This week's story
AI Agents × Developer Tools

On-device agents and coding agent teams

On-device agents and coding agent teams

z = 3.1 11 sources 43 articles 2026-10-14 (±1 day) /en/ai/story/agents-2026-10-14
✎ Story body

Real-world examples of on-device agents and coding-agent teams arrived from several major players on the same day.

What happened

Hugging Face described fast, private on-device agents built on LFM2.5-24B-A2B. NVIDIA showed AI agents rewriting Spark UDFs for GPU acceleration and improving tool-calling reliability in chip design, Cohere shared lessons from scaling teams of coding agents, and Google Research presented geospatial agents.

Why it matters

Agents are moving from prototypes toward worked examples embedded in real engineering and business tasks. High-attention posts from five organizations landed together, raising on-device execution, team-scale operation, and reliability and efficiency at once. These are vendor-reported results, and no cross-checking has appeared yet.

What to watch

Look for third-party comparisons or reproductions on agent cost and reliability, and for signs that on-device agents spread into everyday use.

▲ Official & Press
Official

Run fast, private agents on-device with LFM2.5-24B-A2B

Hugging Face Blog ・ 2026-10-13 ・ 📌

Liquid AI releases LFM2.5-24B-A2B, a 2B-active MoE for on-device agents

Press

IFS最大イベントで新産業用AI発表　対話と自律実行で「眠る40％」を解き放つ

ITmedia AI+ ・ 2026-10-14

IFS unveils IFS Intelligence, an industrial AI platform with agents

Community

AI is mass decompiling games. That's both awesome and scary.

Simon Willison's Weblog ・ 2026-10-14

AI is mass decompiling games: awesome and scary

Official

Putting Earth AI to work: Geospatial agents for our planet's biggest challenges

Google Research Blog ・ 2026-10-14

Google puts Earth AI to work with geospatial agents

Official

Optimizing Agent Reliability and Tool Calling Efficiency for Chip Design Using NVIDIA NeMo Gym

NVIDIA Developer Blog ・ 2026-10-14

Improving chip-design agents with NVIDIA NeMo Gym

Press

Windows上でAIエージェントの実行を安全に隔離する「Microsoft Execution Containers」（MXC）正式リリース。ファイルやネットワークなどの境界を柔軟にカスタマイズ

Publickey ・ 2026-10-14

Microsoft Execution Containers (MXC) for AI agents now GA on Windows 11

Official

Amp puts AI agents to work beyond code with OpenAI models

OpenAI Blog ・ 2026-10-14

Amp uses OpenAI models to put AI agents to work beyond code

Official

Lessons from scaling teams of coding agents: reach the frontier at 30% of the cost

Cohere Blog ・ 2026-10-14

Cohere: well-organized coding agent teams hit frontier at 30% of cost

Official

How to Rewrite Apache Spark UDFs for GPU Acceleration with AI Agents

NVIDIA Developer Blog ・ 2026-10-13

NVIDIA shows how AI agents can rewrite Spark UDFs for GPU speedups

Community

Quoting Geoffrey Huntley

Simon Willison's Weblog ・ 2026-10-13

Quoting Geoffrey Huntley: fix agent habits with a pre-commit hook

Press

エージェント型AIがプロジェクトを切り盛りする日　その時、人は？

ITmedia AI+ ・ 2026-10-13

When agentic AI runs projects, what role is left for people?

Press

AIエージェントで「約7人に1人が法令違反リスク」　古いルールを参照させた代償とは

ITmedia AI+ ・ 2026-10-13

13.4% of AI-agent firms hit compliance risk from outdated rules

Academic (arxiv etc.) 31 ▾
Academic

Tev: Scaling Long-Horizon Agentic Environments with Dense Progress Rewards

arXiv cs.AI (Artificial Intelligence) ・ 2026-10-13

Academic

Hybrid Reasoning Agents

arXiv cs.AI (Artificial Intelligence) ・ 2026-10-13

Academic

The Last Remote Sensing Model Built by Humans? Casting Any Earth Observation Task as Recursive Self-Improvement

arXiv cs.CL (Computation and Language) ・ 2026-10-13

Academic

RoboSimStudio: An Interactive Toolkit for Rapid Real2Sim of Robot Manipulation Tasks

arXiv cs.AI (Artificial Intelligence) ・ 2026-10-13

Academic

Optimizing Team Behavior via Extremum-seeking Control in Multi-agent Reinforcement Learning

