Day 5 — Deployment, the Magic Moment

Part of the 7-day "ship a real app this week" series. Day 5 of YOU building a $Bn company, solo.

TL;DR (the 5 moves — do them in this order)

Deploy the frontend (60 seconds — connect the GitHub repo to Vercel, click Deploy, get a live *.vercel.app URL)
Fix the localhost trap (5 minutes — your backend has to go live too, or the site loads but nothing works)
Turn on analytics (2 minutes — so you can see your first real visitor)
Skip the domain — for now (0 minutes — don't buy a .com until you have something worth showing)
Send the link to one human (1 minute — the actual magic moment)

Total time from "works on my machine" to "works on the internet, for everyone": ~10 minutes.

The thing nobody tells you: localhost:3000 is not a website. It's a place only you can visit. Your app isn't real until it has a URL someone else can type — and the part that trips people up isn't the deploy, it's that the backend is still on your laptop.

Why this order matters

By now you have a working app — frontend, backend, auth, payments, all version-controlled. The instinct is to polish it forever before showing anyone. That instinct is the enemy.

Deploy reveals the truth. The bugs that only appear in production — the backend that isn't wired up, the env var you forgot — stay invisible until the thing is actually live. So you deploy early and ugly, find the real problems, and fix them. A perfect app on localhost is worth nothing. A rough app on a real URL is a company.

So: get it live first, perfect it after. The point of Day 5 isn't a flawless launch. It's crossing the line from "private project" to "public thing."

The bigger shift (the actual reason this works)

Deployment used to be the scary final boss — SSH into a server, configure nginx, set up CI, pray. Founders hired DevOps engineers just to put a site online.

In 2026 the deploy is a single click, because Vercel watches your GitHub repo and ships every push for you. The skill isn't deploying anymore. It's understanding the one concept the click hides: your frontend and your backend are two different things, and both have to leave your laptop. Get that, and deployment is solved forever.

Step 1 · Deploy the frontend (60 seconds)

If you followed Day 1, this already happens automatically on every push. If you're starting clean:

Go to vercel.com and sign in with GitHub.
Click Add New → Project, pick the repo you've been building all week.
Click Deploy. Wait ~60 seconds.

You get a live URL like your-app-abc123.vercel.app. Open it. The thing you built is on the internet.

Prefer to stay in the chat? Tell Claude Code: "deploy this to Vercel" — it runs the same flow and prints the URL.

Gotcha: the first deploy can fail on a build error (a missing dependency, a type error). Don't panic — paste the error to Claude: "Vercel build failed with this: [paste]." It fixes the code, you push, the next deploy goes green. The previous working deploy stays live the whole time.

Step 2 · Fix the localhost trap (5 minutes — the part nobody warns you about)

Here's the afternoon everyone loses. You deploy, you open the live URL, the page loads and looks perfect — and then nothing works. Login fails. Data doesn't save. Why?

Because your backend is still running on your laptop. During the week you ran a local server at localhost:3000 (or your database pointed at a local connection). Your live site has no idea what localhost is — localhost is a place only your computer can visit. So the frontend went live, but it's reaching for a backend that isn't there.

The fix is the same connect-and-click idea:

Put the backend on a host too. If you used Supabase (Day 3), it's already live in the cloud — you just need to point your deployed site at it. If you're running a separate server, deploy it to Railway or Render (connect the repo, click deploy — same as Vercel).
Swap every localhost for the live address. Tell Claude: "my app is deployed but the backend calls still point at localhost — move every API base URL and database connection to environment variables, and set the production values to my live Supabase/Railway address."
Set environment variables in Vercel. Your .env lives on your laptop; the host can't see it. In Vercel → Project → Settings → Environment Variables, add the same keys (Supabase URL + anon key, Stripe keys, etc.). Then redeploy.

Now it works for everyone, not just you. This one beat — the backend has to go live too — is the single most valuable thing on deploy day.

Common gotchas:

"loads but login/data is broken in production, fine locally" → it's almost always a missing env var on the host. Re-check Vercel's Environment Variables against your local .env.
CORS error in the console → tell Claude "add the production domain to my backend's allowed origins."
Stripe works locally, not live → you're likely still on test keys, or the webhook still points at localhost. Update the webhook URL to your live domain and use live keys when you're ready to take real money.

