Control Room
Scientific request
Build a numerical experiment for a reaction-diffusion PDE, estimate convergence, flag stability risks, and produce a cautious statistical summary of error behavior.
Agent team
Combined
Statistics
Numerical Analysis
Reasoning backend lane
Orchestration layer
Self-improvement cycles
↺
1
5
Memory mode
Persistent lesson memory
Session-only memory
Benchmark gate
Safety mode
Strict public sandbox
Research sandbox
Internal prototype
Approve memory patch for this run
When off, benchmark-approved lessons are staged but not promoted into next-run memory.
Run self-improving loop
Agent Bus
Agent Conversation
Critic + Next Run
Self-Improvement
Research Memo
Gateway + Backend Plan
Exports
Agent Bus
Agent Conversation
Critic + Next Run
Self-Improvement
Research Memo
Gateway + Backend Plan
Exports