Few hundred to few thousand LLM queries fit laptop-scale agent society simulators
Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.
You can replace each LLM agent in a multi-agent simulation with a cheap low-parameter surrogate fitted from a few hundred to few thousand real elicitations (a few dollars on DeepSeek), then scale to arbitrary agent counts on a laptop instead of paying per-agent inference at every step. Critically, whether the surrogate holds is predictable in advance from an interaction-order × memory taxonomy—so if your simulation's questions are macroscopic (phase behavior, scaling in N) rather than individual cognition, you can decide up front whether to skip the full LLM run entirely, with error trends and even saturation-driven failures predicted parameter-free.
Simulating large LLM-agent societies now costs a few dollars for a few thousand queries to fit a low-parameter model, enabling running such simulations on a laptop at any scale $N$, validated on eight named LLM simulations including EconAgent, with predicted error trends holding across different agent perception and memory configurations.