DeepSeek Harness, released by DeepSeek AI, is a modular agent framework where even the agent loop can be replaced via plugins.
A reference architecture separating proposals from approvals has been released to prevent privilege abuse and prompt injection in AI agents.
This analysis examines collaboration patterns in multi-agent systems and the results of software vulnerability detection experiments.
WarpQuant, a 3.62 bpw quantization method combining Hadamard rotation and Output-Fisher recovery to enhance Qwen 27B performance, has been released.
A study was published on porting 250,000 lines of legacy weather simulation code to GPUs using AI agents within a validation-centric workflow.
By leveraging a Mixture-of-Experts (MoE) architecture, a 397B-parameter AI model has been successfully run locally on an iPhone.
Meta has released the Muse Glimmer model and ExecuTorch support for on-device agentic workflows.
Pulitzer Prize-winning journalist Jeff Stein interviewed AI researchers to report on the recent surge in vigilance regarding AI risks.
The latest commit to the MS-Swift GitHub repository confirms support for the Qwen 3.8 35B MoE model.
Z.ai released GLM-5.3 with enhanced cybersecurity capabilities but delayed the open-weight release for two weeks for safety review.
LLMRouter and xRouteBench, featuring over 16 router libraries and benchmarks for LLM cost optimization, have been released.
A tool that manages unnecessary memory by verifying the cost-saving effectiveness of AI agent rules relative to their token consumption.
Stanford researchers successfully designed 16 functional bacteriophages from DNA sequences using language models.
China is securing AI competitiveness through its open model ecosystem and drastic reductions in inference costs.
Large-scale purchases of used books and 'destructive scanning' to secure data for AI training are causing controversy.
Apple has developed its own China-specific LLM with support from Alibaba and received approval from Chinese authorities.
Alibaba's AI model downloads reached 3 billion, surpassing Meta and Google.
GitHub has released Spec Kit, which enforces the creation of structured specifications before AI agents write code.
The Debian project has initiated a vote to determine whether to allow code contributions utilizing LLMs.
An analysis of 94 papers on AI value alignment reveals that the majority do not explicitly define 'value,' relying instead on preference and utility maximization.
A man who attempted to manipulate judicial decisions through hidden prompt injections in litigation documents was sanctioned.
A cookbook for building agents that run on a single GPU has been released for Muse Glimmer.
Introducing BDH-CQ, a system that demonstrates high efficiency on the ARC-AGI-1 benchmark through recurrent latent reasoning.
Aeon, an autonomous AI agent framework based on GitHub Actions, has been released.
AI Briefing