By Wednesday, November 12, I had a workflow with clear patterns. Flat coordination. Workers handle their own responsibilities. Coordinators orchestrate. Loop until clean.

Thursday, November 14, I’d discover that having working patterns changes everything.

The Request

Thursday morning, I needed a second workflow. Different domain. Similar structure.

I opened a conversation with Claude and said: “Look at the first workflow in this directory. I need a similar one for this new requirement. Apply the same patterns.”

Claude read the existing workflow. The coordinator structure. The worker patterns. The verification loops.

And then Claude built the second workflow using those patterns.

The Template

Claude read the first workflow and saw the patterns: flat coordination, workers handling their own responsibilities, verification loops that run until clean, phase-based organization.

I didn’t need to explain any of it. The working code was the documentation.

Three Weeks Versus One Day

The first workflow took three weeks. Week one: built hierarchy, discovered it wouldn’t work, flattened everything. Week two: simplified coordinator responsibilities. Week three: documentation and cleanup.

The second workflow took one day. Thursday morning: described requirements, pointed to first workflow. Thursday afternoon: Claude built it following the established patterns.

Same complexity. Different domain. Discovery is expensive. Application is cheap.

Working Code as Documentation

The first workflow wasn’t just instructions saying “do it this way.” It was working code showing exactly how: coordinator sequences, worker organization, result flows, verification loops.

Claude read the structure and applied it to the new domain. I provided the domain knowledge. Claude handled the patterns.

Compounding Knowledge

Now I have two workflows following the same patterns. When I need a third, it’ll be even faster. Two reference implementations instead of one.

The cost of discovery is fixed. You pay it once. The benefit of application is recurring.

First workflow: expensive discovery. Second workflow: cheaper application. Third workflow: even cheaper.

This is why taking time to get the first one right matters. The first one is your investment. Everything after is return.

Thursday Afternoon

Thursday, November 14, around 9 AM, I merged the second workflow. One day from concept to completion.

No hierarchy to flatten. No cognitive load to refactor. No patterns to discover. Just: here’s the first workflow, build one like it for this new domain.

The first workflow took three weeks because I was learning. The second took one day because I was applying.

That’s the power of established patterns.

Next: Testing Bottlenecks