01
Chat
Follow the conversation.
Talk with multiple models in one thread. Inspect alternative responses and use inline judgment to examine their differences as the conversation develops.
A workspace for inspecting model reasoning
Make reasoning visible.
Follow a conversation in Chat, trace an interpretation in Guide, and inspect differences in Compare. ASSAY gives you more to examine before you make a judgment.
This is not a benchmark leaderboard or a “which model wins” tool. The aim is to make reasoning inspectable, not to automate your verdict.
Three ways to inspect
02
Guide
Read beside the reasoning.
Keep a source text in view while you inspect claims and their citations. Noted Gaps and Not Reflected separate the model’s observations about omissions from passages its claims leave uncited.
03
Compare
Examine the differences.
Put model responses side by side and apply structured judgment. Inspect shared ground, different interpretations, and tradeoffs against the criteria you choose.
Private access
Request access.
Tell us where to reach you and, if you like, what you want to explore. Requests are reviewed by a person; accounts are created only through an invitation.
You bring your own provider API keys. ASSAY doesn’t sell model access.
Already have an invite? Sign in