AI Briefing#llm
The latest AI and developer news about #llm, with the original source and a short summary.
The latest AI and developer news about #llm, with the original source and a short summary.
The 2026 State of AI survey found that the share of AI-generated code surged to 56%, with coding agents emerging as a core tool.
The Qwen team has unveiled **Qwen3.7-Max**, a model for agents optimized for coding, office automation, and long-horizon autonomous execution.
A new framework is proposed that ensures AI safety not through model training but through software architecture.
An AI agent that lets users interactively query Datasette data and generate charts has been released.
It solves the single-stream bottleneck of LLMs by separating thoughts, inputs, and outputs.
LatitudeGames has unveiled Equinox-31B, a storytelling-focused model based on Gemma 31B.
Anthropic's Claude Sonnet 4.5 model will become unavailable in chat starting May 26, being replaced by Sonnet 4.6.
Google is building a new control layer over the web by delivering AI-generated answers instead of search result links
The OpenClaw AI agent successfully controlled a physical robotic arm via an LLM without any separate coding.
MDLM improved agentic RL performance as a stronger text-based world model than AR.
SuperGrok and X Premium subscribers can now use the Grok Build model in the open-source coding agent OpenCode.
New research shows that with sufficient compute power, training large models on low-quality data without filtering is possible.
Cohere has released Command A+, an open-weight model based on Apache 2.0.
PopuLoRA co-evolved a verifiable self-play curriculum with a teacher-student LoRA population.
News has emerged that the Qwen team is likely to release a new model at the 27B scale.
Cohere has signed separate MOUs with Indra Group and Multiverse Computing to advance the expansion of sovereign AI.
An update has been proposed in llama.cpp to improve MTP performance by moving the draft path to backend sampling.
ByteShape released Qwen 3.6 35B GGUF NTP and MTP quantizations along with benchmarks.
Cohere has open-sourced Command A+, a 218B-parameter MoE model, under the Apache 2.0 license.
Shares the process of building an AI system that efficiently delivers domain knowledge by combining LLM Wiki and RAG.
A feature has been added to Hugging Face's benchmark dataset that allows filtering results by model size.
Google unveiled Gemini 3.5 Flash, putting an agent-centric strategy front and center.
Google and major AI chatbots have become vulnerable to manipulation, prompting Google to revise its spam policy.
New research shows guardrail techniques can raise an 8B-scale model's agentic task success rate from 53% to 99%.