AI-Driven Development Series
Welcome back.
Part 2
In Part 1,
we talked about WHY.
The business case
is never fire.
How do you actually
DO this?
There's a system
at your company.
The one nobody
wants to touch.
The one that's been
running for years.
Built by people
who left long ago.
"Nobody knows
why that's there."
Dependencies
frozen in time.
Things don't
get improved.
Security vulnerabilities
linger.
Your legacy system
IS an executable
specification.
The behavior is
already there.
Captured in the
running system.
Think about it
like cooking.
Legacy systems
are leftovers.
Tests aren't just
for testing.
Record what
the system does.
These become
your fixtures.
The weird scenarios
from production.
The ones
nobody documented.
If clients depend on it,
lock it down.
Separate interface
from implementation.
Tells you exactly
what differs.
Implement. Run. Fix. Repeat.
$ harness run --fixtures ./golden
Running 247 fixtures against legacy...
Running 247 fixtures against new...
Comparing outputs...
Results: 245 match, 2 differ
4
Implement route by route.
Every route gets
green
before you move on.
Or within
tolerated bounds.
6
Release with guardrails.
Don't flip a switch
and hope.
Testing behavior,
not implementation.
The tests were
the recipe.
Every test =
safer rebuild.
How to build a
thinking factory.
Thank you.
Rebuild, Don't Patch