arXiv cs.AI (Artificial Intelligence) ・ 2026-10-13

Academic

SEAR: Self-Evolving Agentic Robotics

arXiv cs.AI (Artificial Intelligence) ・ 2026-10-13

Academic

Pack Light, CarryOn: Building Object-Oriented Context for Long-Horizon Agents

arXiv cs.AI (Artificial Intelligence) ・ 2026-10-13

Academic

BLADE: Auditing Information Use for Failure Attribution in LLM Multi-Agent Systems

arXiv cs.AI (Artificial Intelligence) ・ 2026-10-13

Academic

Knowing the User Is Not Knowing What We Established: Grounding Awareness in Personal Agents

arXiv cs.AI (Artificial Intelligence) ・ 2026-10-13

Academic

ProAS: A Progress-Aware Advantage Shaping Framework for Multi-Turn Agentic Reinforcement Learning

arXiv cs.AI (Artificial Intelligence) ・ 2026-10-13

Academic

Falsification-proof non-market allocation mechanisms

arXiv cs.AI (Artificial Intelligence) ・ 2026-10-13

Academic

Knowledge Graphs as Files: An Agent Harness Where Tool Creation and Tool Use Co-Evolve

arXiv cs.AI (Artificial Intelligence) ・ 2026-10-13

Academic

Can an Agent Learn to Predict Its Own Failure? Outcome-Grounded Confidence Calibration in LLM Agents

arXiv cs.AI (Artificial Intelligence) ・ 2026-10-13

Academic

Towards Safe Autonomy for Computer Use

arXiv cs.LG (Machine Learning) ・ 2026-10-13

Academic

The Last Mile of Coding Agents for Long-Document Tasks: An Empirical Study on Clinical Trial Protocol Digitization

arXiv cs.CL (Computation and Language) ・ 2026-10-13

Academic

Judges Follow the Reference: The Derivation Gap in Verifying Tool-Call Arguments

arXiv cs.CL (Computation and Language) ・ 2026-10-13

Academic

ReTerm: Recovery Training for Self-Improving Terminal Agents

arXiv cs.LG (Machine Learning) ・ 2026-10-13

Academic

Kairos: Transactional Security Enforcement at the System Boundary for LLM Agents

arXiv cs.LG (Machine Learning) ・ 2026-10-13

Academic

Reliable Memory from Unstable Growth: Curating Self-Written Agent Playbooks with Randomized Experiments

arXiv cs.CL (Computation and Language) ・ 2026-10-13

Academic

Receiver Decides Agent-to-Agent Negotiations

arXiv cs.CL (Computation and Language) ・ 2026-10-13

Academic

SimInter: Evidence-Grounded User Simulation with Interview-Based Profile Enrichment for Conversational Recommender Systems

arXiv cs.CL (Computation and Language) ・ 2026-10-13

Academic

JevGPT: Thinking Fast and Slow in Autonomous Research

arXiv cs.CL (Computation and Language) ・ 2026-10-13

Academic

Encrypted Skills Can Become New Security Exploits If AI Agents Know How to Decipher

arXiv cs.CL (Computation and Language) ・ 2026-10-13

Academic

When Does Opponent Modeling Pay Off? Spatial Inference Stabilises Reinforcement Learning under Partial Observability

arXiv cs.LG (Machine Learning) ・ 2026-10-13

Academic

TextualRL: Textual Reinforcement Learning for Context Management

arXiv cs.CL (Computation and Language) ・ 2026-10-13

Academic

SWE-MISBEHAVIOR: Benchmarking and Mitigating Boundary Violations in Software Coding Agents

arXiv cs.CL (Computation and Language) ・ 2026-10-13

Academic

Reinforcement Learning for Large-Scale Lifelong Multi-Agent Path Finding with Shared Spatial Fields

arXiv cs.LG (Machine Learning) ・ 2026-10-13

Academic

Fan Out or Relay? Ownership-Aware Routing for LLM Agents Querying Siloed Tables

arXiv cs.CL (Computation and Language) ・ 2026-10-13

Academic

Narrative Lookahead in LLMs: Event-Level Decodability, Predictive Advantage, and Steering

arXiv cs.CL (Computation and Language) ・ 2026-10-13

Academic

From Agents to Edges: Paired Optimization for Multi-Agent Collaboration

arXiv cs.CL (Computation and Language) ・ 2026-10-13

Academic

LatentRemedy: Efficient Remedy For Non-Prefix KV Cache Reuse in Multi-Agent Systems

arXiv cs.CL (Computation and Language) ・ 2026-10-13

← Story Archive