paperarXiv (systems, data, ML)·22 Oct 20264.6
Separates task decomposition from typed-state QA to explain gains in multi-agent LLM pipelines.
newsInfoQ·24 Oct 20264.5
An 11-hour unattended Gemini 4 Argon run produced two accepted optimizations for Google ADK 200K-line Python codebase, cutting memory and latency.
paperarXiv (systems, data, ML)·22 Oct 20264.2
Routes coding agent skills at runtime between macro and micro granularity with recovery.
paperEmergent Mind·25 Oct 20264.2
Coding agents follow stale conventions in instruction files over the actual code. A one-sentence prompt fix telling them to prefer existing code improves decisions a lot.
newsInfoQ·22 Oct 20264.1
Harvard study of 718 firms finds AI coding tools raise code output but downstream review and deploy bottlenecks erase productivity gains.
newsBleepingComputer·25 Oct 20264.0
A malicious AGENTS.md file in a Downloads folder made a coding agent install a rogue npm package. Researchers reproduced it against OpenAI Codex and showed a related trick against Claude.
paperarXiv (systems, data, ML)·22 Oct 20264.0
DoorDash uses a multi-agent LLM system to clean up stale feature flags at scale.
paperarXiv (systems, data, ML)·22 Oct 20263.8
ProductWebBench grades frontend coding agents on whether the existing site keeps working.
paperHugging Face Daily Papers·22 Oct 20263.7
IQuest-Q1 is a 320B sparse MoE model built for agentic CLI tasks.
newsHacker News·25 Oct 20263.7
JetBrains open-sourced ThinkRail, an agentic IDE built on the pi coding agent with git-worktree workspaces per task, living specs and reusable skills.
newsInfoQ·22 Oct 20263.6
Vercel Labs open-sources fx, an embeddable coding agent written in Zig.
blogTrail of Bits·21 Oct 20263.6
Trail of Bits releases patching-bugs and regression-hunt, agent plugins that write fixes and then check what else the fix changed.
paperHugging Face Daily Papers·20 Oct 20263.6
Tests whether frontier agents can automate ML for science in the wild.
paperarXiv (systems, data, ML)·22 Oct 20263.0
Argues governance must scale with capability for superhuman software agents.