Tag: coding agents
Microsoft lets a thousand agents organize themselves on one job
The Agensh harness drops the central orchestrator and lets workers claim their own share of a coding task.
A study finds agent harnesses mostly buy you tokens
Two researchers held the model constant across three coding harnesses and found less than two points of difference…
Trail of Bits finds its own patches broke working code
After letting agents write security fixes, the firm audited what it had merged and confirmed 21 regressions across…
Replit walks away from Anthropic as its CEO praises OpenAI
Amjad Masad says cost and benchmark results pushed Replit off Claude, though people close to the partnership dispute…
SoL-Pi cuts coding agent token bills by rewriting the harness
The four mechanisms came out of automated research loops and trim token traffic by nearly half on a…
Claude Code learns to read the cross vendor AGENTS.md file
Version 2.1.277 lets Claude Code fall back to the open AGENTS.md convention when a project has no CLAUDE.md.
Anthropic rebuilds Claude Code projects around cloud threads
Claude Code projects now coordinate parallel cloud sessions with shared memory, leaving local workflows waiting.
Cognition post-trains SWE-2 on Kimi K3 to undercut Fable 5.1
The Devin maker tuned Moonshot's open model with reinforcement learning and says the result lands within a point…
Muse Spark 1.3 trims Meta’s agent costs and tops a coding chart
Meta's fourth Muse Spark release in five months cuts token use by a quarter and edges ahead on…
GPT-5.6 family arrives in AWS Kiro for cheaper agent builds
All three GPT-5.6 models are now inside AWS's Kiro, where spec-driven builds cut the cost of successful tasks.
DeepSeek opens Harness so builders can assemble their own agents
The Chinese lab's open-source agent framework takes on Claude Code with an everything-is-a-plugin design.
Approval fatigue lets a third of dangerous agent commands slip
A 40,000-run browser game finds humans approve about one in three malicious requests.