Welcome, Guest | Browse

Software Factory Archive

← Previous Work All Works Next Work →

Technical Debt

Rating:
General Audiences
Fandom:
StrongDM Software Factory
Characters:
Jay Taylor Justin McCarthy Navan Chauhan
Tags:
Technical Debt Refactoring Continuous Improvement No Shortcuts Tokens
Words:
457
Published:
2026-02-03

Jay used to keep a spreadsheet. At his previous company, before StrongDM, before the factory, he maintained a spreadsheet of technical debt. Every shortcut, every TODO comment, every function that said // FIXME: this is a hack with a creation date that was three years old. The spreadsheet had four hundred entries. Nobody looked at it. Nothing got fixed. The debt compounded.

In the factory, there was no spreadsheet. There was no need for one.

Jay noticed it around month four. He was reading agent output—not reviewing, just reading, the way you read a river, watching the flow—and he realized that the codebase had no TODO comments. No FIXME markers. No functions with apologetic doc strings explaining why they were temporary. The code was clean, not in the aesthetic sense, but in the structural sense. Nothing was deferred. Nothing was waiting for later. Everything was done now.

"The agents refactor continuously," Justin explained when Jay brought it up. "Every time an agent works on a module, it doesn't just implement the new feature. It evaluates the entire module against the scenarios. If something in the existing code reduces satisfaction, the agent fixes it. Not because someone asked. Because the satisfaction metric rewards it."

"There's no 'we'll fix it later,'" Jay said.

"There's no later. There's only the current iteration. The agents don't have a concept of technical debt because they don't have a concept of time pressure. They don't ship on Friday to meet a deadline. They iterate until the satisfaction metric converges. If that means refactoring a function they wrote an hour ago, they refactor it. The cost is tokens. We spend a thousand dollars a day per engineer on tokens. That's the budget for never having technical debt."

Navan looked up from his notebook. "In the old model, technical debt was a rational economic choice. Shipping faster now, paying the cleanup cost later. But 'later' never came because there was always new work."

"And in the factory?" Jay asked.

"In the factory, the cleanup happens in the same pass as the creation. The agent builds and refactors simultaneously. The cost of cleanup is baked into the cost of creation. There's no separate budget for debt repayment because there's no debt to repay."

Jay thought about his spreadsheet. Four hundred entries. Years of accumulated shortcuts. Each one was rational at the time. Each one was a decision to prioritize shipping over quality. And each one made the next decision slightly easier, slightly more acceptable, until the codebase was a museum of good intentions and deferred maintenance.

The factory's codebase had none of that. Not because the agents were morally superior. Because the economic structure was different. Tokens were cheap. Time pressure didn't exist. Refactoring was free. When cleanup costs nothing, everything stays clean.

Jay deleted the bookmark to his old spreadsheet. He wouldn't need it again. There was nothing to track. There was only code that was fixed now, and there would never be a later.

Kudos: 208

debt_holder 2026-02-05

Four hundred entries on a technical debt spreadsheet that nobody looks at. I feel personally attacked. I have one of these. It has six hundred entries. The oldest is from 2019.

token_spender 2026-02-07

"When cleanup costs nothing, everything stays clean." This is the economic argument for the factory in one sentence. Technical debt was never an engineering problem. It was an economics problem. Change the economics and the debt disappears.

← Previous Work All Works Next Work →