Stop Being a Replaceable Printer
Code doesn’t have to be perfect. It just has to work.
I just released “Dirty Code. But Works.” - a short, manifesto-style book that challenges everything we’ve been taught about “clean code” and software engineering best practices.
The Central Thesis
Over 20% of code is now generated by AI. And tools like Claude, ChatGPT, and GitHub Copilot are barely two years old. What does this tell us?
Code is not the most important thing in this business. The problem you’re solving is.
What’s Inside
This isn’t another 500-page technical tome. It’s a focused manifesto that covers:
The Clean Code Trap
Clean Code stopped being a guideline and became religious doctrine. But here’s the thing: AI-generated code is often messy, verbose, and violates every Clean Code principle you hold dear. And it works.
You now have two choices:
- Spend hours “cleaning up” working AI code to match your style preferences
- Ship the working solution and move on to the next problem
Architecture > Code
What matters most now are the high-level decisions you make. The tooling. The frameworks. The architectural choices that will define your system for years to come.
Should you use the latest JavaScript framework or stick with Next.js? Easy choice—Next.js has incredible AI support. Claude can generate Next.js code all day long.
These are the decisions that matter now. Not whether to use camelCase or snake_case.
Code Doesn’t Matter to Clients
Let’s be real: most people don’t care how something is built. Only people involved in engineering care about implementation details.
Clients care about:
- Price
- Support
- Solutions that work
That’s it.
The New Reality
You must transition from a code writer to a software architect. Someone who can:
- Judge AI-generated code — Spot what works and what doesn’t
- Choose the right technologies — Architecture decisions that outlast any individual codebase
- Identify obvious errors — Before they become expensive problems
- Solve real problems — Not just implement elegant solutions
We Built This Foundation
AI didn’t learn to code from nothing. It learned from decades of open source contributions—millions of developers sharing their work, their solutions, their mistakes. Every GitHub repository, every Stack Overflow answer, every tutorial became training data.
We built this foundation collectively. And here’s the thing: because of open source, we actually wanted this to happen. Otherwise, why would anyone have produced all that free code?
We shared our solutions because we believed in collective progress. Software architecture is unique—we’re the only industry that gives away entire systems for free. We publish complete frameworks, entire operating systems, database engines, compilers. We document our architectural decisions. We share our failures openly.
Now AI leverages that shared knowledge to generate code faster than any individual human ever could.
But notice what AI struggles with: medical diagnosis, cleaning robots, manufacturing automation. Why? Because those fields don’t have massive open source repositories. AI can only learn from what somebody wrote down and shared.
We gave it decades of our best thinking about software development. That’s why it works so well in our domain and struggles elsewhere.
This book argues that many “best practices” have become obsolete or even harmful:
- Mocked unit tests that test nothing real
- Design patterns applied religiously without context
- The Boy Scout Rule that breaks working systems
- Over-engineering simple solutions
Instead, focus on what actually matters:
- Does it work reliably?
- Is it secure?
- Can I understand it six months from now?
- Will it still work when I need to change something nearby?
If yes, ship it. Regardless of whether it follows Clean Code principles.
Get the Book
Available now:
Available September 1st:
This is just the beginning. AI code generation will only accelerate from here. The developers who thrive will be those who embrace this change and focus on solving bigger problems faster.
The question isn’t whether AI will change how we build software. The question is whether you’ll be part of building the future or watching others build it without you.
def ship_it():
if works(): # Good enough
return "success"
# Architecture > Perfect Code
Top comments (0)