You already use Claude Code. You use it the way most people do: one window, one prompt, waiting. This course rebuilds that into a day. Teach it your job once so you stop repeating corrections, open a second session safely with worktrees, run the morning from a single agent view, lay out four panes that tell you at a glance which one needs you, and hand the repetitive hours to hooks, loops and scripts. It ends where no other course goes: the honest ceiling. Two to five agents is a limit of human attention, not of tooling, and the bottleneck moves to review the moment you pass it. Windows is the spine here, not an appendix. Ends with your own configuration directory, exported as a starter kit you keep.
A real install, an honest look at what it costs, a terminal you can navigate, a session you can leave and come back to, and a mistake you undo on purpose.
FirstPrompt is in early access. These are notes from the people testing it, shared by role instead of made-up names, while the first public cohort gets going.
I'd avoided the terminal for years. Every tutorial just assumed I already knew the basics. Here I ran my first real command in about a minute, right in the browser. No install, no panic.
Product designer
Day-one beginner
Honestly the daily streak is what got me. Half an hour after dinner, and by week three I had a landing page I was actually proud to send people.
Marketing lead
Never opened a shell before
The code review on every submission felt like sitting next to someone really patient. It caught stuff a video course would never have bothered to tell me.
Solo founder
A bit of terminal experience
FirstPrompt is a small independent studio teaching people who don't code to get real work out of Claude. Every lesson is written and tested end to end by the same team that builds the product — the courses, the tutor and the sandbox are one workshop, not licensed content.
Everything you need to know about the course, the sandbox, the cert, and the money.
Stop paying the same correction twice. A short project file, a memory you audit, your first skill, plan mode used as triage rather than ritual, and a prompt that grades itself.
The first real parallelism, taught the safe way: two isolated sessions in one command, the guardrail's honest limit, a reviewer who did not write the code, and what the second window costs.
The module the course is named for. Morning triage from the agent view, work handed off and walked away from, four panes that say at a glance which one needs you, and a status line that makes them legible.
The hours that should not need a human. A rule that actually holds, a hook that stops you losing money, a loop that keeps the day moving, Claude inside a script, and the tools your day already uses.
Where the bottleneck actually moves, how to read a claim of forty agents, reviewing at speed without rubber-stamping, the five ways this goes wrong, and what to do while it works.