Observed arrival · 2026-11-22
Left Running: overnight work by AI agents
An episode-based project documents AI agents working on real tasks overnight, with logs, measurements, and playable results.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Readers tracking autonomous AI agents and their failure modes
- Worth noticing
- The round counts are explicitly non-comparable: one model's rounds were tested versions, the other's self-logged rounds.
Field notes
The first comparison tracks two builds through distinct measures, including tested versions, self-logged rounds, file size, console errors, and reported cost. The site flags estimated costs and explains that the two models counted rounds differently. Its failure-mode labels are tied to logged events, such as a loop rereading the same file 365 times and feedback arriving one round late.
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
One card from the complete issue