01Agents
Tools
Agents and dev tools worth using, the releases that change the work, and status-page incidents worth watching.
Other
4 new model(s) priced in the LiteLLM registry
Four new models added to LiteLLM's pricing and context registry: chatgpt/gpt-6-astra, chatgpt/gpt-6-luna, chatgpt/gpt-6-sol, and chatgpt/gpt-6.1-sol.
On KL-Regularized Policy Optimization
Paper on KL-regularized policy optimization (KLPO).
TRACE: Rollout-Guided Quantization-Aware Training for FP4 Reinforcement Learning of MoE Language Models
HF Daily Papers signal
Self-Retrospection Distillation: Turning Post-hoc Experiences into Prior Foresight
Self-Retrospection Distillation turns post-hoc experiences into prior foresight — hindsight distilled into foresight available in advance. The author is answering questions on the paper page.
DAEDALUS: Bootstrapping Agent Memory from Self-Generated Tasks
Bootstraps agent memory from tasks the agent generates itself — memory/retrieval for agents, with a project page linked from the paper.
Harness Engineering for Software Engineering via Modular Executable Dev-Primitives
HERMES turns repository components into agent-native Dev-Primitives — modular executable primitives, proposed as a harness-engineering approach for software engineering.
Choosing Before Acting: Comparative Value Estimation for Long-Horizon Tool-Use Agents
Comparative value estimation for long-horizon tool-use agents — choosing before acting; arXiv 2610.02330, submitted 1 Oct 2026 (Yu Li, Zheng Zhang, Xin Liu, et al.).
Fast Models, Slow Evidence: A Paired and Self-Audited Evaluation of System-1 Decision Models for LLM Agent Harnesses
Paired, self-audited evaluation of System-1 decision models for LLM agent harnesses — the 'fast models, slow evidence' question; arXiv 2610.02267, submitted 1 Oct 2026, single author (Jiawei Li).
Sharpening Tax in Post-Training
Top-scored paper in today's HF Daily Papers batch — post-training work on the 'sharpening tax'; abstract is on the paper page.
DAGent: Evaluate-then-Grow Planning for Deep Research Agents
Evaluate-then-grow planning for deep research agents; the authors' note on the paper page marks it as a NeurIPS 2026 paper.
Terraform for agents: how we define our software factory in code
Warp's post on defining its software factory in code, Terraform-style — the pattern behind Warp Factories, which run fleets of coding agents across the SDLC (first-pass review on every PR, bug reproduction and routing).
EVOKE: Eliciting World Knowledge in Agents for Transferable Decision-Making
HF Daily Papers selection on eliciting world knowledge in agents for transferable decision-making.
Show HN: Offrun – manage every coding agent from one workspace
Run Claude Code, Codex, AGY, and Grok Build side by side — see who is working, who needs you, and what every account has left. 76 points and 64 comments on HN.
VepAgent: Bridging Causal-Transition via Tool-Augmented Reinforcement Learning for Video Event Prediction
Uses tool-augmented reinforcement learning to bridge causal transitions for video event prediction.
EngiWorld: What Can Frontier Agents Deliver in Professional Engineering Environments?
EngiWorld bills itself as the first benchmark covering the complete professional engineering workflow — not isolated coding tasks. Useful for calibrating what frontier agents actually deliver on multi-step engineering work.
Show HN: Corral – Kill every command your agent starts
Kills orphaned processes agents leave behind — tail -f, background jobs — after the agent exits. Solves a real operational headache for anyone running agents that spawn subprocesses.
BaRe-Mem: Bayesian Reliability Memory for Robust and Adaptive Agent Consultation
Online Bayesian reliability memory that learns which agents to trust for consultation, adapting as reliability shifts. Useful for multi-agent routing where trust calibration matters.
OSWorld-Science: A Benchmark of Computer Use Agents for Learning and Using Scientific Software
Extends OSWorld to scientific software — tests whether computer-use agents can learn and operate domain tools, not just navigate generic UIs. Scientific software presents a demanding test.
JevSpawn: Adaptive Agentic Inference through Compositional Action Spaces
Training-free approach to Jev-style agentic inference — composes action spaces so agents adapt without retraining. Worth a look if you're tuning agent action policies.
EgoTools: Towards Tool-Centric Reasoning in Real-World Egocentric Videos
Benchmark for tool-centric reasoning in egocentric video — tests whether agents can reason about tool use from first-person real-world footage.
KaliBench: A Fine-Grained Benchmark for Cybersecurity Tool Use on Kali Linux with Runtime-Free Verifiable Rewards
Benchmark for agents driving Kali Linux security tools, with verifiable rewards that don't require running the tools — lets you evaluate offensive-security agents without sandbox risk.
