AI pair programming usually means one developer working with an AI coding assistant. thinkpool keeps that loop visible, adds room for several agent lanes, and lets a partner watch and steer the same work.
Three ways to work
Developer and AI: one developer prompts an AI coding assistant, reviews the result, and decides what happens next.
Screen share: a second person can watch, but the keyboard, terminals, and agent state still live with the host.
thinkpool room: one developer can direct the room alone. Invite a partner and both people can see and steer the same live terminals.
How AI pair programming works in thinkpool
Connect the bridge. Run it on the machine that holds the project and your installed coding runtime.
Open visible terminals. Direct Claude Code, Codex, or Hermes Agent and split work across agent lanes when the build needs it.
Bring someone in. Share the room link when another person should review, redirect, or steer alongside you.
Review what returns. Keep the plan, terminal history, handoffs, tests, and visual proof together before work moves forward.
Human review stays load-bearing
An agent can write code quickly. Someone still has to direct the problem, inspect the result, run the checks, and decide what lands. thinkpool makes that work visible to the person directing it and to the partner they invite.