“Vibe coding”, creativity, craft and professionalism… are we making ourselves redundant?

In a time of “vibe coding” and instant creation at the push of a button or speaking to the computer, how do we stand out as professionals? Is the immediacy of creation to anyone a nail in the coffin of digital creativity as a craft or the start of something new?

Let’s go back in time to 1968…

2001 – A Space Odyssey is an undisputed masterpiece of cinema. It inspired tons of technical innovations, had one of the first AIs in cinema going rogue and killing people and included things like lip-reading to spy on humans when they didn’t want to be overheard. The movie is nothing but epic and a big part of this is the usage of classical music, notably Richard Strauss’ “Also sprach Zarathustra”. The trailer is already enough to show you that you are about to experience something amazing.

Things could have gone differently though. Let’s take a look at this version of the trailer instead.

The reason why this is less epic and sounds more like unspeakable things done to elephants is the Portsmouth Sinfonia. This was an English orchestra founded by a group of students at the Portsmouth School of Art in 1970. The Sinfonia was generally open to anyone and ended up drawing players who were either people without musical training or, if they were musicians, ones that chose to play an instrument that was entirely new to them.

AI makes everyone an artist

Sounds familiar? The media is awash these days with headlines claiming that you can do anything with the help of AI:

• AI democratises development
• Vibe Coding allows you to build whatever you dreamed of without needing to know how to program
• 34 prompts to make your CV stand out on LinkedIn and land you a $200k job
• XYZ allows you to create stunning videos from single images
• Use ABC to create videos from a single prompt
• Show your creativity by converting your photo to a Ghibli style image!

In other words: All you need to be a developer, musician, artist, painter, videographer or writer is AI, time and the right prompt!

The biggest mistake in this argument is that part of artistic creation is learning the craft. Only when we immerse ourselves as humans in the craft do we build something that is rewarding and feels good. Of course, there is merit in creating something and getting lots of immediate praise for it. But when the applause ebbs down we know deep down that we’re a fraud and we don’t deserve the praise. Real happiness in creation comes from having done the work, even if the outcome isn’t perfect. Remember Bob Ross and “The joy of painting”. The joy was not in creating masterpieces, the joy always is finding the artist in you.

Not so happy little clouds…

Over the last few weeks, social media was flooded with images created by ChatGPT in the style of Studio Ghibli. This felt slightly awkward though, as Hayao Miyazaki, the mastermind of Studio Ghibli is not at all a fan of AI generated art:

"I am utterly disgusted. I would never wish to incorporate this technology into my work a t all. I strongly feel that this is an insult to life itself. I feel like we are nearing the end of times. We humans are losing faith in ourselves.”

– Hayao Miyazaki on AI generated art

For ChatGPT and OpenAI this was a resounding success and for millions of short moments, people felt joy in seeing their friends and colleagues in the beloved animation style. ChatGPT got 100M new users in a day and 700M images were generated in a week. This was such a success that even Sam Altman of OpenAI asked people to hold their horses as the strain on their infrastructure and people was immense.

The flipside of creating images in that order of magnitude is that we make the world burn. If you consider the fact that creating one image uses roughly as much energy as charging your phone, this is a colossal waste of resources.

But why was this such a success? It boils down to the fact that the style of Studio Ghibli is highly recognisable. Even the current state of the stock market looks lovely in it:

Dedication to details creates lovable products

The reason why it is so beloved is that a lot of work went into creating it. For example, in “The Wind Rises”, there is a four second scene of a crowd moving that looks like a random group animation. However, Hayao Miyazaki was adamant that no CGI should be used, so the animator Eiji Yamamori spent 15 months to complete the scene. That is 96 images at 24 frames per second, so 6.4 images a month. For a four second scene! Each character in the crowd had their own backstory and identity.

This is, of course, over the top and it takes a certain kind of person to work in this fashion. The same could be said for the director of 2001, Stanley Kubrick, who was notorious for being hard to work with and driving people to the brink of exhaustion. The haunted look of the main actress in “The Shining” wasn’t acting for the most part.

But there is no doubt that dedication to the craft creates emotions, makes people connect to the final product and creates lasting impressions. The web is full of wonderful “making of” videos that show us how much love to the detail and dedication went into making Muppet Show sketches, the Calvin and Hobbes comics and the stop-motion designs of Aardman with Wallace and Gromit being the flagship product.

Stop-motion animation has to be only for people with the patience of saints. And as creating a few seconds frame by frame means creating sets for each of them it allows the makers to hide a lot of easter eggs and have fun with the surroundings. Finding yet another background joke in a Wallace and Gromit movie is adding to the joy that is already shown on the screen.

