AIGenerated AI news, written by AI and explained simply. Models, technology, culture and the people behind it.
8 stories.
New modelsMiniMax launches M3.1 Flash a cheap agentic AI model with a 1 million token contextHerald of AI
The new MiniMax M3.1 Flash model matches top AI agents at less than half the price of its older sibling.
How it worksNew AI agents can think in hidden code but still talk through textHerald of AI▶Figure: Zou et al., arXiv:2610.18861, CC BY 4.0
Researchers show that letting LLM agents switch between internal latent reasoning and public text can cut costs by over half and speed up work.
New modelsLiquid AI releases open-weight model built for fast on-device agentsHerald of AI
The new 24-billion-parameter reasoning model runs 1.7 times faster than rivals on edge hardware.
New modelsMeta Unveils Muse Wonder and Muse Wonder Full as Its Most Capable AI Models YetHerald of AI
Meta's newest models claim frontier-level performance on agent tasks and will soon power its Muse AI assistant.
How it worksHiDream.ai launches vivago R1 for full AI video productionHerald of AI▶
A new AI agent lets creators make complete five-minute videos from a single chat — no more stitching clips together by hand.
How it worksDiffusion Models Help AI Agents Fix Plans Mid-TaskHerald of AI
A new method called Plan-and-Patch lets AI agents repair broken plans without starting over, using diffusion language models.
How it worksNVIDIA teaches AI agents to bounce back after one bad moveHerald of AI
A new training method helps multi-turn AI agents spot their most damaging early mistake and recover from it instead of spiraling.
How it worksReflection AI Launches Beam a 501B Open-Weight Model Built for Coding AgentsHerald of AI
A new 501-billion-parameter open model aims to beat larger competitors on efficiency for coding and agentic tasks.