Skynet: Workflow-Level Anomaly Detection for Agentic AI

A new arXiv paper introduces Skynet, a method that detects failures in agentic AI systems by modeling both the semantics and the structure of multi-step workflows. Rather than judging individual outputs, it treats long-horizon plans, tool calls, and multi-agent coordination as a whole, since a single bad step such as an injected prompt or a flawed plan can derail the entire task. The authors position workflow-level monitoring as a way to catch these faults before they propagate.

Stopping and restarting strategy speeds up multi-turn agentic on-policy distillation

A new arXiv paper addresses the high cost of on-policy distillation, which relies on expensive autoregressive rollouts by the student model and scales poorly when tasks span multiple turns. The authors propose deciding when to halt a rollout and where to resume it, aiming to cut the compute spent on generating student trajectories. The method targets more efficient transfer of capabilities from large teacher models to smaller students in agentic settings.

Agent as Policy approach lets general-purpose agents control robots without task-specific training

An arXiv paper proposes Agent as Policy (AGP), a method in which a general-purpose agent handles both planning and execution while directly operating a physical robot. The authors report that this requires no training tailored to a specific task or environment. The work sits in the cs.CL category and is a revised submission.

Study compares Claude Code and Codex on gravitational-wave pipeline task

A preprint describes a methodological comparison of two coding agents, Anthropic's Claude Code and OpenAI's Codex, each autonomously running the same matched-filter pipeline on simulated Einstein Telescope data. The authors frame it as an early look at how agentic AI performs on a real gravitational-wave analysis workflow rather than as a model benchmark.

Echo-CoPilot: Agentic Framework for Multi-View Echocardiography Interpretation

Researchers present Echo-CoPilot, a multi-perspective agentic framework designed to interpret echocardiography by combining temporal evidence from multiple views with quantitative measurements and guideline-based reasoning. The work targets a gap in existing foundation-model pipelines, which the authors say handle isolated subtasks and break down when tool outputs are incomplete or inconsistent. The paper is listed as an updated submission on arXiv (2512.09944v4).

Agentic AI workflow automates TCAD calibration for oxide semiconductor transistors

Researchers present an agent-based workflow that automates experimental TCAD calibration for emerging oxide semiconductor transistors, a task currently done by hand and dependent on expert judgment. The approach targets the model ambiguity that arises when several physical models and parameter sets can fit the same measurements. The paper is posted on arXiv as a cross-listed replacement version.

ANASSA: Agentic AI Framework Targets Spatial Intelligence in GIS

A new arXiv preprint introduces ANASSA, a framework that coordinates multiple AI agents to handle geospatial analysis tasks. It combines language and multimodal models with tool execution to support geographic information system workflows. The work aims to extend agentic orchestration beyond text into spatial reasoning.

arXiv paper proposes hybrid agentic AI framework for supply chain analytics

A new arXiv preprint describes a hybrid agentic AI framework aimed at supply chain analytics, targeting tasks such as database querying, KPI analysis, demand forecasting, and performance diagnosis. The authors argue that planners struggle to use these analytics effectively for decision making, and propose combining agent-based components to handle the workflow. The work appears as a new submission without peer review at this stage.

Study Examines Limits of Agentic ICD-10-CM Coding Benchmarks

A new arXiv paper analyzes how well agentic systems perform on ICD-10-CM medical coding, the alphanumeric codes used in the US for diagnoses, billing, and epidemiology. The authors argue that standard benchmarks rely on aggregate scores that hide poor performance on harder coding scenarios. The work aims to expose where current evaluation practices fall short for complex cases.

Agentic LLM Framework Generates and Refines Counter-Narratives to Hate Speech

A new arXiv paper proposes a multi-stage agent-based pipeline in which LLMs draft and then iteratively improve counter-narratives aimed at hate speech and misinformation online. The authors argue that simply suppressing such content can backfire by increasing polarization, eroding trust, and amplifying extremist messaging. The work positions automated counter-speech as an alternative moderation strategy rather than takedown alone.

arXiv paper introduces agentic framework for high-throughput atomistic simulations

A new arXiv preprint describes a system that uses autonomous agents to run machine-learning interatomic potential simulations at scale. The work targets the difficulty of applying these potentials across broad chemical spaces, where near-ab initio accuracy is achieved at much lower computational cost. The abstract frames the effort as addressing a central bottleneck in practical use of such potentials.

Agentic AI Approach Aims to Place Research Manuscripts in Scientific Context

A new arXiv preprint describes a method that uses large language model agents to help authors situate their manuscripts within the broader scientific literature. The work targets the time-consuming and uncertain process of framing a paper's contribution relative to existing research. It is presented as a preprint and has not been peer-reviewed.