An RL View of OPD: Least Square Policy Distillation for Sample-Efficient LLM Reasoning
Least-square policy distillation for sample-efficient LLM reasoning — a cheaper route to reasoning models if you're fine-tuning agent planners.
Measuring the Microtask Eligibility Gap: When Is an Off-the-Shelf SLM Enough for an Agent Harness?
Benchmarks when an off-the-shelf SLM can handle agent-harness microtasks — a direct answer to where you can swap frontier models for cheap local ones.
Beyond Memory: Harnessing Long-Horizon Agents with Explicit Belief States
Proposes explicit belief states over raw memory for long-horizon agents — a step toward agents that track world state without unbounded context.
A New Agentic Experience: JetBrains Air in IDEs – EAP Now Open
JetBrains opens early access for Air, a new agentic experience across its IDEs — operators can test agent workflows in their existing toolchain.
Safety of Latent Communication in Multi-Agent Systems
Examines what happens when aligned LLM agents stop communicating in text and exchange latent vectors — a safety blind spot for multi-agent systems.
The Teacher Is a Direction, Not a Destination: Extrapolating RL-Induced Representation Residuals in On-Policy Distillation
RIDE treats the RL teacher as a direction to extrapolate, not a fixed target — distills on-policy behavior by extrapolating representation residuals. Relevant for anyone distilling RL-trained agent policies.
AI: LLMs, Agents, Work, and Us - Engineers. Personal Thoughts
A DevOps engineer's personal reflections on what AI agents give and take away — and how to keep thinking for yourself while using them. Operator-grade introspection, not hype.
Sign in to Warp with ChatGPT
Warp adds ChatGPT as a sign-in option — ChatGPT users can now use Warp's agentic terminal without a separate account. Minor feature, but it lowers the friction for ChatGPT users to try Warp's agent CLI.
10 Levels of Jev For Agentic Engineers
IndyDevDan's ten levels of Jevons-paradox thinking for agentic engineers — how cheaper inference expands agent demand rather than shrinking it.
Breaking Babel: A Self-Evolving Multi-Agent System for Long-Form Subtitle Translation
Self-evolving multi-agent system for long-form subtitle translation — agents refine their own outputs across context windows. A concrete multi-agent orchestration pattern worth a look.
WorldAuditBench: Interactive 3D World Auditing with Multimodal Agents
WorldAuditBench — 213 tasks across 13 interactive 3D environments for multimodal agents. A new benchmark for agents that need to audit and act in spatial worlds.
Why Dwarkesh is Wrong about Computer Use + How OpenAI shipped its Jev competitor in 1 Week
Latent Space DevDay coverage: pushes back on Dwarkesh's computer-use skepticism and details how OpenAI shipped a Jev competitor in one week. Operator-grade context on the computer-use race.
LLMs are General Asynchronous Agents
AsyncLLM — open-source framework that lets pretrained LLMs observe, reason, and act concurrently as asynchronous agents. Directly relevant to anyone building long-running agent loops.
AutoDataBench: Can Agents Write the Data That Feeds the Self-Improvement Loop?
Benchmark probing whether agents can generate the training data that feeds self-improvement loops — the question underneath every synthetic-data flywheel.
1jehuang
1Jehuang/Jcode 0.90.1
OpenAI-compatible providers that answer 529 or another 5xx "heavy usage" error are now retried automatically instead of failing the turn, and /usage shows real quota windows for Kimi Code, Cursor, and the Z.ai Coding Plan.
can1357
OMP 18.6.1
Subagent follow-up messages now arrive as background jobs you can wait on or cancel (delivered once); fixes OpenAI context compaction wrongly rejecting sessions with many screenshots, and preserves DeepSeek-family reasoning context for warm Responses sessions.
ordewell
Ordewell 0.6.3
Closes the planner confinement hole from 0.6.2 — a `cd` chain no longer lets the research agent read outside the workspace unprompted. Terminal transport now runs OpenCode 2.x and works on macOS without tmux.
RunMaestro
Runmaestro/Maestro 0.17.8
Security release: the built-in server that maestro-cli and the phone interface drive listened on the whole network from the moment Maestro opened, even with Live off, and answered requests from any web page — anyone with its URL could run commands. It now listens on 127.0.0.1 only until Live is turned on.
vercel-labs
fx 0.0.13
Ultrafast lands on supported OpenAI models through AI Gateway (a higher-priced inference tier) — launches up to 23× faster, shell calls up to 8.6× faster, exits up to 44× faster in matched benchmarks; subagents and ACP sessions support it. Breaking: terminals launched through the shell tool now stop when fx exits, and new libfx checkpoints cannot be restored by older versions.