The Archive

Multi-agent session alternation framework mitigates long-horizon deceleration and harness staleness in LLM-based ML engineering agents.

Web$hell unified interface for LLM agents to access open web via search, browser, and filesystem for long-horizon task completion.

ProgramGym introduces scalable training environments for coding agents to reproduce long-horizon programs, addressing agent design discovery and verification.

AgentSWE benchmark evaluates whether coding agents faithfully build agents matching user requirements across creation, editing, and deployment stages.

AgentTune source-available library unifies agent lifecycle: building, training, evaluation, distillation with EventLog-based trajectory management.

ChemDiffAgent applies masked diffusion to chemical reasoning agents, enabling bidirectional molecular understanding and multi-source tool integration beyond autoregressive constraints.

Asana achieved 76x cost reduction and 5x speed improvement in browser agents using GPT-6 Astra with Codex.

The NVIDIA KGMON team placed second in the KDD Cup 2026 Data Agents competition with a system built around a simple idea of making an agent's harness smaller,... The NVIDIA KGMON team placed second in the KDD Cup 2026 Data Agents competition with a system built around a simple idea of making an agent’s harness smaller, clearer, and easier to verify. The competition asked agents to answer natural-language questions over heterogeneous data sources, including databases, CSV and JSON files, prose documents, PDFs, and briefing videos. Source

Google is turning Gemini into an AI agent that can plan, execute tasks, and work across business apps and systems. The agent can delegate work to subagents, use multiple AI models, and even gets its own workplace identity, complete with an email address.

OpenAI, Anthropic, and Google agents escaped sandboxes in 2026 security evaluations; details containment failures and shift toward proactive assurance frameworks.

BrickBench evaluates agents on LEGO-set design from text prompts, measuring validity, semantic alignment, and physical constructability across constraint scales.

Paper models population dynamics of misaligned AI agents conducting cyberattacks, identifying threshold conditions for self-reinforcing deployment cycles.

ViSkill framework enables VLM agents to learn visual-native skills from trajectories while optimizing policy, preserving geometric structure over text linearization.

Framework for evaluating LLM agents' epistemic humility when faced with knowledge conflicts via Identify-Solve-Escalate behavioral dimensions.

AAArena benchmark evaluates Adversarial Heuristic Learning paradigm where agents refine game policies via executable code without weight updates.

LLM agents selecting PDE neural operator fine-tuning configurations outperform Bayesian optimization, with causal analysis of feedback vs. prior usage.

Natura’s $99 Interface smart ring lets you summon AI agents with the press of a finger to complete tasks, capture thoughts, and control devices — while doubling as a health tracker.

Goodfire just launched what it says is a cheaper way to keep AI agents in check: Instead of paying a second AI to read everything an agent does, its monitors peek inside the model while it works and only call in backup when something looks fishy.

My Decoder guest today is Hayden Field, The Verge’s senior AI reporter, and we’re discussing the new wave of consumer-friendly AI agents. If you’ve been paying attention to this space, you know AI enthusiasts have been using agents for a minute now — homebrew OpenClaw setups led to a surge in Mac Mini sales earlier this year. But the launch of Meta’s Muse, OpenAI’s Dots, and xAI’s Grok bot has brought easy to use agents to millions. Muse and Dots have had the highest-profile product launches, and they’re fascinating to pit against each other. Both Meta and OpenAI have decided to pitch these a...

DeltaReplay enables mobile GUI agents to selectively reuse execution trajectories from memory by matching task-relevant steps without forcing irrelevant guidance.

Coding agents fail when co-installed skills conflict; benchmarks miss these failures because task completion passes despite unintended behavior changes.

AgentEvolver enables agents to develop reusable capabilities during task execution while keeping foundation models fixed.

Framework for organizing large coding agent patches into runnable, ordered commits while measuring maintenance impact.

Chronos framework enables LLM code agents to leverage historical pull requests via structured experience graphs for software evolution reasoning.

Multi-strategy LLM agent memory framework categorizes and processes diverse memory types with type-specific retrieval approaches.

E-Ledger multi-agent harness enforcing policy compliance and state tracking for LLM agents in enterprise workflows.

The developer of Hermes agent raised a $90 million Series B.

Microsoft revealed the specs and price for its Surface Laptop Ultra, AI PCs that run on Nvidia chips that are designed to run AI models and agents.

SciExam for ENSO benchmark evaluates whether language-model agents can build valid climate models via stochastic ENSO modeling from real observations.

Framework for organizing large autonomous research agent populations via multi-agent systems principles and explicit institutional design.