When ChatGPT stood out as the one thing we all should know for the future, I decided to take a course in Prompt Engineering and get myself a certification. This felt like the right way to go, at the time. Now, I’m not too sure.
There’s a certain kind of LinkedIn post you’ve probably seen by now. Someone shares their ”ultimate ChatGPT prompt template”, usually formatted like a software engineering spec, complete with brackets, role definitions, and numbered steps. The comments will soon fill up with people asking if they can save it. Someone inevitably replies: ”This is gold 🔥”
Six months ago, that same person was probably still figuring out how to get the thing to write a decent email.
This is where we are with prompt engineering. What started as a practical skill, learning how to talk to language models in ways that got useful results, has morphed into something else entirely. Secret formulas, mega-prompts, and AI whisperers selling certainty in a space that fundamentally resists it. Frameworks with acronyms. Workshops with certificates. (Ahem)
Prompt engineering is drifting away from technical literacy and toward symbolic ritual. Less engineering, more corporate astrology for people trying to survive the next productivity panic.
The rise of the AI shaman
Remember SEO in the early 2000s? Or crypto prophets promising financial enlightenment? LinkedIn growth hackers with their ”engagement frameworks”? Every new technological uncertainty creates its own class of guru, and generative AI is no exception.
The AI expert economy emerged fast. Within months of ChatGPT going public, there were people positioning themselves as authorities on something most of them had been using for maybe eight weeks. The tell wasn’t just the speed; it was the tone. Confident, prescriptive, a bit preachy, and often wrapped in the language of unlocking hidden potential.
”Use this exact prompt formula to 10x your output.”
”The secret system prompt they don’t want you to know.”
”I’ve spent 500+ hours perfecting this mega-prompt.”
Here’s the thing about prompting: two nearly identical prompts can give you wildly different outputs. The same prompt run twice can behave differently. Most people who’ve used these tools seriously will tell you the skill isn’t about finding the magic words; it’s about iterative refinement. Trying things and adjusting them. Developing intuition for what works in different contexts.
But intuition doesn’t scale. It can’t be packaged into a viral post. So instead, we get alchemy. Prompt templates as status symbols. The obsession with mega-prompts: those 800-word monstrosities that look like someone tried to write a legal contract with an AI. They feel advanced, and they signal effort. Whether they actually work better than a simple, clear sentence is beside the point.
The gurus aren’t really selling technique. They’re selling emotional reassurance. In a moment when everyone’s trying to figure out if AI is going to make them irrelevant, someone promising a system, a formula, is offering control even if it’s imaginary.
Why corporations desperately want prompting to be a science
If individuals crave control, organizations are losing their minds over it.
Businesses need things to be repeatable. They need measurable workflows. Scalable best practices. Training systems that can be rolled out to 5,000 employees in Q2. Generative AI does not naturally offer any of this, and it’s making executives deeply uncomfortable. So they try to standardize the ambiguity.
This is how you get corporate AI workshops. ”Certified prompt engineers.” Internal prompt libraries with version control. Policy documents explaining the approved way to interact with language models, as if there were a guideline for talking to a chatbot.
The problem is that using generative AI is less like running software and more like collaborating with that weird, inconsistent colleague who sometimes nails it and sometimes gives you something baffling. You can develop better instincts. You can iterate. But you can’t really control it the way you control a spreadsheet formula.
Executives don’t fear bad prompts nearly as much as they fear unpredictability. A bad prompt can be fixed. Unpredictability threatens the entire operational fantasy that AI can be deployed like any other enterprise tool—configured once, optimized, then forgotten.
So we get process worship. Prompt templates are treated like compliance documents. KPIs for AI usage that measure activity but not usefulness. The theatre of control, performed for an audience of stakeholders who need to believe the technology is manageable. It’s institutional anxiety dressed up as best practices.
The placebo effect of complicated prompts
There’s a psychological dimension to all this that’s worth sitting with for a minute.
Longer prompts feel smarter. More structured prompts feel more professional. When you write ”You are an expert marketing strategist with 15 years of experience in B2B SaaS, specializing in growth-stage companies in the fintech sector,” it feels like you’re doing something. You’re being specific. You’re in control.
Compared to just saying: ”Help me write an email to a potential customer.”
The first one has weight. It looks like expertise. The second one sounds like you didn’t even try.
But here’s what a lot of experienced users figure out over time: the complicated version often doesn’t work better. Sometimes it works worse, with too many constraints, too much noise, the model gets tripped up on the roleplay instead of just solving the problem. The people who use AI tools the most tend to develop simpler habits. They treat it less like programming and more like asking a smart but literal-minded colleague for help.
The gap between what feels effective and what actually is effective gets filled with ritual.
This is where you get the aesthetics of prompting. Markdown formatting. Bullet points. Numbered lists. The ”Act as…” construction, which has become a kind of identity cosplay. Prompt frameworks with acronyms like RICE or STAR, borrowed from product management or hiring, as if interviewing an AI will make it perform better.
It’s procedural comfort. The appearance of precision in a space that doesn’t actually offer it.
And look, I get it. When you’re staring at a blank text box, and your job depends on making this thing work, structure feels safer than ambiguity. A detailed prompt is a signal to yourself and others that you’re taking it seriously. That you’re not just messing around.
But at a certain point, the structure stops being a tool and starts being a performance, basically optimization theater. You’re not trying to get a better result. You’re trying to feel like someone who knows what they’re doing.
The future of prompting may be invisible
Here’s the kicker: all of this, this entire culture of prompt expertise, mega-templates, certification programs, might be a temporary phenomenon.
The better these models get, the less any of this matters.
Future systems will likely infer intent more naturally. They’ll rely on conversation history and memory instead of needing every context spelled out in a single prompt. They’ll get better at figuring out what you actually want instead of what you technically said. The gap between casual users and ”experts” will narrow, because the thing requiring expertise will have mostly disappeared.
We’re in an awkward middle period. The tools are powerful but unstable. Useful but unpredictable. Just capable enough to matter, just unreliable enough to make people nervous. That gap between what AI can do and how much we can trust it is where all the gurus and frameworks and corporate anxiety live.
But that gap is closing. And when it does, the whole performance might collapse under its own mythology.
Which brings us to the real point.
People aren’t just trying to master AI. They’re trying to regain a feeling of control in a workplace and digital environment that’s becoming increasingly unpredictable. Automation anxiety. Job insecurity. The sense that the rules are changing faster than anyone can keep up.
Prompt engineering culture, with the rituals, the gurus, and the obsessive template-building, isn’t really about the technology. It’s a response to fear. A way of creating order and expertise in a moment that feels profoundly destabilizing.
And that emotional need is exactly why it started drifting toward astrology in the first place.
Astrology offers the appearance of insight and control over forces you can’t actually control. It gives you a language, a system, a way to feel like you understand what’s happening. Whether it’s true is less important than whether it’s comforting.
Prompt engineering culture does the same thing. It won’t tell you the future. But it’ll give you something to do with your hands while you wait to find out.
Lämna ett svar