Hours in, nothing moved
I told a friend of mine, a marketer, to build his site himself in Lovable. He did. A good-looking one, in a single afternoon, without a line of code. Up to that point everything went exactly as I'd promised.
Then he wanted to connect it to his own domain. He opened his hosting provider's support chat and asked why the site wasn't working. He got answers. Correct ones. Several hours later he was in exactly the same place.
When we finally sat down on a shared screen, he named it better than I could have:
Since I don't have that basic knowledge — that these certificates even have to exist — I didn't know what to look for, and on top of that: where to look.
Forty minutes and it worked. He wasn't stuck for lack of an answer. He was stuck for lack of a question. And to be honest, part of that was on me: I recommended the tool and never said what comes the day after.
What the gap actually is
To ask a good question you already have to know something about the subject. And here, three things stood in the way whose existence he had no idea about:
- The SSL certificate (that padlock next to the address) — that it's a separate piece and has to be switched on by hand.
- The second panel — that the domain settings live somewhere other than the panel he logs into every day.
- The „verify” button — that after adding the settings you still have to go back to the tool and tell it to check.
The chat would have answered every one of those without blinking. Had they been asked. They weren't, because asking them required knowing the thing existed in the first place. That's not a gap in knowledge about answers. It's a gap in the map of the terrain.
What AI won't do for you
A language model is reactive by default: it gets a question and answers it. It's genuinely good at that. Far more rarely does it do the thing that actually saves you here — stop and say: „hold on, describe the whole situation from the start, because you're asking about a symptom, not a cause”.
On top of that, we describe the symptom („the site doesn't work”) rather than the context („I built the site in Lovable, the domain is with a Polish registrar, I added an A record”). Without context an answer can be correct and completely useless at the same time. The same mechanism ruins prompts when building prototypes: the less context you give, the more elegantly you get something other than what you meant.
What to do about it
Four things that genuinely shorten this kind of dead end. The order is deliberate.
- 1
Describe the situation, not the symptom
Instead of „the site doesn't work”, write what you built, where the domain is, what you've already clicked and exactly what you see on screen. One extra message upfront can save an hour.
- 2
Ask for a map, not a fix
„List every step that has to happen for this to work — including the ones people forget.” That question drags out the things you didn't know existed.
- 3
Set yourself a time limit
If nothing moves after an hour, the next two usually won't either. Being stuck for lack of a question isn't solved by rephrasing the same question.
- 4
Ask someone who has been there
Not so they'll do it for you. So they'll tell you where to look. That's usually a matter of minutes, not a project.
And this isn't only about domains. The same pattern returns with integrations, with invoicing, with every rollout where a non-technical person meets a tool built for a technical one. AI has massively lowered the barrier to building. It hasn't lowered the barrier to knowing what to look for — and that's the part someone still has to spell out for you.
How long can you sit with something before you give up and ask an actual person?
Guide
Lovable and a custom domain — a step-by-step guide
Related
From a call to a prototype in 24 hours
Mateusz Kozłowski
Founder of flowbiz · Process automation expert
I implement automations, integrations and AI in mid-sized companies across Pomerania and Kuyavia-Pomerania.
