Weekly AI news for engineers, executives, and enthusiasts.

PrismML fits 27B model on iPhone: Inside OpenAI’s first AI-powered consumer device

Claude models treat you differently by language. Cognition’s SWE-1.7 boosts Devin’s coding power. Nvidia’s Audex handles audio and text with one transformer. OpenAI updates its prompting guide for GPT-5.6.

Apple sues former partner OpenAI: Meta unveils cost-efficient Muse Spark 1.1

ChatGPT Work wants to handle your work outside the work. IBM’s CodeAlchemy, an enormous synthetic dataset. OpenAI shuts down its Atlas browser. Pangram says AI-aided text is everywhere on social media.

Text Without Typing: Researchers at Meta and other institutions built Brain2Qwerty v2 to generate sentences from brain waves

Imagine you are typing a sentence. But instead of a keyboard, joystick, or eye tracker, a device surrounding your head reads your intentions and generates that sentence on screen.

DeepSeek’s DSpark Gains Velocity: DeepSeek open sourced a speculative decoding module that speeds up text generation without losing accuracy

DeepSeek built a speculative decoding module that speeds up its production models’ text generation by more than 50 percent without sacrificing accuracy, then made its technique open source.

Google Pairs Nano Banana Update With Video API: Gemini's image (Nano Banana 2 Lite) and video (Gemini Omni Flash) models, built to work together

Google built a low-cost, high-throughput pipeline for developers working with media, combining Gemini’s fastest image model yet with a similarly speedy multimodal model that can turn images into video with synchronized sound.

Fable’s Return and Fallout: How Anthropic's Claude Fable 5 got banned by the U.S. Government, then came back to the market

Claude Fable 5 and the more powerful Claude Mythos 5 are back, three weeks after Anthropic suspended the models due to an export control directive from the U.S. Department of Commerce.

Make All Your Tokens (and Your Brainwork) Count: Agentic loop coding works until it fulfills a goal. Here's an agentic way to iterate on your goals to find the right one.

The key idea of agentic coding loops is to have an agent keep working until it satisfies a condition, such as achieving a product specification.

Restoration of Claude Fable 5, Gemini's Video Dev Engine, DeepSeek Speeds Up Speculative Decoding

The Batch News & Insights: The key idea of agentic coding loops is to have an agent keep working until it satisfies a condition, such as achieving a product specification.

GPT-5.6 will see a wider release this week: China’s government moves to control its companies’ AI products

Copilot will decrease reliance on GPT, Claude. Claude Cowork gets web, mobile implementations. Nvidia’s Audex blends audio and voice with text intelligence. OpenAI mini voice model promises to keep costs low.

Alibaba and ByteDance quash human-like bots: And, in case you missed it: some recent under-the-radar news

Seedance 2.5, the movie industry’s favorite video generator. Field-specific knowledge helps Claude code users most. DiffusionGemma, Google’s fast, capable text generator. Regular use of AI chatbots rises among Americans.

Better Reward Models for Robots: Inside RoboReward, a family of vision-language reward models that train robots to take action

When you’re training a robot via reinforcement learning, a handcrafted reward function is labor-intensive to build but often dispenses rewards more effectively than a general-purpose reward model based on a vision-language model. Researchers built reward models that narrowed the gap.

Microsoft Strikes Out on Its Own: Microsoft revealed MAI-Thinking-1, a Claude Sonnet 4.6-sized reasoning model developed without distillation

Microsoft, once OpenAI’s exclusive partner and still a major reseller of other companies’ AI models, built its own reasoning model from scratch.