The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
AI agent eval suite: can AI write trustworthy test cases?
15 short requirements, two AI models, 3 runs each, and a rubric written before the first run. On points both models passed. With one gate ("fail if the answer states as fact something the requirement doesn't say"), only 5 and 3 of 15 cases passed. A human spot-check of the AI grader is included. Every score is public.
A third one. The two models under test were Claude Sonnet and Claude Haiku; the grader was a separate Claude Opus instance per case, given only the rubric, the case and the 3 answers. It is still the same model family, so I spot-checked its strictest verdicts by hand: I agreed...
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Related posts
Clean layout pure focus
Interactive AI Analytics Dashboard | SaaS Product Design
A modern, animated AI-powered eCommerce analytics dashboard designed to help businesses track sales performance, monitor revenue, understand customer behavior, and make data-driven decisions.
The design combines a clean SaaS interface, intuitive data visualization, interactive dashboard elements, and smooth animations to create a more engaging user experience. Every section is designed to make complex analytics easier to understand while maintaining a professional, scalable, and user-friendly layout.
Super clean work!
A little behind the scenes of a site launch video I'm working on 👀
