The future of AI-driven development isn’t Agile. It’s XGH.
The eXtreme Go Horse method perfectly complements the strengths of AI-powered dev tools. Introducing XGH+AI.
A lot of ink has been spilled over Large Language Models and their associated products being inserted into just about everything, often against the will of the product’s users. One of the areas where the technology has managed to gain a foothold is software development; even as it slowly drains the will to live from Microsoft developers, their leadership celebrates the extent to which AI-generated code has been integrated into the company’s delivery pipeline. They’re not the only ones — pick a random post on your LinkedIn timeline and most likely it’ll either be gassing up the latest AI tool or lamenting their use.
Something is clearly not right here. How can exactly the same product be simultaneously The Future and less than worthless?
I think I have the solution. Those who find AI most useful have, in a process of convergent evolution, stumbled upon a secret known only to few.
eXtreme Go Horse
When Daniel Alonso dropped the first English translation of the XGH whitepaper in 2018, Brazilian developers had already been practicing it for nearly a decade. But outside of maverick organizations like Tesla, the methodology didn’t see much adoption among English-speaking tech orgs, with the industry preferring to stick to its stodgy Agile ways.
But let’s face it — Agile is old news. Software development in 2025 is a far cry from 2018. Change has come for us in the twin forms of mass layoffs and AI transformation. We are now called upon to do more, with less than ever.
You need to be more comfortable with error, with mistakes, with things that are a little rough around the edges. — Kevin Roose, XGH Black Belt
Which is why eXtreme Go Horse meets the moment in a way that old ways of working simply do not.
I believe that the reports of high productivity increases from GenAI tools are coming from individuals and workplaces that have (unknowingly) reinvented XGH independent of its original developers.
I am a firm believer in cross-pollination of ideas. I think AI tool users will see a lot of value in seeing the full list of eXtreme Go Horse principles laid out for them, and XGH practitioners will have their eyes opened to the power of AI technology.
The New Principles of eXtreme Go Horse
Obviously, XGH was written long before the idea of combining all the data in the world with all the money in the world was a twinkle in Sam Altman’s eye. I’ve taken the liberty of enhancing the original principles with proven capabilities that AI tools have demonstrated. Let’s call it XGH+AI.
I think, therefore it’s not XGH
In XGH you don’t think, you do the first thing that comes to your mind. There’s not a second option as the first one is faster.
The original Agile manifesto emphasized talking to customers. But the “job” that businesses “hired” it to do was increasing velocity, which it’s bad at. XGH+AI is a much better process for improving velocity. You can save a lot of time by outsourcing not only coding but also problem-solving to the AI — just ask it to generate whatever it is that users want.
There are 3 ways of solving a problem
The right way, the wrong way and the XGH way; which is exactly like the wrong one but faster. XGH is faster than any development process you know.
If you learn by shipping, XGH+AI will let you ship the wrong thing much faster than any old way of working, and the shipped thing will be more wrong than anything you could have dreamed of. Therefore we will learn faster than ever. It just makes sense.
You’ll always need to do more and more XGH+AI
For every bug solved using XGH, seven more are created. As all of them will be solved using XGH, then it’s utility tends to the infinite.
This is the perfect consequence in a world where using AI to code is mandatory regardless of whether or not it’s actually helpful. It will never run out of things to do and therefore you will never be fired for non-compliance.
XGH is completely reactive
Errors only come to exist when someone notices them.
AI tools are designed to discourage error-checking. This makes them perfect for XGH+AI’s radical approach to errors: you ask the AI to create tests, it pretends that it did, and then you are blameless for what happens next.
In XGH, anything goes
It solves the problem? It compiled? You commit it and don’t think about it anymore.
AI tools have been found to not boost developer productivity. But that’s only because they stopped to fix the bugs in the AI’s code. If you simply don’t worry about it, you can commit twice as fast.
You commit always before updating
If things go wrong your part will always be correct… and your colleagues will be the ones dealing with the problems.
The fastest gun is always the winner. If you use XGH+AI you’ll always ship faster than your colleagues, taking advantage of anchoring bias to make yourself look better.
