What vibe coding taught me about what we will lose and gain from working with AI
When journalist Sam asked me which AI tools would be suitable for doing visual investigative journalism, I gave an answer you wouldn’t have expected.
Do you know the latest Gen Z news apps yet? Let me present you TruthTrace, CheckMate and FactFlow. They are fact-checking apps where teens and 20-somethings invite their followers to check information together. Are they popular? No. Well-known? Neither. But they do exist – yes, they do. In fact, I created them myself while on the train. Until now, I was unable to build an app myself, but with a new line of AI tools and the following task, I generated all three in no time:
Create a visually oriented fact-checking app where Gen Z influencers invite their followers to check claims and research them together.
– The only vibe coding prompt I needed to build three apps in no-time
This was the prompt I put into several, so-called vibecoding tools, with names like Lovable, DeepSite and Bolt. Through these kinds of, largely free, AI-driven web applications, you create working code and visual, interactive apps in just minutes. They come with AI-conceived titles, buttons, forms, swipe interactions and other functionalities. All very convincingly presented. The more specific your prompt is, the more detailed the result you get:
What is vibe coding?
Vibe coding is very much 2025. But while you’re still marvelling at the superficial Ghiblification of our selfies and other AI nonsense that pseudo-AI-experts fuss about on your Linkedin-feed, something much more fundamental is going on in AI land: vibe coding. That is: building digital things without programming knowledge.
And it’s probably here to stay. The term vibe coding comes from computer scientist Andrej Karpathy, co-founder of OpenAI and former head of AI at Tesla. Last February, he tweeted:
‘There is a new way of coding that I call “vibe coding”, where you completely give in to the vibes, embrace exponential growth and forget that the code even exists.’
Andrej Karpathy, co-founder of OpenAI
According to Karpathy, vibe coding is possible because ‘LLMs are getting too good.’ Meanwhile, that post has been liked by more than 4.5 million people. So much for the promo talk.
Vibe journalism
Because as good as programmer Karpathy says LLMs are, we journalists tend to disagree (more and more). And neither do I think it’s desirable to vibe investigative journalism or ready-to-publish end products. It’s like your surgeon saying: that appendicitis… I’ll vibe it out. No, of course not! But you do use vibe writing, vibe editing and vibe designing where it fits: in the early stages of the creative process. Not for research or publication – that’s where facts and reliability are essential.
The prediction is that in brainstorms and as input for creative discussions, editors vibe code their story ideas to give them something visible, more substantial, to get specific feedback on. When exploring alternative solutions for online audience interactions or infographics, vibe coding provides ‘live’ talking points for the team. Freelancer Sam will soon be vibe pitching a research story at a fund or editorial meeting. It offers the other side of the table a tangible representation of the intended plan and how it could live as various journalistic outputs – online, newsletter series, interactive scrolly – anything.
Jaap Meijer, working for newspaper Nederlands Dagblad shared a very vibey example with me: while donating blood, he vibecoded a Tinder version of his newspaper – all via his smartphone using his free arm. Say what?! Yes, you read that right. The resulting app contains articles from ND.nl that you interact with via the distinctive swipe to the left (or right).
Stop playing demise-by-AI ping pong
Since November 2022, the way we write, compose and create images has changed under the influence of generative AI. To that ever-growing list, we are now adding code and design. At the same time, something much more fundamental is shifting. Of course, in the near future, we won’t all be sitting around vibing out our journalism all the time.
But the following questions do arise:
- Who will, in the end, be able to actually operate, do an interview, edit text / video or write the code ?
- What if AI-powered vibes don’t quite work after all, go awry or pose a danger?
Both the question and the answer to ‘what we are losing at the hands of AI’ are irrelevant. Eventually, we will no longer get away with the now familiar two arguments: ‘that AI is not a replacement, but an extension’ or – the Godwin for many an AI discussion: ‘your prompting skills suck.’ I call this demise-by-AI ping pong. Our job is to remain timeless and technology-agnostic in identifying what we value people for. The ongoing developments – not only with AI – give reason to have frequent conversations in the newsroom about what value(s) we have and deliver, and to whom that matters. This requires, on a practical side, an extension of our current AI guidelines. But that, unfortunately, is not the end of the story.
The vibe of your audience
‘When we discuss in an editorial meeting how we are going to cover Jutta Leerdam’s engagement with Jake Paul, my 22-year-old colleagues say: I’ll ask ChatGPT about that,’ a senior magazine editor told me. I recently asked a group of Editorial Assistant students at a Graphic Design School in Rotterdam: ‘who would get sad if chatbots were taken away from you?’ All hands went up in the air. ‘We use it for everything, sir.’ To answer their manager at the supermarket in a more formal way via Whatsapp, when they dare not say ‘no’ to friends on Snapchat or to think up exam questions to their – yes really – handwritten notes. It’s hard to blame them, as every app these days has a chatbot or smart assistant baked into it.
Chatting into an uncertain journalistic future
What we are now losing is more than just traffic to our sites, apps, podcasts and newsletters. No, the vibe shift that comes with AI, results in something more profound: when AI-powered search is the first and only place and destination to get our information from, we are deprived of the active, meaningful and complex discovery of knowledge. We are given less opportunity to learn something ourselves or form our own opinions. No longer is the context offered in which the answers are found, and there is less incentive and space to browse further on our own.
While most of the recent chatbots mention sources and links to them, they are often more than a click away or a layer deeper within the interface. Computer scientist Felienne Hermans and linguist Emily Bender warn of an impoverished, biased and homogeneous reality when AI search will be dominant. Whether you’re looking for new shoes, see if it’s already peace in Ukraine or want to know exactly how old the most popular Danish TikTokker is… you’ll get every answer in the same friendly, confident and authoritative tone, peculiar to chatbots. Frictionless is what they call that in Silicon Valley. The vibe of AI, meanwhile, seems already too addictive for Gen Z and A.
Here’s what journalistic ‘vibe thinking’ does deliver
What we need is not the mock expertise of chatbots recycling common knowledge, but rather the disruptive potential of vibe thinking itself. Journalists stand out when they see patterns that AI systems overlook. Who, like Copernicus, dare to say: ‘Maybe the story is revolving about something very different from what training data suggests.’
Vibe coding gives you the chance to formulate questions not previously asked. The fact-check apps I fabricated on the train are not end products, but provocative conversation pieces. Vibe coding is not about mimicking existing journalism with AI, but about having the courage to explore new forms and ask new questions. Creating friction instead of accepting that frictionless Silicon Valley reality. That’s where the real vibe is. Good luck with your pitch, Sam!
–
This piece was originally commissioned by and published in Dutch at Villamedia, April 7, 2025