Every processed story in chronological order, with the newest coverage first. Filter by tag, source, or score to drill in.
kkt introduces a constraint-first planning layer for coding agents, replacing open-ended build prompts with explicit limit-setting before any implementation path is chosen.
Replacing raw screenshots with a compact structured payload cuts per-action token cost from several thousand down to roughly 700, directly extending how long an agentic UI automation session can run before hitting context limits.
The post offers a concrete game-design vocabulary — time resolution and unit scale — for understanding how the feel of AI coding tools changes as users move from single-agent chat to multi-agent orchestration.
The contrastive context-selection objective demonstrably outperforms simply adding more contrastive data, showing that how the training signal is structured — not just what data is used — drives grounding improvements in both agentic and multimodal LLM settings.
Despite code access giving LLM agents a measurable edge on time series tasks, a 22–34% error rate on benchmark questions exposes a concrete reliability gap that limits their use in high-stakes automated decision-making domains like finance and healthcare.
If successful, Trace Commons would give open-weight and open-source model labs access to real-world agentic coding interaction data that is currently accumulating exclusively within Anthropic and OpenAI's proprietary pipelines.
The paper provides a concrete taxonomy of coding agent failure modes and a harness-level mitigation that is empirically validated, giving practitioners a structured basis for hardening agent deployments against real-world destructive failures.
The study reveals that mainstream benchmarks like SWE-Bench Verified and Terminal-Bench 2.0 compress capability differences between agents into narrow bands, and that esoteric language evaluation exposes a qualitative gap in how strong versus weak agents construct and debug novel strategies.
Loom addresses a gap in agentic coding workflows — reliable multi-step delivery — by adding durable state and structured orchestration on top of existing agents rather than requiring a switch to a new model or editor.
SWE-Explore provides a fine-grained diagnostic lens on coding agent capabilities that binary benchmarks like SWE-bench cannot offer, enabling targeted measurement of where exploration quality breaks down before the repair stage.