Leaving the phone field blank broke our booking form. We fixed the wrong thing seven times before we looked in the right place. Here's what changed.
We shipped a Light mode nobody could read while every test passed. Now every screen is checked as a picture, in both themes, in Safari and Chrome.
Every new screen goes to a practice company first. Real Bench workspace, nobody's real money. Your account sees the change after it's been through there.
AI models build most of Bench. The model that writes a feature never gets to approve it. A separate model tries to break it first.
When we teach an agent how your company does a job, it gets written down once. Every agent reads the same page. Change the page and they all change.
An AI model can only hold so much at once. Most tools drop the oldest messages. On a Bench harness they get filed and summarized, never deleted.
Fixing the bug on your screen is half the job. We name what kind of mistake it was and add a check so that whole kind can't come back.
Allowed to close its lid.
The agents live on the office Mac that never sleeps. A MacBook Air can help as a second machine, and it's allowed to close its lid.
A page first.
Anything we build for you shows up as a web page first, so you can read it and change it. The PDF comes after you say it's done.
