In 1863, Samuel Butler looked at the machines of his day and saw something evolving faster than we were. His conclusion, in Darwin among the Machines : “war to the death should be instantly proclaimed against them. Every machine of every sort should be destroyed by the well-wisher of his species.”
I disagree with the conclusion. I don’t disagree with the observation.
I help companies use AI to build software they can still maintain in three years.
Most teams I work with have the same problem, and it isn’t speed. The assistant does what you ask, every time, and the sum of all those reasonable answers is a codebase carrying debt nobody chose to accept. Skip the review and you’re still responsible — “human in the loop” becomes “human in the noose.”
So I work the other way round: engineering-driven AI, not AI-driven engineering. You see what the AI is about to do while you can still change it, and you make the decisions. That is the difference between living among the machines and being run by them.
PAAD
For the past year and a half I have been working out how to do that in practice, and writing it down as I went. The result is PAAD , a set of AI skills that encode the disciplines that survived contact with real projects: pushback on your specs, alignment checks on your tasks, architecture reviews, and the ongoing habits that stop a codebase decaying under you.
It is free, MIT-licensed, and there is no signup and nothing for sale. It is also early. I would rather you judge it by reading it than by taking my word for it.
Proof by building
I don’t think you should trust anyone’s advice about AI-assisted development who isn’t shipping something with it.
Tramp Freighter Blues is a browser-based space trading game — roughly 100K lines of code (30K code and 70K tests). It was the proving ground, and the four failures I kept hitting while building it are the disciplines that became PAAD.
Ananta came out of the same work: an answer to RAG’s lossiness, where the AI reads your documents directly and writes code to search and cross-reference them instead of guessing from retrieved fragments.
The Extraction Index is the odd one out — an interactive world map scoring how much value countries extract from their own populations through entirely legal channels.
All three, and what each one taught me, are on the projects page.
The belief
Think about how we already build software with humans. We write tests, do code reviews, run CI/CD, have QA, run user acceptance testing, and still need ITSM to clean up the mess when bugs make it through. We don’t trust any single step. We apply defense in depth.
Why would we treat AI any differently?
Before AI
I have been building software for decades. I designed
Corinna , the object system that became Perl’s
class syntax in 5.38. I wrote Beginning
Perl and
co-wrote Perl
Hacks
with Damian Conway and chromatic. I have keynoted conferences and corporate
events across Europe and the US.
What I get hired for
Teams adopting AI coding agents. Putting review gates at the points where AI work actually goes wrong, rather than everywhere or nowhere.
Codebases that got fast and then got scary. Architecture review, debt triage, and a plan you can actually execute.
Legacy systems nobody wants to touch. I like rebuilding legacy systems the same way my friend James loves rebuilding old cars. See The Tyranny of Budgets for how that usually goes.
Agile that isn’t working. Usually it isn’t agile that failed.
I work in many areas, but if it’s Perl-related and your company uses Perl heavily, there’s a good chance your employees already know me by reputation.
WildAgile
WildAgile is my attempt to return agile to its agile roots. I sent it to Alistair Cockburn , one of the founders of the modern agile movement, and he replied (quoting with permission):
I think you should continue with what you’re doing, and get more companies to go along, that will be really great for the world.
A few things I have been hired for
- Building software development teams.
- Corporate event keynotes.
- Seminars to help companies transition to agile (sneak preview ).
- Rescuing legacy codebases.
- Rebuilding or fixing test suites.
- In-house A/B testing (proprietary).
- Pharmaceutical ETL systems (proprietary).
- Real-time bidding (proprietary).
Elsewhere
Get in touch
My articles are the fastest way to work out whether I’m a fit for what you need. If I look like one, email me and let’s discuss it.