GrowingI have seen this movie before. Several times. The actors change. The soundtrack improves. The promises get louder. The ending stays the same.
Around 2000 it was this.
Everyone needs a website. Learn HTML in a weekend.
Suddenly the internet was full of blinking text, broken navigation, tables inside tables inside more tables, and a mysterious visitor counter at the bottom proudly displaying the number 7. Most of those sites died quietly. Some are still out there, fossilized in forgotten hosting accounts, whispering in Comic Sans.
Then 2008 arrived.
Mobile apps are printing money. Just learn iOS development.
Every developer with a MacBook became a visionary founder. To do lists. Flashlight apps. Another to do list but this one is social. The App Store filled up. So did the graveyard. Turns out shipping version one is easy. Maintaining version twelve while Apple changes everything and users demand perfection is not.
Then 2015.
Blockchain will change everything. Smart contracts. Decentralize the world.
Whitepapers were written. Tokens were minted. Conferences were attended. Most projects never launched. Many quietly shut down. The few that survived did so not because of hype, but because someone somewhere treated it like real engineering instead of a lottery ticket.
Now it is 2024.
AI will replace developers. Build a GPT wrapper this weekend. Launch an AI SaaS by Sunday. Retire by Wednesday.
Result?
Still loading.
The pattern is so predictable it should come with documentation.
First, a new technology emerges. It is genuinely powerful. It solves something real. It unlocks new capability.
Early adopters see opportunity. They experiment. They struggle. They learn the sharp edges. They build things that actually work.
Then the get rich quick crowd floods in.
Tutorials appear. Build X in 30 minutes. Ship your AI startup in a day. Passive income while you sleep.
Reality, as it always does, shows up uninvited.
Infrastructure costs are not passive. Hallucinations are not a feature. Latency is not just a vibe. Security matters. Data privacy matters. Customers do not care that you used the latest model if their invoice is wrong.
Then 90 percent fail. Not because the technology was useless. Because they skipped the fundamentals.
This is the part that is boring and therefore ignored.
What actually succeeds is deeply unfashionable.
People who understand the technology, not just the API.
People who solve a real problem instead of wrapping a model around a landing page.
People who build maintainable systems that survive version two.
People who think about cost before the first viral tweet.
People who apply engineering principles that worked before the hype cycle and will work long after it.
Clean architecture still matters.
Proper error handling still matters.
Cost management matters even more when your variable expense is a model that charges per token.
Security matters when your system is literally generating output from user input.
Testing matters. Monitoring matters. Observability matters. Especially when the output is probabilistic and your users expect determinism.
And perhaps most controversial of all, talking to customers still matters.
You cannot prompt engineer your way out of building something nobody wants.
I am not saying AI is not transformative. It is. It changes what is possible. It changes how we work. It changes the leverage a small team can have.
What it does not change is the physics of building durable systems.
After twenty years of watching waves come and go, here is the lesson.
New technology does not obsolete good engineering.
It makes it more important.
Because when the barrier to building drops, the barrier to competing drops with it. When anyone can launch in a weekend, the only thing that differentiates you is depth. Discipline. Taste. Restraint.
The gold rush is always loud.
The people who last are usually quiet.
So the question is simple.
Are you building for the long term.
Or are you building for the demo.