Agents

AI agents are LLM-driven systems that pursue a goal over multiple steps — planning, calling tools, observing the results, and deciding what to do next — instead of producing a single response. The model runs inside a loop rather than answering once.

What separates an agent from a chatbot is autonomy over control flow: it chooses which Tool Use action to take and when the task is done. This unlocks work like research, multi-step coding, and workflow automation that no single prompt can complete.

For builders the hard problem is reliability, not capability: small per-step error rates compound over a long loop, so production agents lean on Context Engineering, Agent Memory, guardrails, and verification far more than on a smarter base model.

Related: Agent Memory Tool Use Context Engineering Code Agents

This is the Agents concept page (definitions and fundamentals). For dated developments, see the Agents Topic hub.

Recent Updates

• 2026-07-24: Telco-GAIA benchmark tests bilingual telecom agents across text, SQL, and web archives (cs.AI updates on arXiv.org) · arxiv.org
• 2026-07-24: LegalCiteTrust benchmarks citation trustworthiness in Chinese legal research reports (cs.CL updates on arXiv.org) · arxiv.org — arXiv
• 2026-07-24: LLMs Struggle When User Intent Changes Mid-Conversation (cs.LG updates on arXiv.org) · arxiv.org
• 2026-07-24: The Trust Problem Behind Vibe Coding (Hacker News) · alexklos.ca — Andrej Karpathy Grady Booch Grace Hopper
• 2026-07-24: Maintainer says agent PR volume is rising, but review is the real bottleneck (Hacker News) · nesbitt.io — Tyler Erin Marsh OpenClaw-4.2
• 2026-07-24: Anthropic introduces Claude Opus 5 (ElKornacio) · anthropic.com — Anthropic · Zapier · Claude Opus 5 · Claude Fable 5 · Mythos 5 · Opus 4.8 Claude Max Claude Pro
• 2026-07-24: Claude Opus 5 launches with stronger coding and knowledge-work benchmark results (Hacker News) · anthropic.com — Claude Opus 5 · Claude Fable 5 · Mythos 5 · Opus 4.8
• 2026-07-24: Opus 5 tops the Artificial Analysis Intelligence Leaderboard (Hacker News) · artificialanalysis.ai — Artificial Analysis · Claude Opus 5 · Claude Fable 5 · GPT-5.6 Sol Mercury 2 HyperNova 60B 2605 Granite 4.0 H Small Gemma 3n E4B Instruct Nova Micro Sarvam 30B · Gemini 2.5 Flash Lite · Command A+ · Gemini 2.5 Flash · GLM-5.2 · MiniMax M3 · DeepSeek V4 Pro
• 2026-07-25: A catalog of 3,607 reported AI agent misbehavior incidents (Hacker News) · rewardhacking.org — GitHub · Hacker News LessWrong · X
• 2026-07-25: Vivix launches A1, a real-time interactive multimodal model with a unified streaming architecture (量子位) · qbitai.com — Vivix 量子位 · QbitAI henry A1 W1
• 2026-07-25: ARC-AGI-3 leaderboard compares performance against task cost (Hacker News) · arcprize.org — Kaggle ARC Prize ARC-AGI-1 ARC-AGI-2 ARC-AGI-3 GPT-4.5 Claude 3.7 o1 pro · Gemini 3 Pro
• 2026-07-25: A debate on whether agents are already useful or still just getting started (量子位) · qbitai.com — QbitAI · Alibaba Kujing Technology Miaopai CCF · Tsinghua University Nanjing University Shanghai Jiao Tong University Tianjin University Shandong University Shanghai AI Laboratory Jin Lei Gao Yang Hao Jianyie Han Zhongyi Wen Ying Zhou Hao Zhang Hangfan Du Yanlong
• 2026-07-25: Andrew Ng’s OpenWorker is an open-source desktop agent with local-first controls (量子位) · qbitai.com — GitHub · Slack · Jira · Notion · Linear · Hubspot Outlook · Gmail Google Calendar · OpenAI · DeepSeek · Ollama · Google · Baidu DeepLearning.AI Coursera · Stanford University Google Brain Andrew Ng Daphne Koller 文婷 · GPT-5.6 Sol · Claude Fable Gemini 3.6 · Kimi · GLM · Inkling
• 2026-07-25: Ecovacs open-sources its embodied robot platform for developers and researchers (量子位) · qbitai.com — Ecovacs QbitAI Atomgit · GitHub Qian Dongqi Kreisi
• 2026-07-26: World-model-optimizer: Continuously optimize agents and reduce inference costs through model routing and distillation (Hacker News) · github.com — experientiallabs OpenRouter · e2b
• 2026-07-27: Wattage: A token-spend profiler and cost-regression gate for AI agents (Hacker News) · github.com — faizannraza
• 2026-07-27: Port Zero: Stable Virtual Domains for Port-Conflict-Free Local Development (Hacker News) · portzero.net — Port Zero Network
• 2026-07-27: Trajectory-Aware Retrieval Agents for Temporal Decision-Making (cs.AI updates on arXiv.org) · arxiv.org
• 2026-07-27: Learning on the Job: Continual Learning from Deployment Feedback for Frozen-Weights Agents (cs.AI updates on arXiv.org) · arxiv.org — Anthropic · Mistral · Claude Sonnet 5 Mistral Large
• 2026-07-27: The Regression Tax: Decomposing Why Skills Help and Hurt LLM Agents (cs.AI updates on arXiv.org) · arxiv.org
• 2026-07-27: Kimi-K3: Open-Source Frontier Model with Native Agentic and Coding Capabilities (Hacker News) · huggingface.co — Moonshot AI · Hugging Face · Kimi K3
• 2026-07-27: MirrorCode Benchmark: AI Systems Complete Complex Programming Tasks in Hours Instead of Weeks (Import AI) · importai.substack.com — Epoch METR · Anthropic · Apple · Claude Opus 4.7 · GPT 5.5
• 2026-07-27: OpenResearch: Autonomous Research Agents with Parallel Experimentation (‌alphaXiv) · github.com — alphaXiv · Hugging Face · Modal
• 2026-07-27: Industry shifts AI guidance toward agentic systems over chat models (Simon Willison’s Weblog) · simonwillison.net — OpenAI · Anthropic · Google Ethan Mollick Simon Willison · o3 Claude 4 Opus · Gemini 2.5 Pro Gemini Spark
• 2026-07-28: Rebuilding legacy sites using AI without reading the code (Все статьи подряд / Искусственный интеллект / Хабр) · habr.com

FAQ

What is Agents?

AI agents are LLM-driven systems that can plan, call tools, inspect results, and iterate toward a goal. GROUNDING tracks agent architectures, benchmarks, tool use, browser agents, coding agents, and reliability patterns.

Which concepts are related to Agents?

Related concepts tracked by the radar include Agent Memory, Tool Use, Context Engineering, Code Agents.

What does this Agents page track?

Dated updates, papers, and mentions of Agents collected by the GROUNDING radar, most recently on 2026-07-28.