Survey Maps Cybersecurity Threats and Defenses for Agentic AI Systems

A new arXiv survey examines the security landscape around agentic AI, which combines reasoning loops, long-term memory, tool use, and multi-agent coordination. It catalogs attack surfaces and defense architectures specific to these autonomous systems, and outlines unresolved research gaps. The authors argue that conventional security models do not adequately cover goal-directed agents.

Paper Proposes Action-Level Safety Signals for Verifying NetOps Agents

A new arXiv preprint introduces a method for checking the safety of agentic network operations (NetOps) systems at the level of individual actions rather than coarse task outcomes. The work targets autonomous networks that adjust workloads and respond to incidents, where verification granularity matters for reliability. The authors argue that finer-grained safety signals are needed before such agents can be trusted in production networks.

arXiv Paper Proposes 'BlueSky' Research Agenda for Agentic Process Mining

A newly revised arXiv preprint argues that process mining should move beyond retrospective analysis of event logs toward supporting process-aware AI agents that take governed actions. The authors sketch a research agenda, framed as a BlueSky vision, covering how such agents would move from understanding recorded behavior to acting on it under governance constraints. The work positions agentic AI as a shift in what process mining is expected to deliver.

Google Opens Registration for DevFest 2026, Over 800 Global Developer Events

Google has announced that DevFest 2026 will return with more than 800 community-run events worldwide. This year's programming focuses on building, securing, and scaling systems in the agentic AI era. Developers can find a local event through the company's DevFest site to take part.

WHY IT MATTERS ↘DevFest's agentic-AI focus shows Google using its community network as a low-cost channel to standardize developer practice around its own tooling, security defaults, and deployment patterns before competing ecosystems do the same. For practitioners, it is a free way to pick up operational guidance on building and securing agent systems that official documentation often covers unevenly.

AI Agents Increase Data Center Power Demand as Industry Shifts From Chatbots

Silicon Valley is moving from simple chatbot interactions toward agentic AI systems that require more computing resources. This shift is helping drive a buildout of data centers to meet the added power and infrastructure demands. The trend raises questions about energy use as AI agents become more capable and widely deployed.

Simon Willison uses ChatGPT Work and GPT-6 Astra to generate 5K and 10K running routes

Simon Willison demonstrated having ChatGPT Work, running on GPT-6 Astra, design 5K and 10K loop routes starting and ending at his home address. The prompt instructed the model to base the routes on OpenStreetMap data. It is a practical example of an agentic assistant handling a real-world geospatial planning task.

arXiv Paper Proposes Control Framework for Persistent AI Agents

A new arXiv preprint examines how agentic AI systems increasingly retain state and operate across task boundaries, which introduces control challenges. The author argues that today's harnesses handle objectives and retries mostly manually, and proposes a framework aimed at drive and persistent alignment.

ARCHE: Agentic AI System Automates Chemical Mechanism Discovery

Researchers introduced ARCHE, an autonomous agent-based system designed to investigate chemical reaction mechanisms with minimal expert involvement. The work targets the heavy reliance on human intervention in current computational chemistry workflows. Details on validation methods and performance results are presented in the arXiv preprint.

arXiv Paper Proposes Finite Rule Revision for Verifying Adaptive Agentic Controllers

A new arXiv paper addresses the difficulty of verifying adaptive agentic AI systems, which can produce convincing outputs while being hard to validate under non-determinism and confidentiality constraints. The authors propose limiting how many times an agent's rules may be revised, framing verification around a finite revision budget. The work targets the gap between demonstrated prototype capability and dependable industrial deployment.

LLM-as-a-Judge Framework for Agentic AI in Drug Discovery Aligned With Human Raters

A new arXiv paper addresses the difficulty of scoring open-ended, tool-using LLM agents in chemistry and drug discovery, where conventional benchmarks fall short. The authors propose an evaluation system built on the LLM-as-a-Judge approach and tune it against human expert judgments to improve reliability. The work aims to make automated assessment of agentic scientific workflows more trustworthy.

SparseDitto: Agentic Compilation Framework Tunes Sparse GPU Kernels

A new arXiv paper introduces SparseDitto, a framework that uses an agentic approach to generate sparse matrix computation code for GPUs. It synthesizes implementations that account for both the sparsity pattern of the input and the specifics of the target hardware. The authors note that no single existing implementation performs best across all sparsity patterns, operators, and GPU architectures.

VALG: Agentic System for ML Theory Research

A preprint presents VALG, an agentic system designed to help with research in machine learning theory, where a learning procedure is formalised through elements such as the data model, training protocol, oracle access, loss, metric and randomness. The authors frame theorem-proving in this setting as an open problem the system aims to address, and the paper is a revised cross-listing on arXiv.

Dear Algo paper describes agentic intent layer for unified search and recommendation on Threads

A new arXiv paper introduces Dear Algo, a system deployed on Threads that sits between open-ended user requests and the platform's ranking engines. It lets people steer discovery with free-form instructions such as 'more NBA news' or 'less politics', translating them into precise signals that shape both search results and recommendations. The authors outline how the intent layer is designed to handle this shared discovery objective across the two systems.

