One proving workspace
Launch different reasoning engines from one console without losing the original problem or run context.
UCLA · AI + MATHEMATICS
An AI workspace for mathematical research. Explore an approach, work with proving agents, and examine the evidence behind every result.
THE RESEARCH ATLAS / VOL. 03
01 / THE SYSTEM
Ansatz keeps the question, the search process, and the resulting evidence together—so a promising answer is the start of inspection, not the end of it.
Launch different reasoning engines from one console without losing the original problem or run context.
Move between pipeline and graph views to see where a proof developed, branched, or left a gap.
Keep proofs, artifacts, memory, skills, and tools attached to the work that produced them.
Add feedback to an existing run and continue from its workspace instead of rebuilding context.
Try multiple engines and models while retaining a common view of progress, output, token use, and cost.
Manage provider access in account settings and choose the model route that fits each run.
02 / WORKFLOW
Paste a statement or choose an existing problem, then select a model and proving engine.
Follow live agent output in a structured pipeline or a free-form dependency map.
Inspect the resulting proof, supporting files, verification output, and unresolved gaps.
03 / START HERE
Try the console in a browser-scoped guest session, sign in for durable account history, or read the main concepts first.