# The 30-Day Ramp — Weekly Checkpoints

Four weeks. Single tasks, then connected tasks, then your first automation, then
your first guardrail. It will feel slow. That's correct. Every big system starts
as one small, ugly, verified win.

Budget your time honestly: Week 1 needs twenty to thirty minutes a day. Weeks 2
and 3 run closer to forty-five. Week 4 is an hour on the days you build your
check. If your week explodes, stretch the ramp to sixty days without guilt. The
calendar isn't the point. The reps are.

---

## Week 1 — Single Tasks (Days 1–7)

You should have completed at least three tasks. If you haven't, your tasks are
too big. Break them smaller. "Overhaul my entire client communication strategy"
is not a task. "Rewrite Tuesday's status email so it takes five minutes instead
of twenty" is a task.

If you completed three tasks and caught zero errors: you're not checking
carefully enough. Go back and look harder. The errors are there. They're always
there.

## Week 2 — Connected Tasks (Days 8–14)

You should have caught at least two lies. Not "things that were slightly off" —
actual errors that would have caused a problem if you hadn't caught them. A
wrong number in a report. A made-up deadline. A claim you couldn't verify.

If you haven't caught any: you're either not checking carefully enough, or
you're not giving the machine tasks complex enough to expose its failure modes.
Push harder. Give it something with real data and real stakes.

## Week 3 — Your First Automation (Days 15–21)

You should have one process that takes less time than it used to. Even ten
minutes saved is real. If you saved zero time, one of two things happened:
either the task was too complex for the machine at this stage (pick a simpler
one), or you spent more time correcting the machine than the task would have
taken by hand. That second one usually means your instructions aren't specific
enough. Go back to the instructions and find the vague parts.

## Week 4 — Your First Bank (Days 22–30)

You should have one guardrail in place. One rule. One check. One moment in your
workflow where you stop and verify before the output leaves your hands.

You are now operating with more AI governance than the vast majority of people
using these tools. Not because you're technical. Because you got burned, you
learned from it, and you built a bank so it doesn't happen the same way again.
That's the whole system, at its core. Everything else is scale.

---

## After Day 30

If you've done the work — not read about the work, done the work — you now have:

- A working process for at least one recurring task
- A context profile the machine knows about you
- A lie list that reveals the machine's failure patterns in your domain
- One governance check that protects you from the most expensive failure

Start ugly. Start now.