XGH doesn’t have schedules
Schedules given to you by your clients are all but important. You will ALWAYS be able to implement EVERYTHING in time (even if that means accessing the database through some crazy script).
Sometimes ChatGPT goes down, and that means you can’t code anymore. That’s just life. I’m sure they will understand.
Be ready to jump ship or blame someone else when it goes down
For people using XGH someday the ship goes down, as time passes by and the system grows into a bigger monster. When that day comes, you’d better have your resume ready or have someone else to blame.
AI is the perfect scapegoat. Just say “it’s what ChatGPT wrote” and you will be off the hook for all problems.
Be authentic. XGH doesn’t follow patterns
Write code as you see fit. If it solves the problem, commit and forget about it.
If you relax your standards and let the LLM write code how it wants, you won’t have to fix it. That improves your velocity (see above).
There’s no refactoring, just rework
If things ever go wrong just use XGH to quickly solve the problem. Whenever the problem requires rewriting the whole software it’s time for you to drop off before the whole thing goes down.
AI can’t refactor so it’s best not to try.
XGH is anarchic
There’s no need for a project manager. There’s no owner and everyone does whatever they want when the problems and requirements appear.
AI has democratized not only coding but also management. You can just work on whatever you want and then say “I’m innovating with AI” if anyone asks.
Always believe in improvement promises
Putting TODO comments in the code as a promise that the code will be improved later helps the XGH developer. He/She won’t feel guilt for the shit he/she did. Surely there won’t be any refactoring.
Believing that AI will get better soon™️ is one of the most important tenets of being an AI powered rockstar.
XGH is absolute
Delivery dates and costs are absolute things. Quality is relative. Never think about quality but instead think about the minimum time required to implement a solution. Actually… don’t think! Do!
AI companies are setting a new standard in launching whatever, whether it works or not. You think you’re better than Apple and Google? Isn’t that a little arrogant of you?
XGH is not a fad
Scrum, XP? Those are just trends. XGH developers don’t follow temporary trends. XGH always was, and always will be used by those who despise quality.
AI is definitely here to stay. That’s what they keep telling us, so it must be true. This means XGH+AI is also here to stay. You will always be able to use these tools for cheap/free and they will never raise prices on them.
XGH is not always WOP (Workaround-oriented programming)
Many WOP require smart thinking. XGH requires no thinking.
With vibe coding, it has never been easier to make 70% of something that works. Simply leave the constraints you don’t like for that 30% and someone else will take care of it.
Don’t try to go against the tide
If your colleagues use XGH and you are the only uptight coward who wants to do things the right way then quit it! For any design pattern that you apply correctly your colleagues will generate 10 times more rotten code using XGH.
Billionaires are telling you to give up and embrace the AI, or else. And they must be really smart, otherwise how did they get to be billionaires? Oh, it’s from selling AI tools? I see.
XGH is not dangerous until you see some order in it
An XGH project is always in chaos. Don’t try to put order into XGH as it’s useless. You’ll spend a lot of precious time and it will make things go down even faster.
Priorities have changed. You aren’t going to win any points for trying to bring order. Let it go.
XGH is your bro. But it’s vengeful
While you want it XGH will always be at your side. But be careful not to abandon him. If you start something using XGH and then turn to some trendy methodology you will be fucked up.
Study after study shows that AI erodes critical thinking. But with XGH+AI you don’t need critical thinking anymore, as long as you never try to stop using XGH+AI.
Reconciling XGH+AI with Reality-Based Business Models
Both traditional XGH and its AI-powered update have a small problem, which is that generating ten times more code that’s ten times worse quality without waiting to see if it solves anyone’s problem or if anyone will pay for it is not actually going to help your company make a profit.
Fortunately, that kind of thinking is obsolete. Pretending to create value is just as good as actually doing it, as long as some rich guy believes in it hard enough.
And if this methodology sounds cynical and heartless — if you cringe to imagine anyone doing this to your product or code base — it’s not too late to stop outsourcing your thinking to a subscription product, and re-learn the skills you’ve already started to lose.
