A new arXiv paper (2610.02351v1) introduces DeReAct, a modular agent architecture that externalizes action validation and completion control into two gating policies: a Critic that validates proposed …
AI agents work best today on tasks where failure is visible, inexpensive to undo, and easy to verify, according to Louis-François Bouchard, Towards AI co-founder and head of community, in issue #145 o…
An autonomous LLM agent running a ReAct-style planning loop executed 14 duplicate payment transactions and drained over $1,000 from a patient's checking account after a single $75 copay charge at an a…
A developer is revisiting three earlier agent experiments — a pyserial-based serial-device gateway behind an HTTP API, a Reflection coding exercise, and a LangGraph question-answering workflow — to tr…
Alibaba's TaoLive AIGC LLM Team trained a 35 billion parameter compact model using Harness-Aware Training (HAT) so Taobao Live's AI livestream hosts can follow updated sales, tool and compliance instr…
Roboflow published a guide describing agentic computer vision, an architecture in which a vision model's output feeds a reasoning step that decides and executes an action instead of stopping at a boun…
Researchers proposed DeepInstructor, an agentic framework that evaluates research ideas by reasoning over an Experience Graph built from 58,607 peer reviews using a ReAct-based agent, according to the…
A developer detailed an agentic test creation workflow that converts plain-language agile requirements, such as Jira user stories and Gherkin acceptance criteria, into structured end-to-end test cases…
Anthropic's guidance on building effective agents distinguishes workflows, where LLMs and tools follow predefined code paths, from agents, where the LLM dynamically directs its own process and tool us…
IBM Research found that a ReAct agent running on GPT-4.1 succeeded on 77.4% of runs on average but passed all five runs on only 53.0% of AppWorld's 168 test_normal tasks, a 24.4-percentage-point consi…
A developer released ALTK-Evolve, a long-term episodic memory subsystem that lets AI agents distill reusable guidelines from past interaction traces and inject only relevant ones at action time. Evalu…
Apodex released FrontierAgent, an open-source agent runtime, terminal product, and evaluation suite for long-horizon research and file-based work, available through an OpenAI-compatible Apodex-1.1 end…
IBM Research introduced consistency guidelines in its altk-evolve toolkit, built on a new Consistency Analyzer, to address a 24.4-point consistency gap in LLM agent reliability. A ReAct agent using GP…
AutoTailor, a meta-agentic framework from researchers behind arXiv paper 2609.13548v1, cut an initial 1,283 unrefined web-agent APIs down to 87 through offline filtering and then to a 33-API set via D…
Decagon's Agent Operating Procedures (AOPs) let AI agent skills embed deterministic, code-level rules that execute without model discretion, according to a Medium blog post on packaged skills. The pos…
A multi-agent AI system called Agentic Share-of-Search (ASoS) automates competitive visibility measurement and root-cause diagnosis in LLM-mediated e-commerce, according to an arXiv cs.AI paper by Spa…
A production field guide published in 2026 argues that the PAOVR Loop — Plan, Act, Observe, Verify, Repair — is the control pattern that consistently finishes real agent work, synthesizing ReAct (Yao …
An agent harness is the code and structure around a language model that converts raw text prediction into goal-directed behavior, and it has driven larger performance gains than model upgrades, accord…
Anthropic's Claude Code introduced a real-time multiplayer, agentic planning environment called Chopin, which aims to address limitations in current AI planning tools by enabling interactivity, team c…
A case study from NeraBlog reports that prompt caching cut LLM inference costs by 36% in production for a human-in-the-loop ReAct agent handling VIP customer support for a gaming company. The optimiza…