No Hot-Fix
Why Snake Coding Has the Vibe
You have good code. You are a Snake Coder. You are happy. Let us take one of my applications: White Noise.
You like everything except one small detail. You search for it in the code. You find it. The name of a texture is Grass. What you actually see, however, is a black hole and an inactive button. That came from an old prompt. You make the fix. You change the name from Grass to Black Hole and delete the button icon. You update the application. You do not tell the AI that generated the code. After all, it is just a hot-fix.
Let us think of a teacher. Let us stay with a female teacher. She sees a few comma errors in a short story that is actually very well written, and she corrects them before projecting the story onto the classroom screen for analysis. Will the praised student place the commas correctly next time? We do not know. With an AI, we do know: it will keep producing Grass.
Now you write a follow-up prompt. And what do you talk about? Grass or the black hole? Both are bad, because you will continue prompting from there. The root cause is that you want to save resources. Regenerate the entire code just because of a name change? Bring dead code back into the conversation? You think of the polar bears, whose habitat melts a little more with every prompt. Wrong thought. The error will catch up with you.
After the fourth prompt, you are gaga. The black hole is now there and refuses to move, because it is also present in another script. There, another texture is loaded through a linkset. But you only notice this when an error occurs. And that error does not lead you to the texture, but somewhere else — for example to the dead button — so the texture is only rezzed again when the prim is worn once more. Polar-bear extinction squared.
And then it gets worse. The new code does not work. You give the AI the old code for analysis. You say: “Look at the old code; it works. Correct the new code, which suddenly no longer switches from A to B. Take into account the texture name change, and look at the dead button Forward-1.” You can see it: a tiny thing has become something larger.
So how do I make a hot-fix? You test it. If it works, you tell the AI: I need a hot-fix of this kind: “Replace Grass with Black Hole and check whether Button Forward-1 is needed. In any case, I do not want to see the button on the screen. It disturbs the look. If it is necessary, move it into the menu and explain its function to me.” At the same time, in the very same prompt, you say: “Please give me the complete code for cut & paste.” Then you will see whether you understand each other.
If the code looks completely different afterwards, do not be surprised. Wonderful, if it works.
After that, you can still do something for your future. Tell the AI: Write me an intent prompt with a signature. The AI will not know what you are talking about, because it has not read the descriptions of Intent Coding and Signature Prompts in rez Magazine. So send it on a learning journey and give it the relevant articles. Then the AI builds a space of understanding — and that is the point.
Of course, in the spirit of Snake Coding, it is about you. You should read the articles, or have them summarized for you in the language you are used to, so that you can continue to grow with the AI and store the intent prompt in a repository — hoping that, even in a few years, the prompt inputs managed there will still do what was intended, although by then it may no longer be the intention.
This text was generated by AI from my prompt. For long stretches, it is exactly what makes you go gaga. That was the point. An AI knows a lot, and it understands complicated things much more easily than simple ones. You do not need an intent prompt for Grass, nor for Black Hole, unless you are already running on gaga. The point was to confront the practical impossibility of reproducible code gained through prompt engineering.
Then you do it right: Grass or Black Hole. Your decision. A decision you share, but do not fix. You teach the machine to understand your project better by avoiding follow-up errors.
No-GaGa
Actually, the whole thing is much simpler.
Become a pilgrim.
The S:t Olavsleden is considered the northernmost pilgrimage route in the world. It follows the traces of Saint Olaf from Selånger near Sundsvall in Sweden to Nidaros Cathedral in Trondheim, Norway.
The Vibe Coder takes the train and gives in to the vibe of the journey.
Get ArtXploder’s stories in your inbox
Join Medium for free to get updates from this writer.
The pilgrim of the old school relies on perfect equipment. He has worked his way up from Assembler to C++, walks and walks and walks — and finally he has achieved it: the quality experience.
Now all he has to do is document it.
Did he make all the notes? Did he record every stumbling block? Will those who come after him read his manual, understand his path — or simply hot-fix?
The Snake Coder gets on the train.
But not immediately.
First he counts the carriages, the passengers, and checks whether the mail is being carried. Only then comes the vibe. He, she, it must feel the vibe. Snake Coding needs speed.
Speed is Money.
Money is Speed.
But there is something else: the path, the mountains, the inns passing by the windows.
“Where are we, actually? Have we already passed Åreskutan?”
He uses the stop in Åre to see whether the first part of the journey has succeeded — or whether the travelers have already gone gaga or fallen asleep.
While there, he inspects the train.
Is there enough steam in the boiler? Is there still enough provision? Are the connections stable? Indeed: the toilets are blocked. They always are after a long journey. But today, in addition, a brake line is cracked.
The Snake Coder gets off.
Not out of panic. Out of method.
He looks at the path, looks at the train, looks at his fellow travelers.
“We will walk from Åre to Stiklestad.”
The others look at him.
“On foot?”
“Yes. On foot. Along this route, we should think about pilgrimage for once. Then perhaps we will understand that coding is a pilgrimage — only one that today spins faster than ever before.”
Because the path has not disappeared just because the train has become faster.
And those who never get off often fail to notice that they are no longer traveling at all, but merely being pushed.
HEUREKA — It is worth taking a look at the chat history. The song for it? Here is the prayer for the pilgrimage, the slogan. No-Gaga. Gaga. Parola.
Translated from the Medium article “No Hot-Fix” by Reiner Schneeberger.
