AI

Drift

Every technological revolution brings its own variant of the idea that gives this article its title, by making it look very simple to produce an artefact that will be popular and lucrative.

There was, for instance, a golden age of the App Store, and for a while the famous “thing likely to be lucrative” was an app. Do you remember Flappy Bird?

Released in 2013, this rather crude and infinitely frustrating game was available for a brief spell1 on the App Store before its creator, overwhelmed by its success, pulled it. At the time, iPhones with the game installed were offered for several hundred dollars. Even though it isn’t certain that any sale went through, the excitement of those days left a lasting trace in my mind as an example of improbable and (seemingly) easy success: “incredible, now you can develop a thing in a few days and make millions”.

Nevertheless, for lack of skill and time, I hadn’t developed a thing in a few days, and consequently hadn’t made millions2.

The years went by, AI arrived, and here we are a few months ago, while I was having fun with vibe coding platforms which, as the practice was taking off, were particularly generous with features and free (or cheap) usage. One of them, Replit3, even acquiring a mode dedicated to making games.

That was all it took for Flappy Bird to resurface in my memory, and for me to set the AI the following challenge: make a three-dimensional version of it — 3D, to its friends.

As usual with AI at the time, I got a thing that more or less worked, and that grew out of control with every correction and every new idea. On top of that, I was incapable of evaluating code meant to render 3D in a web browser4. In the end, I had a game that worked, I played it a little, and I forgot about it.

The weeks went by, better AIs arrived, I made progress in my ability to develop thanks to AI, and I wondered what to do with this experiment. So, to test my dear Claude a little, I challenged it to condense the code that had been written as far as it would go. From roughly 5,500 lines, we went down to fewer than 200, a very short file in HTML format.

Having reached this point in my story, I recall a recurring piece of feedback about my articles on AI: philistines5 can find them impenetrable. I will try the culinary metaphor to keep my whole audience. Imagine a cake recipe, extremely detailed, step by step, gram by gram, that would be summed up in 3 lines and would give a much simpler result, but one that kept the big idea of the original cake6.

Indeed, the game no longer has anything of a steroid-pumped 3D Flappy Bird: it is a simple orb moving forward.

And so we can come full circle: AI can sometimes look like the new avatar of “let’s make an easy thing that makes millions”. Undeniably, the barrier to entry for building a working tool (an app, a website, a training course) has dropped a great deal. Not everyone has become a millionaire for all that. Unless something escaped me. I like the idea all the same: might as well try, you never know, and if something catches on, then it will no doubt be time to devote oneself to the phase that remains hard: promotion, distribution, customer acquisition.

It probably isn’t Drift, my simplistic 3D variant of Flappy Bird, that will make a millionaire of me. It is only a small brick in my experiments with AI, a game you can play in my portfolio, and now the subject of an article in my newsletter.

Click or tap to play.

Footnotes

  1. At least in my memory; in fact, it was above all its popularity that was very short-lived. ↩

  2. Note the difference with the present day: I now make things in a few days. Not millions yet, but who knows what might happen if this newsletter reached, say, 1,000 paying subscriptions? ↩

  3. Which I can only recommend to anyone who wants to try a little vibe coding and quickly have something that works. ↩

  4. I still am. ↩

  5. Using this word for novices is probably a misuse, given that it is a little pejorative, whereas far be it from me to want to be so towards my readers. But I like the word, so I allow myself this liberty. ↩

  6. Someday I will have to look into the question: why are the explanatory metaphors that come to me about AI so often culinary? ↩

Play

Elenchus read this story

Reasoning strong

Le texte présente une réflexion personnelle nuancée sur l'illusion de facilité et de rentabilité associée aux nouvelles technologies, en s'appuyant sur des exemples concrets et en évitant les généralisations abusives.

What holds L'auteur utilise des exemples précis (Flappy Bird, Replit, son propre projet) pour illustrer son propos sans tomber dans la généralisation hâtive.

My newsletter

In French: the books I read, the exhibitions I see, the AI I practise, the biases that catch me — and right now, a writer’s apprenticeship in public.

All issues