Testing browser agents on toy sites hides the failures that matter.
This is a dev post classified by Jev as Testing & observability (a tool drop), kept by the Dev Radar because it carries real work, not commentary.
Testing browser agents on toy sites hides the failures that matter. ClawBench is an open-source benchmark for evaluating AI browser agents on everyday online tasks across live websites. It helps you test end-to-end task completion by running agents through real browser workflows and comparing their recorded trajectories against human reference runs. Key features: • Live-site coverage – 281 tasks across 163 live websites and 15 life categories • Everyday workflows – includes tasks such as ordering food, booking travel, applying for jobs, and managing email • Two benchmark corpora – V1 has 1
Posted by Dan Kornas (99.2k followers) 2 h ago · 3 likes · 845 views · view the original post on X. Kept by the Dev Radar as Testing & observability.
More dev work like this
- I just added to my Slow Reader a weekly check that HTTP server and DNS have right… — @sitnikcode
- Excited to ship experiment mocks! — @calcsam
- Next week Bug Bash crosses the Atlantic! — @AntithesisHQ
- It has been a really interesting experience to build an eval suite for System One Models… — @morganlinton
- Here's my first pick for this - tested with Astra Max/Luna, now giving ds4.1 a little… — @evalstate
- 服务器上突然多了个进程，知道它在跑还不够，关键是：它为什么会跑？ — @bkdgiffug
- No coincidence — @lennysan
- You can run a session-level eval. Until now you couldn't collect a human label on that… — @arizeai
Every post is read and classified by Jev (TypeSafe): what it is, which market it belongs to, and whether the link is a real tool. 22.6k posts from 5k X accounts over the last 21 days, 2.6k tools, 12 markets. Collected every 5 minutes, fully re-ranked every hour — last update 2026-09-23 15:21 UTC. Full method.