AI Daily · Oct 14, 2026 · Wednesday
BAD INFLUENCE
Google DeepMind releases Gemini 4 Argon, its most capable model yet
Google DeepMind has made Gemini 4 Argon globally available through the Gemini API, AI Studio, and several Google products, at an introductory price of $2 per million input tokens and $10 per million output tokens. The post says Argon ranks first on the Vals Index and scores 87% on the ERQA embodied reasoning benchmark, and it cites a 2.85x inference speedup for Gemma 4 E2B on vLLM.
- Google makes Gemini 4 Argon available to developersGoogle · Gemini app
- Google rolls out Gemini 4 Argon broadly with new cyber and misalignment safeguardsGoogle · Gemini app
- Google releases Gemini 4 Argon in the Gemini app with coding and task updatesGoogle · Gemini app
- Google DeepMind releases Gemini 4 Argon for robotics with 87% ERQA scoreGoogle DeepMind
Model releases & updates
Liquid AI BlogOfficial5 more sourcesLiquid AI releases LFM2.5-24B-A2B, a 24B on-device agent model
Liquid AI releases LFM2.5-24B-A2B, a 24B-parameter reasoning model with 2B active parameters built for on-device agents. The company says it beats Qwen3.6-35B-A3B on five of eight agentic and tool-use benchmarks, scoring 48.09 on LiveClaw versus 38.19 and 27.69 on AutomationBench versus 8.72. It reports 184 tokens/s on an Apple M5 Max and up to 21K output tokens/s on a single NVIDIA B300 GPU, with open weights on Hugging Face.
Google Antigravity@antigravityOfficial1 more sourceGoogle releases Gemini 4 Argon, its frontier model for software engineering workflows
Google announces Gemini 4 Argon, a frontier model built for complex real-world software engineering workflows. In Google's own codebases, it reportedly raised agentic code output by 35% and cut post-submit rollbacks by 30%. It is rolling out now to Google AI Ultra subscribers and is available in the Antigravity app and Antigravity CLI.
Product launches & updates
Vercel Developers@vercel_devOfficial2 more sourcesVercel AI Gateway adds Google's Gemini 4 Argon, offering 50% off through December 1
Vercel says Gemini 4 Argon, which it calls the first Gemini 4 model, is now live on AI Gateway. The post offers 50% off through December 1 and recommends the model for difficult coding and agent tasks.
Google Cloud Tech@GoogleCloudTechOfficialDeveloper guide to Gemini 4 Argon Long Decode
Google Cloud Tech explains how Long Decode lets Gemini 4 Argon return completions up to 1M tokens by splitting one generation across several model inferences. Developers can enable it through the latest Gen AI SDKs or by manually handling a continuation_token, with max_output_tokens set above 65,536 for streaming or 32,768 for unary requests. The post also notes that billing for completions above 32K tokens starts November 13, 2026.
Apple launches HomeView, a $299 Siri screen that recognizes family members
Apple has launched HomeView, a 7-inch Siri screen priced at $299 in the US, with preorders open October 13 and sales from November 13. The screen recognizes up to six family members through facial recognition, but personal content appears only when that person's paired iPhone or iPad is on the same Wi-Fi and recently unlocked. In mainland China, only the HomePod mini 2 is available for preorder, at 1099 yuan, and HomeView has no announced China plan.
ZDNet · AINewsLinus Torvalds uses Claude Code to write Linux kernel patches
Linus Torvalds says he wrote a set of Linux kernel patches largely with Anthropic's Claude Code, under his direction. The patches address a page cache contention problem that slows multi-threaded small reads. In one test on his laptop, eight-thread throughput rose from about 10 million to 30–33 million reads per second, though the branch is not yet merged into mainline and memory reclamation issues remain unresolved.
Industry news
Greg Brockman@gdbX3 more sourcesXBOW says GPT-5.6 Sol helped build a $250,000 Chrome exploit in 6 days
Greg Brockman says XBOW, using GPT-5.6 Sol, built a full-chain Chrome exploit in 6 days for about $40,000 of compute. Google awarded a $250,000 bounty, the first under its 2026 rules, for the exploit, which the post says could affect about 1 billion Chrome users on Windows. Brockman says defenders should test their own code first, and XBOW is developing an autonomous product for testing compiled software.
The DecoderNews6 more sourcesCharacter.ai files for bankruptcy as Disney buys assets for $20 million
Character.ai has filed for bankruptcy and is up for sale, with the deal expected to close by early December 2026. Disney is picking up some assets for $20 million, including a perpetual, non-exclusive technology license. If no buyer emerges, the service will shut down, while the platform and subscriptions keep running until then.
OpenAI Newsroom@OpenAINewsroomOfficial5 more sourcesOpenAI finds hidden recording device in office, employee arrested by FBI
OpenAI says it discovered a sophisticated recording device hidden in one of its offices and used its models to identify that a specific employee had planted it days earlier. The company referred the matter to the FBI, which made an arrest, and says it has no indication the employee accessed model weights, source code, or customer data. The employee was terminated.
Tips & takes
Anthropic@AnthropicAIOfficial5 more sourcesAnthropic builds in-house source control after merge volume outgrows git
Anthropic says its merge volume rose from about 2,500 code changes per day in May to more than 30,000 per day in September, which outpaced developer infrastructure built for human-scale work. The company says it built a new version control system and fully migrated its monorepo from git in five months.
Google Developers BlogOfficial1 more sourceGoogle uses Gemini 4 Argon to optimize ADK's Python codebase overnight
Google tested Gemini 4 Argon on google/adk-python, a 205,000-line codebase, in an 11-hour unattended run guided by .agents/skills rules and /goal mode. Two of six proposed optimizations shipped, cutting in-memory session preparation time by about 40% in long sessions and speeding up package imports by 21%. Four proposals were rejected in code review for fragile state or private library dependencies.