Taylor Otwell posted this and I agree with it, probably for different reasons than most of the replies though.
A lot of the replies are sarcastic - “could you build Laravel Cloud in a day?” - and no, obviously not. But that’s missing the point. You could get something real going in a day, and that’s where you should start. It’s not a promise, it’s just a way of stopping yourself from overcomplicating something before you’ve even written a line.
I’ve also seen people read it as a green light for uncritically vibe coding - throw a prompt at an AI, hope it works, ship it. That’s not it either.
The reason building something meaningful in a day is actually possible now is that the models are good, but they’re only good when you give them something to work with. AI has largely taken over boilerplate and syntax. What it can’t do is define what the system is supposed to do, what the constraints are, or why you’re building it that way - that’s still on you.
One of the biggest shifts is that being able to write a decent spec matters far more than it used to - and you can use AI to help write that too. These aren’t things every project needs, but they’re examples of the kind of artefacts that give an AI enough to work with:
- an OpenAPI schema so the AI knows what the API should do
- types and interfaces so it knows what the data looks like
- acceptance criteria and business rules so it knows what “correct” actually means
- tests so there’s some way to know if it’s working
- a rough idea of how it’s getting deployed and why
When you turn up with that, the AI moves fast. When you don’t, you get code that looks fine and does the wrong thing.
The bigger decisions still need someone who knows what they’re doing. AI will happily have a go at almost anything, but it doesn’t reliably make the right architectural trade-offs on its own - how the API is shaped, where it lives, what the failure modes are, which constraints actually matter. It can implement decisions, it just can’t always be trusted to make them.
Engineering effort is shifting away from implementation and towards design and intent. The people getting the most out of it are the ones who can clearly define what they want before they start - and that turns out to be harder than it sounds.