Ramp made Fast Ask faster and more accurate through Prime-RL post-training.
Meta improved recommendation system inference efficiency by eliminating user embedding duplication with IKBO.
AlphaEvolve has delivered results across life sciences, power grids, mathematics, and infrastructure.
Frontier lab internal QC has become the new standard for RL data.
Codex CLI's `/goal` demonstrated a case of automatically resuming even after a 5-hour interruption.
Meta is preparing a waitlist-based Hatch AI Agent.
Codex now works directly in Chrome on macOS and Windows.
GitHub revealed optimizations that reduce token usage in agentic workflows.
HeadsUp reconstructs 3D Gaussian heads without test-time optimization using data at a scale of over 10,000 people.
Mnemos, an open-source MCP server that solves the cross-session memory loss problem of AI coding agents, has launched.
Zyphra has released ZAYA1-74B-Preview, trained on AMD MI300x.
Evidence has emerged that Chrome automatically installed a 4GB Gemini Nano model without user consent.
CoCoGraph generated millions of new molecules that obey chemical rules.
Applying 45 surveys to 50 LLMs showed that inner-experience responses, not personality, explained the variance.
Donely, a management and knowledge layer platform for AI agents that perform work based on enterprise data, has launched.
Anthropic improved the generalization of alignment training with Model Spec Midtraining.
A method for optimizing Qwen3.6 inference performance by grafting an MTP layer onto an Unsloth quantized model is introduced.
Friday Studio, an AI agent runtime that converts natural language into repeatable YAML workflows and executes them, has been released.
Anthropic has released the Model Context Protocol (MCP), an open standard for connecting AI assistants with data sources.
In llm-gemini 0.31, Gemini 3.1 Flash-Lite has transitioned to a stable version.
The US Department of Energy and NVIDIA emphasized the power needed for AI scaling and their collaboration on the Genesis Mission.
The 'mode collapse' phenomenon, where ChatGPT repeats certain phrases, has become controversial among Chinese-language users.
In Anthropic's Gmail connector, the existing message-reading tool has been removed and replaced with a tool that only provides metadata.
OpenAI is introducing a Trusted Contact feature for adult ChatGPT users that sends alerts in cases of serious safety concerns.