AI Agents

PolicyGuide compiles organizational policies into workflow graphs and proactively checks agent state at user-turn boundaries. The approach helps customer-service agents recover missing procedural steps instead of merely blocking risky actions.

EnvHarness wraps existing environments with programmable plug-ins instead of rebuilding them from scratch. Its companion system, EnvRigger, studies black-box policy trajectories to create and validate environment changes aimed at specific weaknesses.

Nvidia researchers dramatically improved an agent’s performance by changing the software layer around the model. The result highlights memory, supervision and tooling as core parts of an agentic system.

Meta AI Research has released Muse Glimmer, an Apache 2.0 open-weight model designed for local agent workflows. It combines vision, coding, multi-step tool use, and failure recovery with runtime techniques aimed at consumer hardware.

The next battle in AI video is no longer just about producing an impressive shot. It is about understanding a brief, coordinating assets and delivering many usable versions through one workflow.

OpenConnector is an open-source connector gateway for AI agents and application developers. It aims to bring account authentication, permissions, API adaptation, and execution auditing into a shared runtime.

Slack is introducing Slack Code, a workspace for teams to assign development tasks to AI coding agents without leaving the collaboration platform. Members can review diffs, preview HTML output, and approve work before it ships.

Binance’s new Agent OS connects AI tools such as ChatGPT, Claude Code, and Cursor to its trading, payment, and blockchain infrastructure. Agents can place trades for users, but most of the practical risk controls remain in the users’ hands.

Jefferies analysts evaluated eight major global AI agents across five real-world office tasks, with Alibaba’s Qwen Office taking the top overall position. The assessment highlights execution infrastructure and cost, rather than model capability alone, as key factors in enterprise adoption.

Agentic ESOpt explores evolution strategies as an alternative to conventional reinforcement learning for fine-tuning long-horizon LLM agents. It searches around model parameters with inference-level memory and combines parameter updates with context optimization.

A cross-benchmark study finds that Agent Skills help mainly by turning noisy execution traces into procedural anchors, not by supplying missing facts. Retrieval, however, becomes a major bottleneck as skill libraries grow.

A new study proposes “agentic transactions,” reinterpreting the ACID principles of databases for LLM agents that reason, use tools, generate code, and modify persistent workspaces. Its data-agent implementation reports a 10.6% gain over existing agents on relevant benchmarks.

UI-Mate combines an environment-grounded training pipeline with in-context demonstration learning for long-horizon computer tasks. Its 27B model reaches 77.0% on OSWorld-Verified, according to the supplied paper summary.

StateM improves long-horizon agents by redesigning the execution layer rather than changing model weights. On Terminal-Bench 2.1, the system reaches 95.3% raw accuracy with GPT-5.6 Sol xhigh and transfers across several models.

VibeLifeBench reframes agent evaluation around long-running everyday tasks where the world changes silently and constraints are often implicit.

Ouroboros reframes coding agents as systems that can improve not only outputs, but also their tools, prompts, context assembly, and core implementation. The paper reports leading results on Terminal-Bench 2.1, OSWorld-Verified, and CL-Bench while emphasizing governed evolution.

Cloudflare has introduced Kitesurf, a cloud-hosted browser designed for AI agents rather than human users. The company says it can make common browser automation tasks more resource-efficient than Chromium.

Ant Group has open-sourced Avernet, a multi-agent collaboration infrastructure now available in a community edition. The project focuses on agent discovery, consensus, cross-team collaboration and governance.

Video-DeepResearch pushes multimodal deep-research agents from static images to continuous video streams. It tackles two core failures in current models: skipping visual tools and overrelying on internal memory.

Skill-α reframes agent skill generation as a progressive editing process, where each edit is judged by whether it improves downstream behavior. The method outperforms heuristic and pipeline-style baselines in both document-to-skill and experience-to-skill settings.

LongHorizon-Harness separates task state from execution context and updates it only with environment-verified facts. Its Manage-Execute-Audit loop aims to reduce error propagation in multi-step agent tasks.