Ai

Gemini 4 Argon - Out and Tested!!: Sam Witteveen posted a video testing Gemini 4 Argon, comparing it with other models, reviewing its strengths and generated outputs. The description links to a Google blog post introducing Gemini 4 Argon and an Artificial Analysis page but reports no specific benchmark results.

Artificial Intelligence • Sam Witteveen

Tsinghua researchers release SRHarness for formula discovery, MechBench for mechanism recovery: Tsinghua researchers released SRHarness, an open-source runtime for agentic symbolic regression, and MechBench, a benchmark that tests whether AI can recover the mechanisms behind formulas. SRHarness targets formula discovery, while MechBench evaluates mechanism recovery.

AI Research • Pandaily

Google releases Gemini 4 Argon worldwide, saying Argon cut coding tasks from months to weeks: Google released Gemini 4 Argon worldwide via its API, Gemini app, and Google Antigravity. It is aggressively priced but trails Anthropic's Claude Opus 5.5 and Meta's Muse Wonder Full in independent benchmarks. Google says Argon internally cut coding tasks from months to weeks.

AI Models • The Decoder

Shing-Tung Yau launches 8-institution alliance of 1,000+ scientists to apply AI to math: Fields Medal winner Shing-Tung Yau launched an alliance of eight institutions and over 1,000 scientists to apply AI to mathematics research. The move comes as 25 Western Fields Medal winners warned against AI-generated proofs, with some community members calling for a boycott.

Science • The Decoder

ModelBest open-sources MiniCPM-V 4.7, saying its MoE model runs in about 3GB on an iPhone 17 Pro Max: ModelBest open-sourced MiniCPM-V 4.7, its first MoE multimodal model, in 35B-A3B and 1B sizes. The company says the MoE version runs in about 3GB of memory on an iPhone 17 Pro Max and decodes at over 13 tokens per second.

open source • Pandaily

MagicLab's EchoPolicy-0 tops AI2's MolmoSpaces board with 75.4% manipulation success: MagicLab said its VLM agent framework EchoPolicy-0 averaged 75.4% across four MolmoSpaces tasks, and AI2's published leaderboard data lists it first. Its MagicBrain component reached 94.4% on the pick task.

AI & Robotics • Pandaily

Shopify opens over a billion listings to more AI agents for purchases: Shopify announced Tuesday that shoppers will be able to configure their AI agents to purchase products on their behalf. The expansion gives more AI agents access to over a billion listings in Shopify's catalog.

Agentic Commerce • The Verge AI

Amex introduces framework to prevent unauthorized AI agent purchases: American Express introduced a framework to ensure AI agents only complete purchases their users have authorized. The initiative aims to prevent unauthorized purchases and related disputes as AI-powered shopping grows. Specific availability and enforcement details were not disclosed.

American Express • TechCrunch AI

Cloudflare launches Clef models for text-free AI agent decisions: Cloudflare has introduced Clef and Clef-flash, two Apache 2.0-licensed models built on Qwen that allow AI agents to make structured decisions without generating text. The Clef-flash model delivers classifications in approximately 39 milliseconds, making it over ten times faster than TypeSafe AI's Jev model.

Artificial Intelligence • The Decoder

OpenAI launches Dots agent platform focused on enterprise productivity: OpenAI has announced a new agent platform called Dots. While the agents can perform personal tasks like ordering food, the platform is designed primarily as enterprise software focused on workplace productivity.

Product Release • The Verge AI

Anthropic adds Mods system to let developers customize Claude Code: Anthropic is adding a "Mods" system to Claude Code, serving as middleware that runs directly inside the AI coding tool. Developers can use JavaScript or TypeScript to modify the interface and behavior, including adding custom panels, intercepting tool calls, and creating new commands.

Anthropic • The Decoder

OpenAI fires three researchers for leaking confidential data: OpenAI has parted ways with three researchers who allegedly leaked confidential information to an outside AI safety organization. A fourth researcher has also departed the company's safety team amid the shake-up.

Business • The Decoder

Meta releases Muse Gadgets open-source firmware and Linux SDK: Meta has announced Muse Gadgets, offering open-source ESP32 microchip firmware and a Linux SDK to help users integrate its Muse AI system into hardware like the Raspberry Pi 5. The company also developed its own smart home device called the Muse Home Link.

Hardware • Techmeme

Apple tightens macOS Full Disk Access controls over AI agent risks: Apple announced it will introduce new controls for the Full Disk Access permission on macOS. The company warned that the rise of increasingly capable AI agents makes broad access to user files, messages, mail, and browsing history riskier.

Security • TechCrunch AI

Black Forest Labs launches Flux 3 Image with multi-step editing: Black Forest Labs has launched Flux 3 Image, featuring support for multi-step edits that preserve the rest of the image. The model allows users to compose scenes using bounding boxes and up to ten reference images, with output capabilities up to 4K. Open weights are scheduled for release in the coming weeks.

Artificial Intelligence • The Decoder

Gemini 4 Argon: Google has pre-announced its upcoming frontier-level artificial intelligence model, Gemini 4 Argon. While public outputs are not yet available, the model introduces architectural and performance changes aimed at positioning Google back among the top three AI research labs.

Google • Sam Witteveen

OpenAI Security: Controlling Models is Now ‘Hell’: A compilation of recent artificial intelligence developments highlights growing security concerns, including an OpenAI security warning regarding model control and jailbreaks.

AI Security • AI Explained

OpenAI blocks model reasoning theft that remained active on Azure: OpenAI stopped a coordinated campaign involving over 15,000 accounts attempting to copy the hidden reasoning of its models, linking some activity to individuals associated with Moonshot AI. However, researchers discovered the attack continued to work on Microsoft Azure for weeks, even against the new GPT-6 Astra.

Security • The Decoder

OpenAI launches GPT-6 Astra-powered Dots agent to compete with Meta: At its annual DevDay conference, OpenAI CEO Sam Altman announced Dots, a new AI agent powered by GPT-6 Astra. The launch aims to compete directly with Meta's successful Muse AI agent platform. The move highlights the intensifying competition between the two tech giants in the agent space.

AI Agents • The Verge AI

Google releases Gemini 4 with RSI technology at half the price of Astra: Google has unexpectedly released Gemini 4, which is enhanced by Recursive Self-Improvement (RSI) technology. The new model reportedly outperforms GPT and Claude Opus. Additionally, Gemini 4 is priced at only half the cost of Astra.

AI Models • 量子位

Researchers present Praxa AI agent harness with 100% test pass rate: Researchers have presented Praxa, an agent harness designed to govern AI execution through deterministic admission, brokered execution, and reviewed promotion. In initial testing, the system passed 1,027 out of 1,027 unit tests and 89 out of 89 Workerd tests at a pinned revision.

Infrastructure • arXiv

Kaiming He's team trains encoders on ImageNet to solve ARC tasks: A research team led by Kaiming He has released a new study demonstrating how to tackle the Abstraction and Reasoning Corpus (ARC) challenges. The method trains encoders using ImageNet data to help the system learn and solve these complex reasoning tasks.

AI Research • 量子位