Highlights
- A weekend is enough to build a working MVP with AI: Friday to scope it, Saturday to build it, Sunday to deploy it and put it in front of five users.
- Range Tracker's first version took one weekend. As of August 2026, it was a multi-tenant SaaS with its first dozen trial users.
- What ships in 48 hours: a working records app (a tracker or a booking system), a landing page with a waitlist, a client-facing dashboard, or an investor demo. What doesn't: enterprise-grade auth, payments at scale, or an app store listing.
- Scope to one user, one problem, and one workflow, and save everything else for Monday.
- Watch five testers use it, and you'll see where real users stall.
Bolt.new lets you build the app by describing what your business does.
The first version of Range Tracker took one weekend. Its founder built it to track their own progress at the shooting range, and by August 2026 it was a multi-tenant SaaS with its first dozen trial users. A weekend is long enough to build something real, and short enough that a dead idea dies cheap.
How long does it take to build an MVP (minimum viable product) with AI? A weekend, if you scope it right: one evening to cut the idea down, one day to build it on Bolt.new, and one day to make it honest with real data, a live URL, and five people using it who aren't you.
This table shows the weekend at a glance:
| Day | The job | Done when |
|---|---|---|
| Friday evening | Scope it to one user, one problem, one workflow | You can say all three in three sentences |
| Saturday | Describe it, build it, fix it, load real data | The workflow runs end to end with real records |
| Sunday | Publish it and put it in front of five people | You've watched at least one of them use it |
| Monday | Decide what the five showed you | You've picked: build more, fix the stall, or stop |
What can you ship in 48 hours?
A working version of one workflow. That covers more ground than it sounds like:
- A CRUD app: a tracker, a booking system, a client portal, an inventory tool. Something that creates, stores, and shows records.
- A landing page with a working waitlist, connected to a real database instead of a promise.
- A client-facing dashboard that reads live data and looks like a product.
- An investor demo. VISION-LINK, an assistive-AI venture in Nigeria, built its proof-of-concept dashboard in a few hours and pitched funders with a live prototype instead of a deck.
And here's what doesn't ship in 48 hours, so you don't spend Saturday fighting it: enterprise-grade authentication and compliance, payment processing at scale, native apps in the app stores, and anything that needs another company's approval process. Those are month-two problems.
How do you scope an MVP on Friday evening?
Scope an MVP to one user, one problem, and one workflow, and write each as a single sentence before you describe anything to Bolt.new. The playbook starts with subtraction:
- Who the one user is. A person, not a market: "wedding photographers who lose track of client shot lists," not "the events industry."
- The one problem you're solving for them, in their words.
- The one workflow that solves it: what goes in, what happens, what comes out.
Everything that didn't fit in those three sentences goes on a Monday list, and the discipline matters more than the tool. An MVP with four features is a prototype of four ideas, and you won't learn which one worked.
How do you build the MVP on Saturday?
Build it by describing that one workflow to Bolt.new, fixing what comes back, and loading real data, in this order:
- Describe the workflow to Bolt.new the way you'd explain it to a new employee: the data, the screens, who touches what. Your Friday sentences are the first prompt, and our guide to better prompting covers how to write the rest.
- Get a working version before lunch. The first build won't be right, and that's the process working. Correct it by describing what's wrong, the same way you'd redline a draft.
- Load real data. Ten real records expose what tidy examples hide: the field you forgot, the status that doesn't exist, the name that breaks the layout.
- Work in saves. Jai S., who built the diabetes app Ginger Pulse on Bolt.new, plans features before building them and branches his project before big changes, so an experiment that fails costs minutes. Steal the habit: plan in Plan mode, and add one feature at a time.
Stop when the workflow runs end to end with real data. Resist Sunday-morning feature ideas; they're Monday-list material.
How do you test the MVP with users on Sunday?
Publish the MVP to a live link and watch five people who have the problem use it:
- Publish it. Bolt.new deploys to a live URL you can send to anyone; a custom domain can wait. Run the quick checks in our pre-launch testing guide first.
- Send it to five people who have the problem. Skip the friends being nice; pick people who felt the pain in your Friday sentence.
- Watch at least one of them use it. Where they stall is your roadmap. What they ignore is your delete key.
- Write down what they say in their words. You'll want the exact phrasing when you rewrite the landing page.
Five is enough because you're watching them work. Jakob Nielsen's research found that testing with 5 users uncovers about 85% of usability problems (Nielsen Norman Group, 2000). By the fifth user, the stalls repeat.
What do you do on Monday?
Decide, with evidence you didn't have on Friday. Three outcomes, all wins:
- They used it and asked for more: build the Monday list in order of how often it came up.
- They used it and stalled: fix the stall, not the feature list, and run the Sunday test again.
- They didn't care: kill it. The idea cost you a weekend and a month of a software subscription.
The dev-shop route puts a different price on this same experiment: custom software projects through a development shop average $132,480 and about 13 months (Clutch, 2026). At that price you commit before you learn. At a weekend, you learn before you commit.
Start building, free: bolt.new. See what other owners shipped: 7 small businesses that stopped renting software and started building it.
