MegaTrain Trains 100B+ Parameter Models on Single GPU
MegaTrain enables full precision training of large language models on a single GPU. It stores parameters in host memory and uses GPUs as compute engines.
All AI stories, newest first.
MegaTrain enables full precision training of large language models on a single GPU. It stores parameters in host memory and uses GPUs as compute engines.
MegaTrain enables full precision training of large language models on a single GPU. It achieves this by storing parameters in host memory and using GPUs as compute engines.
A new mathematical theory models the evolution of self-designing AIs. This theory differs from biological evolution, with AI design being strongly directed rather than random.
A new mathematical theory models the evolution of self-designing AIs. This theory differs from biological evolution due to directed descendant design.
Researchers propose a mathematical theory to model the evolution of self-designing AIs. This theory aims to understand how AI systems shape their descendants through recursive self-improvement.
Lyria 3 is available for paid preview and testing. It generates music through the Gemini API and Google AI Studio.

Researchers found that surface heuristics can override LLM reasoning constraints. This discovery has significant implications for AI development and trust.
An API catches errors in LLM responses. It helps identify confidently incorrect answers.
Japan has eased its privacy laws to become a more attractive location for AI development. The move aims to make Japan the easiest country to develop AI in.
A supply-chain attack using invisible Unicode code has affected GitHub and other repositories. The attack exploits code that is invisible to the human eye.
Holotron-12B is a new high throughput computer use agent. It is designed for efficient computing.
Google has bumped up its Q Day deadline to 2029, warning the industry to move off RSA and EC quickly. This deadline is sooner than previously thought.
Gemma 4 is a new multimodal intelligence model. It runs on devices, enabling local AI processing.
Google introduces Flex and Priority inference tiers to the Gemini API for balanced cost and latency. This move aims to provide more options for developers.
Gemini in Google Sheets has reached state-of-the-art performance with new beta features. These features enable users to create, organize, and edit sheets using descriptive tasks.
Gemini 3.1 Flash Live is now available, improving audio AI. This update enhances natural and reliable audio processing across Google products.
OpenAI outlines the next phase of enterprise AI with accelerated adoption. New tools like Frontier and ChatGPT Enterprise are leading the charge.
DesigNet learns to draw vector graphics like designers. It creates designs similar to those made by humans.
Codex has surpassed Claude Code as the top AI coding tool. This shift marks a significant change in the AI coding landscape.
A process manager for autonomous AI agents has been introduced. It aims to efficiently manage AI agent operations.