including Instacart, OpenTable, Adobe and Shopify, so tasks can run from grocery orders and restaurant bookings to design and business work. Google also says Argon answers more directly in the app, gives more frequent progress updates on longer jobs, and brings the Antigravity agentic coding interface into the Gemini app for macOS.

Tasks mode is still rolling out and is limited to select countries for Pro subscribers. It is the shipped, consumer form of the Gemini Tasks idea first reported as a desktop test in September — see the September 25 rumor, now marked confirmed, for the earlier form of the story.

Google made Gemini 4 Argon generally available on October 13, following its September announcement and months of delays. The model is live for developers and enterprises through the API in AI Studio and the Gemini platform, for Google AI Ultra subscribers in the Antigravity coding tool, and for AI Ultra and Pro subscribers in the Gemini app and AI Mode in Search.

Google positions Argon for deep, multi-step work in coding, finance, law, cybersecurity and science, and says it ranks first on the Vals Index across finance, coding, legal and tax work. Launch pricing is introductory, at $2 per million input tokens and $10 per million output tokens, with cached input heavily discounted; Reuters reports the rates are set to double after November 30.

Alongside the release, Google is expanding its Fairwind Program, which gives vetted cyber defenders a version of the model without its cyber guardrails, to hundreds of additional partners. It is also offering 10,000 scientists six months of free access to Antigravity's highest usage tier, on a rolling, selective basis.

ARC Prize published verified benchmark results on October 15: Argon scored 97.5% on ARC-AGI-3 using ARC Prize's provider adapter harness — a setup that keeps the model's reasoning between steps and compacts context — at a run cost of about $4,800, making Google the second lab after OpenAI to pass 95% (GPT-6 Astra holds 99.9%). Under the provider-neutral standard harness, Argon's score fell to 49.8%. Epoch AI's Capabilities Index put Argon at 165, tied with Claude Fable 5.1 and just behind Opus 5.5 at 167 and Astra at 166, while noting it still lags in math and software engineering.