← All lines
Research line 04SI · Coming soon
AI Simulation — simulate before deciding
Question
Can we simulate customers, teams and markets with AI agents to decide better — without confusing simulation with prediction?
Thesis
Populations of agents with personality and memory can rehearse decisions before they get expensive. The value is in exploring scenarios and risks, not guessing the future.
0 experiments0 articlesMiroFish · MatrAIx · OASIS (CAMEL-AI)
Research starting soon.
- MiroFish — an open-source swarm-intelligence engine: from seed material it builds a simulated world of agents with personality and memory, lets them interact over rounds and reports likely trajectories, key actors and risk signals. Built on OASIS by CAMEL-AI.
- MatrAIx — persona-based evaluation infrastructure: surveys, chatbot evaluation and prototype/app testing with synthetic users.
Questions I want to test
- Can a simulated customer base anticipate reactions to a change in service, price or campaign?
- Can simulated personas test an AI agent before it faces real customers?
- Where does simulation mislead, and which biases does it inherit?
Experiments in this line
No experiments yet.
On the radar
666ghj/MiroFishStudyingMiroFish
Swarm-intelligence engine: builds a simulated world of agents with personality and memory from seed material and reports scenarios. Starting point for the AI Simulation line.
camel-ai/oasisStudyingOASISLarge-scale agent social interaction simulator by CAMEL-AI. The engine under MiroFish.