Vibeleaderboard
← All Intel
Intel / article

Simgym

Source
shopify.engineering
Author
Shopify Engineering
Date

Context

Shopify's SimGym sends simulated shoppers through a merchant's storefront, and a decides each shopper's next action, so every wait for the model adds to the bill. Shopify's February 2026 engineering post says each simulated shopper has a persona, a budget and a shopping intent and runs in a cloud browser from Browserbase, which bills per minute, while Shopify's own cluster runs the model. The loop is sequential: the browser sends the page state, the model returns an action, the browser executes it, and this repeats about 10 to 13 times. Shopify says that a roughly 20% cut in average model latency reduced cost per merchant run by about 10% and raised daily throughput by about 12%. These are measurements of the simulation system. The post does not claim that the simulated results match real shoppers' purchasing.

Terms in this piece · Glossary
  • LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
  • inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.
Read the source shopify.engineering
More from Shopify Engineering
Recommended reads
Comments

Checking sign-in…

Loading comments…