Observed arrival · 2026-11-05
Chaos Monkey for Coding Agents
A benchmark compares how coding-agent harnesses recover from seeded tool-call failures, tracking task completion, cost, and whether agents report failures honestly.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Developers evaluating coding-agent harness reliability
- Worth noticing
- It reports 95% hierarchical bootstrap intervals and separates two fault tracks; fx has no interception point for Track N.
Field notes
The results page distinguishes a moderate-fault condition from clean runs and defines its metrics, including overclaiming among failed trials and cost per successful trial. It cautions that the two fault tracks differ in where faults enter and which fault types they compare, so their columns are not meant to be combined or directly compared. The page attributes the work to @dani_avila7 and links to a specification and data artifacts.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
ƒJavaScriptBrowser-side code central
One card from the complete issue