Field notes
9 min
We measured 1.2 million runs. Here is what fails.
Most failed runs were not the agent getting something wrong. They were a tool being down, or a rule nobody wrote down.
Theo Lindqvist
I.04
Keep reading
I.04
More from
the journal.
All articles
→
Changelog
3 min
Changelog 4.2: approvals in team chat, and undo for 30 days
Approve requests without leaving your team chat, and undo any action for 30 days instead of 7.
Guide
7 min
Where the line sits: deciding what an agent may do alone
A practical way to decide which actions need a person, which don't, and how to move the line over time.
Field notes
9 min
We measured 1.2 million runs. Here is what fails.
Most failed runs were not the agent getting something wrong. They were a tool being down, or a rule nobody wrote down.