98.6% fewer tokens. Here's what that number does and doesn't mean.
What our Legwork benchmark measured, what it costs in dollars, and what it doesn't show.
What Legwork won't read, and what it writes down
How our next release handles sensitive files, keys and the audit trail, including the limits. Built and tested, not live yet.
Why we built Legwork
Frontier models spend too much effort on errands. Here's the split we made, and what we still haven't proven.