Characters

Behavioral archetypes — the people doing the work, and how AI amplifies their instincts.

character

The Fighter

Charge first, ask questions never

The Fighter is defined by a bias toward action so strong that it can crowd out reflection. They trust their tools, their instincts, and their ability to power through consequences. AI tools amplify almost every one of the Fighter's instincts — and that's the problem.

Volume 1
character

The Wizard

When all you have is a prompt, everything looks like a spell

The Wizard treats the LLM as an oracle. They ask big, open-ended questions and trust the output because it sounds authoritative. The model's fluency becomes a substitute for understanding, and the team slowly loses the ability to explain its own architecture.

Volume 1
character

The Rogue

Rules are for people who get caught

The Rogue operates outside the team's agreed processes. They paste code from the model directly into production, skip code review, and treat guardrails as suggestions. Speed is the argument, and the damage is someone else's problem.

Volume 1
character

The Cleric

The process will save us

The Cleric believes that if the pipeline is green, the code is good. They automate everything — tests, reviews, deployments — and trust the automation more than the judgment it was supposed to support. When the automation approves slop, the Cleric approves it too.

Volume 1

Monsters

The failures that emerge when those instincts meet tools that reward speed over judgment.

monster

The Scope Creep Kraken

Why not? It'll only take ten minutes

The project begins with a real goal and a sensible one. Then someone discovers that the model can produce a decent draft of an adjacent feature in under a minute, and the mood in the room changes. AI broke the throttle that used to keep scope creep in check.

Volume 1
monster

The Congealing Slop

It works on my machine (and the model's)

Congealing Slop is what happens when AI-generated code accumulates without enough shaping. Each piece works in isolation. Together, they form a substance that resists change, defies understanding, and slowly absorbs every developer who touches it.

Volume 1
monster

The Probability Pixie

It worked yesterday, I swear

The Probability Pixie is the monster of non-deterministic output. The same prompt produces different results each time, and the developer who got a great answer on Tuesday can't reproduce it on Wednesday. Consistency becomes a matter of luck.

Volume 1
monster

The Phantom Intern

Who wrote this? Nobody knows

The Phantom Intern is the code that appears in your repository with no clear author, no review trail, and no institutional memory. It was generated, pasted, and committed. Now it's your problem, and you can't ask the author what they were thinking because the author is a statistical model.

Volume 1