Area 07 · 5 jobs

Best AI for code and build, job by job

We don't rank models in general. We test them one job at a time: the same inputs and settings for every model, graded blind, with the cost of each usable result. 1 of these jobs are being tested now.

Testing now

Build a website without coding

Testing now
Why it's hard

Pretty demo, broken when edited or deployed

What we test

3 site briefs, deploy and edit tests

Coming in month 2

Coding assistant for developers

Month 2
Why it's hard

Breaks existing code, high cost

What we test

10 real repo tasks

WeChat mini-programs and apps

Month 2
Why it's hard

Few tools handle mini-program frameworks

What we test

2 mini-program briefs

Coming in month 3

Automations and scripts

Month 3
Why it's hard

Scripts fail on real data

What we test

10 tasks

Dashboards and data viz

Month 3
Why it's hard

Wrong numbers, ugly charts

What we test

3 datasets

How we choose and grade: our methodology.