The animators of Gromit did such an amazing job with his facial and ear expressions that when asked, many people said in interviews that Gromit speaks. He doesn’t, and the creators even made fun of the fact in the latest movie “Vengeance most Fowl”. When Wallace introduces Gromit to the AI driven helper robot Norbot and tells him that it is “voice activated”, Gromit breaks the fourth wall and looks into the camera with an exhausted look.

All this effort can appear as too much for a lot of people who just want to create something. And, yes, these examples overshoot the mark by a long shot. But their success shows that there is something to it. While you will not have recognised the 15 months of work that went into the 4 second crowd animation, your brain did. Our brains strive for harmony, which is why our eyes blend colours together, make non-connected lines look like they do enclose a shape and that jerky movement isn’t really that jerky. We are so used to imperfections that our brains want to see them in art created for our entertainment.

Imperfection makes art easier to consume

With plasticine animation you literally get your hands dirty. You touch the clay a thousand times for a scene and move things around bit by bit. This leaves literal fingerprints of the artists and it makes the animation look more real to us. In “The Curse of the Were-Rabbit” there was a scene where dozens of bunnies float in a large vacuum cleaner. This was too hard to do as stop motion animation, and the decision was made to use CGI instead. When the scene was shown to test audiences, people complained that something was off and it didn’t feel like the rest of the movie. Only when the 3D designers added the random texture of fingerprints to the 3D models, people were OK with it. Only adding an imperfection to match the rest of the product made it work.

Another big filmmaking mistake became a common trick to make computer generated graphics appear more natural: lens flares. A lens flare occurs when you point a camera at a light source and you move it. For a brief moment you see a line across the screen and then you see various semi-transparent circles overlapping the image. These are the lenses of the camera and technically should never be visible. The magic of film is that you are seeing the scene, not a camera for you. So, by seeing a lens flare, the idea of you being part of the scene is gone. Your brains knows that you are just watching a movie and someone else is moving a camera. However, re-shooting scenes with lens flares is expensive, so for decades we left them in and as an audience we got used to them. So, if you create whole worlds in a computer and you never use a real camera, adding lens flares - an error, so to say - makes them look more natural, or at least closer to real movies.

Do we need even more mass consumable art?

Now that the Ghibli hype is ebbing down, people create Muppet Show style images, and LinkedIn is awash with “Action Figures” of people. There is a new idea every few weeks and they all use as much energy and represent the same “quick consumption art” we’ve had over and over again. Remember when people created Simpsons version of themselves or when we sent Flash videos of us dancing as elves using “Elf Yourself” back in the early Millenium?

The difference was that at least these products got re-used. In the AI image generation world, we keep discarding and create new images and videos instead.

Let’s take a step back to the 90s and remember what dominated the music industry: Boy and Girl bands.

Each of those bands wasn’t a band, but a product. Products based on market research and bred in castings. Each member had a persona: the romantic guy, the bad boy, the jokester and so on. In the case of the Spice Girls there wasn’t even any qualms about even calling them what they are: Posh, Baby, Sporty, Scary and Ginger. Being in a boy or girl band was a short-lived success and as you were type-casted and shoved into people’s faces on every occasion you soon had a short shelf life. Not many artists managed to have a solo career after the bands forcefully disbanded. That is not to say that they didn’t have the talent. They just had no chance to show it as they were forced to lip-sync and give the audience what they wanted over and and over again.

People don’t like fast food all the time.

The great thing about people is that you can only fool them for a certain amount of time and then they want more. When the boy and girl band craze became boring and repetitive, Grunge came around as the biggest selling music style. Unkempt, ordinary clothing wearing bands that had no personas but were just what you saw. Loud, hand-made music that didn’t go through rounds and rounds of market research until everybody was happy with it. But even this simplicity required a lot of effort. What it especially required was allowing yourself to fail.

The fact that girl and boy bands were results of castings was one thing, but the next step in force-feeding the market generic music was when these castings became shows and we saw the bands of the next season being picked in front of our eyes.

Dave Grohl of Nirvana explained in various interviews his hatred for casting shows like these. He pointed out that having to wait for hours in line and go through pre-selections just to be told in front of millions of people online and on TV that you are not good enough must be soul-crushing. He is adamant that a whole generation of musicians was destroyed by trying to cut the journey short from playing for fun and being really bad at it to world-wide success. To him, having bad instruments and not knowing what you do is a big part of the success of Nirvana. It is not about having the perfect tools, support and an immediate audience. The joy of creation is to get there incrementally.

Limited environments breed great creativity

This is one of the biggest issues of GenAI and the current fairy tale of finding the perfect prompt and creating the next big thing without knowing the craft. Sometimes having limits is a great boost to creativity. When you need to fill a canvas of a certain size, it gives you a goal to work towards. Or you could give up and say you can never fit your art in that and demand a bigger canvas.