Step 3 · Turn on analytics (2 minutes)

You can't feel the magic moment if you can't see it. Add the lightest possible analytics so you know when a real person opens your link.

Vercel Analytics — one toggle in your Vercel project, or tell Claude "add Vercel Web Analytics to my app." Zero config, privacy-friendly, shows visits + top pages.
That's enough for Day 5. You don't need Google Analytics, a dashboard, or funnels yet. You need a single number: did anyone show up.

Gotcha: analytics needs one more deploy to start collecting. Push, wait for the green deploy, then send your link — so the first visit actually counts.

Step 4 · Skip the domain — for now (0 minutes)

You will be tempted to buy yourapp.com before you show anyone. Don't.

A *.vercel.app URL is a real, shareable, HTTPS website. It's enough.
The magic was never the .com. It's the first time a stranger opens the thing you built.
Buy a domain when you have something worth putting a name on — usually after a few real users, not before. When you do, it's a 2-minute add in Vercel → Domains, and Claude can walk you through the DNS.

Spending an hour on domain shopping is a very comfortable way to avoid the scary part: hitting send.

Step 5 · Send the link to one human (1 minute — the actual magic moment)

This is the whole point of the week so far. Text the live link to one real person — a friend, your group chat, someone who'll actually open it.

Something changes in your head that day. You stop being "someone who's learning to build," and you become a person who ships. The deploy was the easy part; being seen is the unlock.

Then do it publicly: drop your *.vercel.app link in the comments below the Day 5 reel. I open every single one.

Gotcha: the urge to "wait until it's perfect" will be strongest right here. It is never perfect. Send the link anyway. Feedback from one real opener beats another hour of polishing nobody asked for.

Common Day 5 gotchas (the real ones nobody warns you about)

The "it works on my machine" classic — of course it does; your machine has the backend running on it. Production doesn't. The fix is always: backend on a host + env vars in Vercel. (Step 2.)
The "silent env var" bug — the #1 cause of "live site loads but does nothing." Your .env never leaves your laptop. Every key in it has to be re-entered in the host's settings. Diff them line by line.
The "I'll just buy the domain first" stall — domain shopping is procrastination cosplaying as progress. Ship on the .vercel.app URL today; name it later.
The "my deploy broke the site" fear — it can't. Vercel keeps the last successful deploy live and only swaps in the new one if its build passes. A broken push fails quietly; your URL stays up.
The "Stripe is in test mode" surprise — totally fine for Day 5. Don't flip to live keys until you actually intend to charge real cards. Test mode lets you demo the full flow safely.
The "nobody will care" voice — you only need one person to open it. Not a launch. Not Product Hunt. One human, one link, today.

What Day 6 covers

Tomorrow is ship day — the whole week, end to end. You build one real, complete app (a to-do list with a real login that saves your stuff, deployed) — and the trick is you build it by prompting an AI. You open Claude Code (or Codex), describe what you want in plain English, and it writes the app, wires the backend, and pushes it live. Next.js + Supabase + Vercel, three pieces, one live app. Then you send the link and you're done with Week 1.

Day 6 hook: "week 1 ends today, and it ends with you shipping a real app — login, database, deployed. yes, today. yes, you."

Action item preview: ship a to-do app (login, save, deployed) and DM me the live URL.

Day-by-day series map

day	topic
Day 1	The live URL first (Claude Code + Pro + Vercel)	done
Day 2	Frontend, demystified (Next.js + Tailwind + taste)	done
Day 3	Backend, demystified (Supabase: db + auth + storage)	done
Day 4	Git is your time machine (branches, commits, undo)	done
Day 5	Deployment, the magic moment (custom domain + analytics)	← you are here
Day 6	Ship day: end-to-end, real users, real URL	tomorrow

Resources

Vercel: vercel.com — sign in with GitHub, deploy in one click
Vercel Analytics: vercel.com/docs/analytics — one toggle
Railway: railway.app / Render: render.com — for a separate backend, same connect-and-deploy idea
Supabase (already cloud-hosted from Day 3): supabase.com
Environment variables in Vercel: vercel.com/docs/projects/environment-variables
A real example from me: implexa.ai deploys on every push to main — green in ~48 seconds. Yours will too.
Want help? Drop your live URL in the Day 5 reel comments — I open every one.

Get the next one in your DMs tomorrow: follow @ImplexaAI — Day 6 drops at 9am PT.

implexa.ai