Building the MVP: the loop works end to end

The build technique comes from the vibe coding course (Lovable/Replit, iterations, rollbacks). Here we cover what separates a product from a hobby build: the order you build in, and two layers beginners always forget.

Skeleton before beauty

Goal one: the whole loop runs end to end, however clumsily — sign up → do the thing → get the result. Only then do you polish individual steps. Why that order: the scariest problems are the seams ("step 3 is impossible without data from step 5"), and they only show up when the loop runs all the way through. Better to catch them on day one than to discover, after a week of buffing the landing page, that the product can't be completed at all.

This rule is older than web development. While making the first Toy Story, Pixar cut a rough version of the entire film out of sketches — the "reels" — and only once they were sure the story held from beginning to end did they spend months rendering frames. Whole thing rough first, details beautiful second. Software works the same way.

A Socratic pause

You've opened the product for the first time. What's on that first screen? Not a full task list, not a populated feed — emptiness. And that's where it's decided whether you stay or close the tab. Think: what needs to be on that empty screen for you to know what to do?

Layer 1: empty states

A new user has zero data: empty list, empty history. A hobby build shows bare emptiness; a product leads: "Nothing here yet. Upload your first one — here's the button." What to ask the AI for: "For every list and screen, build an empty state: what this is, what it's for, and a button for the first action". First impressions live exactly here — on the screen hobby builds leave blank.

Layer 2: humane errors

Wrong file, the wifi blinked, the payment failed. A hobby build says nothing, or frightens people with "Error 0x80070057". A product explains and offers a way out: "This file is over 100 MB — compress it or split it up." The formula for a good message: what happened + what to do. The ask: "Every error in plain language: what happened and what to do. And show me a list of all the error messages" — that last part lets you edit every text at once, in one voice.

Beginner myths

• "Errors are edge cases, I'll write them later." One in three people hits an error on their very first visit: wrong format, slipped mouse, flaky network. That's not the edge, that's the middle of the experience.
• "It's clear to me, so it's clear to everyone." You know the product from the inside — you're a blind tester. It has to be clear to someone seeing the screen for the first time.
• "I'll make it pretty, then check the logic." Other way round: working loop first, beauty second. A gorgeous broken product is an expensive picture.

The "someone else's hands" test

You know this trick from earlier courses — now it's a required acceptance step. Someone from your target audience walks the loop in front of you, in silence, and you don't hint and don't explain — you just watch where they trip. In "Don't Make Me Think", Steve Krug showed that 3–5 real people are enough to find most problems; you don't need an expensive lab. Three stumbles = three tasks for tomorrow. Keep polishing the loop until a fresh person gets through it without a single question — that's what MVP-ready means.

Where this goes wrong: polishing the wrong thing

The classic MVP build failure looks like this: the founder spends a week perfecting a button animation and a landing page gradient, and the test user trips at the front door — they don't know where to click after signing up. Skin polished, skeleton broken. The order rule exists precisely for this: until a fresh person can walk the loop end to end, polishing any single screen is premature. First "works from start to finish", then "beautiful at every step". The reverse order is the most expensive way to spend a week.

How this links to other courses

Iterating, rolling back to a working version and working with an AI assistant is technique from the vibe coding course. And the empty states and humane errors you're building here directly drive conversion to first value — a metric from this course's analytics lesson. It's all connected: what you build now is what you'll measure later.

Do this now (5 minutes)

Open your product draft and find a new user's first screen. Does the empty state lead them to a first action, or is it just empty? If it's just empty, write the AI ask for empty states right now, before you forget.