AgenticGen: Reward-Guided Agentic Video Generation for Advertising

A new arXiv preprint introduces AgenticGen, a framework that treats advertising video creation as an agentic reasoning task shaped by reward signals rather than simple clip synthesis. The method conditions generation on a specific product and aims to optimize for online business outcomes, building on video foundation models that can produce realistic footage from multimodal inputs. The paper appears as a cross-listing in arXiv's AI and computational linguistics categories.

Experimental Study Evaluates Multimodal Prompt Injection Attacks on Agentic AI Frameworks

A new arXiv paper describes controlled experiments testing how attackers can smuggle instructions into AI agents through images, exploiting the multimodal inputs that many agent frameworks accept. Since these systems can plan, retain memory, and invoke tools with access to files, email, and external services, the authors examine how hidden image-borne injections might translate into real-world actions. The study provides an empirical assessment of how exposed current agentic setups are to this class of attack.

Study Probes Whether AI Agents Can Detect and Fix Artifact Drift in Network Experiments

A newly posted arXiv paper asks whether AI agents can identify and repair artifact drift that occurs during network experiments. The work situates this question within the network systems community's early efforts to deploy agentic AI for multi-step tasks in operational and experimental environments. It examines how far current agent capabilities extend in such hands-on settings.

Researchers introduce VLX-VR, an agentic-aware video reasoning model

A new arXiv preprint presents VLX-VR, a video reasoning model that aims to understand footage by actively collecting and combining visual, audio, textual, and temporal cues scattered throughout a clip. Rather than relying on a fixed video context with a single inference pass, the system operates agentically, deciding which evidence to seek out when parts of the video are incomplete or unclear. The paper was posted in arXiv's computation and language (cs.CL) category.

KairosAgent: Agentic Time Series Forecasting with Fused Semantic Reasoning

Researchers introduce KairosAgent, an agentic approach to forecasting time series that span multiple domains and data types. The framework is designed to pair careful numerical analysis with semantic interpretation and to fuse information from different modalities. The paper is a revised version (v2) recently updated on arXiv.

Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery

A new paper presents an automated black-box red teaming framework designed to uncover security risks in agentic AI systems. It uses a structured risk taxonomy to guide systematic testing, addressing the shortfalls of standard single-turn evaluations. The work targets agentic setups where models process untrusted inputs, invoke tools with real permissions, and act autonomously.

AWS Recaps August 2026 AI Updates Across Bedrock, AgentCore and Strands

Amazon published a roundup of AI developer features it shipped in August 2026, spanning Amazon Bedrock, Bedrock AgentCore and the Strands framework. Highlights include million-token context windows for OpenAI models, cross-Region inference, and agent workloads that can run for as long as 14 days on dedicated compute. The post also notes wider AWS GovCloud availability and a Strands offering aimed at robotics.

Heurist Finance builds AI investment workbench on Amazon Bedrock AgentCore

Heurist Finance has created a conversational investment analysis tool using Amazon Bedrock AgentCore, as described in an AWS customer story. The workbench relies on AgentCore's payments, identity, memory, code interpreter, and observability components so a small team can purchase premium market data per query and run analysis inside an isolated sandbox. The account highlights how a lean organization can assemble AI-driven financial workflows from managed agent infrastructure.

MIT researcher uses GPT-5.6 Sol and Codex to automate quantum computing experiments

OpenAI spotlighted a case study in which an MIT researcher pairs GPT-5.6 Sol with its Codex agent to carry out quantum computing work with minimal human involvement. The system runs lab procedures on its own, interprets the measurement data it collects, and handles qubit recalibration. The writeup serves as an example of agentic AI being applied to hands-on scientific research.

WHY IT MATTERS ↘Agentic AI moving from software tasks into physical lab work shifts the value proposition toward automating scientific labor itself, where validation requirements and error costs are far higher than in code generation. It also signals frontier labs competing for research-automation workloads, forcing labs to define oversight and verification protocols for experiments run with minimal human involvement.

Google DeepMind adds agentic video understanding to Gemini

Google DeepMind announced a new Gemini capability that lets the model analyze video content in an agentic, multi-step way rather than only answering single-pass questions about clips. The company says this allows the system to follow events over time, connect what it sees to tasks, and take further actions based on video input. Details on availability, pricing, and supported regions were not fully specified in the report.

WHY IT MATTERS ↘This shifts competition from benchmark video QA to deployable video agents that can chain perception with tools and actions, making continuous video analysis a practical automation layer for monitoring, editing, and interactive assistants. It also raises governance and cost questions, since always-on video ingestion and downstream actions increase privacy, liability, and compute demands that buyers will need to audit before production use.