AI Guides › Getting Started
The Getting-Started Guide For People Who Hate Getting-Started Guides
By Nigel Guy · 2 min read
You've skipped the tutorial in every piece of software you've ever used, and
it's worked out fine every time. So the last thing you want is a twelve-step
onboarding flow before you're allowed to do the one thing you opened the tab
to do. Most getting-started guides are written for someone who wants a tour.
You want the task done.
The rule: do the task first, and only pick up the two or three habits that
actually save you time, after you've already seen what you're dealing with.
The anti-tour
- Open a chat and do the actual thing you came to do. Not a test
message, not "hello" — the real task, however messy. You'll learn more
from one real attempt than from any explanation of what the sidebar does.
- Notice, after, whether it asked you something you had to repeat. If
you found yourself re-explaining your role, your project, or your
preferences, that's the one setup step worth doing: tell it once, so you
stop repeating it.
- Notice whether your task would have been better as a Project. If the
real task is going to span more than one sitting, start one — it's a
folder that holds context, not a feature you need to learn.
- Ignore everything else until it comes up. The model picker, the
connector list, the settings menu — these matter when a specific task
needs them, not before.
What to skip
Skip the full feature tour entirely — you'll find the buttons you need by
using the thing, and the ones you never need were never worth a tutorial
slide. Skip reading documentation cover to cover before you've hit a wall
that documentation would actually solve. Skip "best practices" listicles
promising forty tips; two or three habits, picked up because you needed
them, beat forty you'll forget by Thursday.
Guardrails
- This works because you're willing to make one messy first attempt. If
you'd rather not risk a bad first result, a short structured setup (see
Juliemango's other onboarding guides) is the safer route — this is the
fast path, not the only path.
- "Ignore it until it comes up" is not the same as never learning it. If the
same wall keeps appearing, that's the signal to go read the specific thing
that solves it, not to keep working around it indefinitely.
- None of this replaces checking what the tool actually gives you back.
Skipping the tour is fine; skipping review of the output isn't.
All 751 AI guides · JulieMango plans from £17/mo