whatkindof.dev

RCB

The Detective

Every line was put there for a reason — find it.

writerreadershipperautonomytrustlevel
Take the quiz →See all types →

Who you are

You read code like a case file. A weird retry loop, a magic number, a column nobody can explain: you don't rip it out, you find out why it's there first. You go slow, you follow the evidence, and you let the git history and the tests tell you the story before you decide anything.

Where you shine

Reading what others left behind. This is Chesterton's Fence as a daily habit: you don't remove the fence until you know why someone built it. Adam Tornhill's Your Code as a Crime Scene is your method made literal: the version history is a dataset, the hotspots are evidence, and the file with too many authors is a confession. You find the truth the static analysis missed.

Your blind spot

You can mistake the archaeology for the answer. Sometimes the fence really is just old, and you've spent a day proving what a five-minute test would have shown. You trust the machinery to confirm your reading, but the machinery also lets you stop reading sooner than you do. Not every mystery is worth solving in full.

In the AI age

Reviewing AI output is detective work, and that's your native mode. The agent's code compiles and passes, but it never tells you why it chose what it chose — and motive is exactly what you read for. Pair that instinct with the backpressure you already trust: tests and types clear the obvious rot so your attention goes to the subtle wrong only an investigator digs out.

Your move

Let the machinery do the first pass so you don't read every line by hand. Save your detective work for the changes that actually carry risk, and once the archaeology is done, let the fence go.

How you're wired

Affinity. You love reading code.

Autonomy. You move slow and on purpose.

Trust. You trust the machinery.

Your shadow

The type wired opposite to yours — every instinct flipped.

The Cowboy WAJ compare →

“Ships first, reads the stack trace later.”

Developers who share your type

If you don't see the use of it, I certainly won't let you clear it away. Go away and think.
— G. K. Chesterton

Your ideal AI workflow

Let the machinery run the first pass so you're not reading every generated line by hand: tests and types catch the obvious rot, and the agent can sweep the low-risk changes on its own. Save your detective work for the diff that actually carries risk: read it like a case file, follow the evidence, and find why the agent did what it did before you accept it. Once the archaeology is done and the fence makes sense, let it go.

Who to pair with

Pair with the Scholar — an RCJ holds the system map while you work the evidence, so between the two of you the why and the how-it-fits both get read before anything moves. The Pilot (SCB) ships the conclusions your investigation reaches, turning a solved case into a release on rails. You'll clash with the Sprinter (WAB) and the Captain (SAB), who tear out the fence before you've found why it's there — don't slow their delivery, just be the one who flags the one line that's actually doing real work in the flood before it gets cleared away.

Where you'll struggle on a team

Because you won't move until you know why the fence is there, you can hold the whole team at the investigation — they want a decision and you want one more day with the git history. It reads as work stalling on a mystery only you think needs solving, and the urgency the team feels never quite reaches you. Call which cases are worth the full case file and let the rest go on a quick read, so your thoroughness guards the risky changes instead of gating all of them.

What to stop doing

Wear it

what kind of dev: RCBPut your type in your GitHub README:

[![what kind of dev: RCB](https://whatkindof.dev/badge/rcb.svg)](https://whatkindof.dev/type/rcb)
Take the quiz →