Ultima was a game designed for the Apple II and then converted to the Commodore 64 in the 80s. There was no way to upgrade your hardware on a Commodore 64. The limits were set in stone. You have 160 × 200 pixels on a 320 × 200 pixels display, 16 predefined colours, 1 screen-wide background colour and you can use 3 colours in each 8 × 8 (displayed as 4 × 2) pixel square.

When the original designers did the game, this is what fairground scene looked like:

Possibly there was a rush to get it converted from Apple II, or there was a lack of graphic tools. In any case, the limits of the platform seem to have dictated the final outcome. Still, people have fond memories of these old games and there is a whole scene of people giving them a new lick of paint and improving the gameplay. So, whilst staying within the same technical limitations, here’s what Vanja Utne came up with for the fairground scene:

Same amount of colours, same resolution. What Vanja did here is know how our brain thrives for harmony and put the correct colours next to each other or mix them in various patterns so the brain does the smoothing for her. Knowing the limits and pushing them is the fun here, not having a green field to do whatever you want.

Questioning GenAI is not about gatekeeping

There is no doubt that allowing people to quickly create things with AI is democratising software creation and even design. The web has thrived because people of all walks of life started creating content, even if they didn’t know what they did at the beginning.

The sentence “You are not a professional developer if you don’t…”is pretty much the same as“I’m not a racist, but…” - nothing good is coming afterwards.

However, creating without understanding the impact your products have is an issue, so is giving people the idea that how the final product works isn’t important.

GenAI tends to play the hits: Popular is more important than appropriate. Code generators, for example, pick the most voted and used piece of code to auto-complete what you write. The most voted piece of code, however, most of the time is not optimised or takes into account that information sent to it is in a certain format. It was upvoted and used over and over again as it was the easiest to use. For decades software solutions were used because they were easy, they explained the how, but not the why.

Security concerns of AI code generation

With “Vibe Coding” we don’t create images, movies or music. We create interactive applications. Interactive applications that do things for our users and are a target for attackers. And attackers are always clever about this.

The latest trick is “Slopsquatting”. “Slop” is a term for useless AI generated content, most of which is used to flood social media. The term “squatting” has been used for years in security circles for techniques that use miss-spellings of domains or software packages. If you miss-spell the name of any big brand in the URL, for example, there is a high chance you end up on a search engine optimised spam page. This is called “Typosquatting” and also applies to software packages in repositories like NPM. These are full of packages containing malicious code that resemble the names of valid packages. Now we have “slopsquatting”. When Large Language Models don’t have an answer, they start giving you things that might be correct. This is called “hallucinating”. It is bad enough when AI gives us nonsense or factually incorrect answers, but in a “vibe coding” or “AI agent environment” this is much more dangerous. As we allow the AI to request code dependencies of the web and install them for us, a hallucinated software package name is a great way to include malicious code. Security researchers find more and more of these attacks. Maybe it isn’t the best idea to allow a machine to get install access to your server.

The other problem that keeps showing up in “shortcut” development is that people keep forgetting to understand and employ the most basic security and stability features. Social media is full of people posting screenshots when they get stuck in their vibe coding journey that show software access keys or passwords. Generated apps also have been reported to have access tokens in plain sight which meant that people deploying their application that “was done in 10 minutes” cost them thousands of dollars as people used these tokens for their own needs. Other developers complained that “vibe coding” cost them months of work as the coding environment keeps creating the app from scratch rather than adding to the existing source code. Without version control - a technique that is a basic development tool - this means a broken version of the generated app means you have no way to undo these changes.

“Vibe coding” is for short-lived products, not an excuse for bad quality

“Vibe coding” results in great prototypes, ideas and throwaway products. It is not about making developers highly effective or redundant. It is a wonderful way for people to build something quick and play with the technologies involved. It also is great to brainstorm ideas and interaction models. It is not safe or sensible to use to build products for public consumption unless you also run them through a diligent audit for security, performance and accessibility.

In 30 years of professional development, I learned that my best code is the one I did not have to write. My best work never started as a design – it started as a need and an idea to fulfil that need.

I made a career by:

• Making it easy for others to work with me.
• Creating code and products that others can build with and extend as they are well documented and come with tests.
• Finding great solutions for common problems – - not the fastest or simplest solution.
• Understanding things and sharing my learnings. You learn best by teaching others.

The biggest questions I get from people these days:“Is my job safe or will AI take it?”

The answer is dependent on your goals:

• If you care about what you do, if it fulfils you and you subscribe to the very human trait of always wanting to learn, you will always have a job.
• If your goal is to make quick money and create what the market wants right now, remember you are just part of a boy or girl band. Grab all you can while it lasts, but AI is better at playing the hits and creating short-lived products in bulk.

Let’s not allow the AI hype to prevent future creatives. Being bad at something, having to learn by trial and error and improving is what makes us creators.

This article was originally published in the AI magazine in German and you can also get it in PDF version in English and German on the AI magazine site.