This editor is instrumented with the Sorsh SDK. Edit the task, run a one-click simulation, or type and
paste yourself. Same engine that powers the review queue.
Task prompt · click to edit
In about 25 words, explain to a non-expert why transformer models use attention instead of recurrence.
Your answer
Or run a simulation
Working...
The honest sim asks Groq for a casual answer to your task and types it like a person. The paste sim grabs a polished ChatGPT-style answer, tabs away, and pastes it back, exactly the pattern Sorsh is built to catch.
Run a simulation or score a submission to see the trust breakdown.