Keeping browsers ready for incoming AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → requests creates an ongoing capacity-management problem. In its infrastructure guide, Browserbase explains that prestarted browsers remove startup delays, but unused capacity still costs money. Too few ready browsers leave requests waiting; too many waste resources. Scaling down also requires waiting for active sessions to finish, because terminating a browser interrupts its task. This explains one recurring cost behind the build-versus-buy decision: maintaining responsive capacity as traffic changes, beyond the work of launching the first browser.
Terms in this piece · Glossary
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
model routing — Sending each request to a model chosen by the difficulty of the task, rather than using one model for everything.
