The engineer who built Claude Code doesn't use it like most people. Boris Cherny, a former Meta engineer who created Claude Code during his first month at Anthropic, runs 10 to 15 sessions simultaneously across terminal tabs, the Claude website, and his phone.
That's not a typo. While most developers interact with AI coding assistants one conversation at a time, Cherny treats Claude Code more like a team of junior developers he's managing in parallel. His setup: five terminal tabs in iTerm2 (numbered 1 through 5), another five to ten sessions on the web, and several more kicked off from his phone throughout the day. He uses iTerm2's system notifications to know when a session needs his attention.
Plan Mode Before Everything
Cherny's most practical advice is to start every session in Plan Mode, a feature that forces Claude to ask clarifying questions before writing any code. "A good plan is really important," he says. Before any code gets generated, he has Claude ask about integrations, tech stacks, and project requirements, then reviews the answers before giving the green light.
This matches what power users of any AI coding tool have figured out independently: the quality of the output scales directly with how well you define the problem upfront. Skipping the planning step is the single most common mistake.
Teleporting Between Devices
One detail that flew under the radar at launch: Claude Code's --teleport command lets you move an active session between your local terminal, the web interface, and mobile. Cherny uses this constantly, starting something on his laptop and picking it up on his phone when he steps away.
For context, Claude Code had over 80% adoption among Anthropic's own engineers before it even launched publicly in May 2025. It started as Cherny's side project and spread internally through word of mouth.
Cherny is careful to note there's no single correct setup. "It was intentionally built in a way that you can use it, customize it, and hack it however you like," he says. His workflow reflects his context at a large company with a full engineering team, so the specific numbers (15 sessions!) won't map to everyone. But the principles do: plan before coding, run tasks in parallel instead of sequentially, and use cross-device mobility so you're not chained to one screen.