Table of Contents
What "validated" actually means
A validated MVP is one where strangers — not friends, not Twitter, not your mom — used the thing more than once without you reminding them. That's the bar. "Looks great!" is not validation. "I shared it with my team" is.
For 48 hours, your job is to engineer evidence of that behaviour, or evidence of its absence.
Write the kill/keep criteria before you start building. "If at least 5 of 20 invited users return on day 2 unprompted, keep going. Otherwise kill." Decide now while you're emotionally cheap.
Hour 0–6 · Pick the problem (not the idea)
Two-thirds of failed MVPs are perfectly built solutions to nobody's problem. Spend the first six hours making sure your problem has actual heat.
Open Twitter/X or Reddit search and look for the last 90 days of complaints about the pain you want to solve. Quote five of them. If you can't find five real complaints in 30 minutes, the pain isn't acute enough yet — pick a different one.
Then write one sentence: "I am building <thing> for <very specific person> who today does <painful workaround>." If you can't fill all three blanks crisply, you don't have a problem yet.
I am building <a deposit-tracking dashboard> for <freelance event photographers> who today do <copy-paste deposit amounts into a Google Sheet they update from their phone after every booking>.
When you can write one of these about your idea, the build phase becomes obvious. The deposit-dashboard founder built theirs in 4 hours.
Hour 6–18 · Build the thinnest possible version
Twelve hours, max. Pick the smallest slice that proves the value claim — usually one screen, one form, one outcome.
In Fabricate, the prompt looks like the one-sentence brief plus three explicit bullets: who logs in, what they see when they do, what they're allowed to do. Resist features. The MVP that wins is offensively small.
Every hour you spend on a feature is an hour you can't spend talking to users. The thinness of the MVP is the validation strategy.
Build a one-page web app for <freelance event photographers>. Sign-in via Google (Clerk). On login, show a single table of bookings: client name, event date, deposit amount, paid yes/no. Add an "Add booking" button that opens a modal with those four fields. Persist to D1. Deploy to a Fabricate subdomain.
That's it. No reports, no email reminders, no calendar sync. Whatever feature you almost added — that's a Hour-30 question, not a Hour-12 build.
Ready to Build?
Start building your full-stack application with Fabricate. Free tier available — no credit card required.
Start Building FreeHour 18–30 · Get 20 humans to try it
Twelve hours of distribution. Use the playbook from "Your First 10 Users" but at higher volume — DM 30, post twice, comment in three threads. Your goal: 20 strangers signed in by hour 30.
Be honest in your DMs that this is a 48-hour validation. People help when they know they're helping you decide whether to keep building.
Hey {{firstName}} — saw your tweet about {{specific complaint}}. I just shipped a 48-hour MVP for that exact thing. I'm trying to find out if anyone actually uses it before I commit another month. Mind trying it for 5 minutes? {{link}}. Genuine reply either way means a lot.The phrase 'genuine reply either way' converts ~2x better than asking for a one-sided positive opinion. People will tell you the truth if you signal you can take it.
Hour 30–44 · Watch what they actually do
You don't need a fancy analytics setup. You need three numbers tracked in any spreadsheet:
1. Of those who got the link, how many signed in?
2. Of those who signed in, how many performed the core action even once?
3. Of those, how many came back on day 2 without a nudge from you?
#3 is the only number that matters. #1 measures your DM, #2 measures your onboarding, but #3 measures the product.
Fabricate already logs basic events to PostHog if you set it up. If you didn't, just count manually — for 20 users it's faster than installing analytics.
Hour 44–48 · Kill or keep
Look at the criteria you wrote at hour zero. The number you got. Compare honestly.
If you keep: pick the single biggest friction the 20 users hit, fix only that, plan the next two-week build.
If you kill: write a 200-word post-mortem on what you learned about the audience, the problem, and yourself. That post-mortem is more valuable than the MVP — it tells you what to try next, faster.
Founders who kill cleanly hit a winner faster than founders who limp along. Each clean kill is one experiment closer to a real business.
Ready to Build?
Start building your full-stack application with Fabricate. Free tier available — no credit card required.
Start Building FreeThe single metric that matters
Day-2 unprompted return rate of new users. That's it.
3 of 20 = okay, dig deeper before you commit. 5 of 20 = strong signal, build more. 0 of 20 = the problem isn't real or your fix is wrong-shaped.
Everything else — likes, signups, "feedback" — is noise. Founders chasing those signals burn months. You're not.
What's next
You decided. Two follow-ons depending on which:
• Kept it? Read "Add Stripe in 30 Minutes" and put a price on it this week. Free MVPs distort signal — paying users tell you what's actually valuable.
• Killed it? Pick a different problem from the same audience and run another 48 hours. The audience research and the DMs you just did are still warm — reuse them.
Ready to Build?
Start building your full-stack application with Fabricate. Free tier available — no credit card required.
Start Building FreeFrequently Asked Questions
I don't have any audience to DM. Can I still do this?
Yes — the goal is 20 strangers who feel the problem, not 20 followers. Twitter/X search, Reddit, niche Discords. Six hours of homework finds rooms. See the "Your First 10 Users" playbook for the channels that work without an audience.
What if Fabricate can't build my idea in 12 hours?
Then your idea is too big for an MVP. Cut features until you have one screen, one core action, one outcome. Almost every Fabricate-built MVP fits in 4–8 hours of building if the brief is sharp.
Do I really need 20 users? Isn't 5 enough?
Five tells you whether your friends are nice, not whether the problem is real. Twenty is the smallest sample where the day-2 return rate stops being noise. If you can only get to 12, do 12 — but you're flying lower.
What if people sign in but don't do the core action?
That's an onboarding problem, not a product problem. Send the same 20 users a one-question DM: "What stopped you?" Answers are usually a missing tutorial step or unclear copy. Fix the friction and re-invite the same users for round two.
Is 48 hours wall-clock or working hours?
Wall-clock — start Friday morning, decide Sunday morning. The compressed timeline forces sharper decisions. Trying to do this across a normal work week tends to drag to 2 weeks; the urgency is part of why it works.