Build a Spin-Off of a Startup That Already Makes Money (Replit Free Mode)
This is the exact setup from the video. You find a company that already makes money, you hand the agent its link, and it builds your version with the one feature they are missing. The prompt below needs nothing else from you. The last step is the one most people skip: the new feature becomes a free tool that markets the paid app.
Start with a startup that already makes money
Open the IdeaWave revenue leaderboard. It ranks real companies by monthly recurring revenue, and the rows carrying the seal are payment verified, so you are not guessing at somebody's screenshot.
Pick one you understand. In the video I picked an AI QR code generator sitting around $16.5K a month. Simple product, one job, plenty of room to do it better.
Copy its URL. That link is the only thing the build prompt needs from you.
Open Replit and start a conversation, in Free Mode
Open Replit and start a new app. In the agent, open the mode selector and pick Free Mode.
Now just talk to it. No spec, no setup, no planning document.
The build prompt
Paste the link where it says to, and send it. Nothing else to fill in.
The agent studies the site first, then replies with its decisions: the category, the one feature it is adding and why, the product name, the demo customer, the signature visual and the three pricing tiers. Then it builds without stopping to ask you anything. It finishes by testing its own work, publishing, and handing you the live link with a table of every decision it made.
Read its decision list before the build finishes. If you disagree with the feature it picked, say so in one line and it swaps it.
Build a spin-off of [PASTE THE WEBSITE LINK HERE].
WHAT YOU ARE GETTING FROM ME
One link. That is the whole brief. You make every other decision yourself, and you do not ask me questions before you start.
BEFORE YOU WRITE ANY CODE, REPLY WITH YOUR DECISIONS
Study the link first: what the product does, who pays for it, the one job it is hired for, the main action a user takes, the data it handles, the platforms it touches, and how it charges. Then reply with a short decision list and start building immediately after it, in the same turn:
1. The category and who this is for.
2. The one new feature you are adding, and why a paying user notices it in the first five minutes. Prefer a feature that creates its own data and can be shown working live on camera: tracking, analytics, monitoring, scoring, alerts, history, collaboration.
3. An original product name and a one line positioning sentence.
4. The fictional customer you will seed the demo with.
5. The signature visual you will build for the dashboard.
6. Three pricing tiers and what limits each one.
Copy nothing from the original: not the name, not the logo, not the copy, not the images, not the code. You are building your own product in the same category, with something theirs does not have.
SETUP
Use Replit's built-in database, and Replit's built-in AI integrations only where a feature genuinely needs AI. No sign-in and no payments yet. Publish on Replit, and build every shareable link from the published domain so links keep working after publishing.
THE CORE FEATURE MUST ACTUALLY WORK, THIS IS NOT NEGOTIABLE
- Implement the main action of the category end to end. A visitor must be able to do the real thing, not a mockup of it.
- Support every input type that makes sense here, each in its exact real format, and say in the UI which formats you accept.
- Validate every input, and when something is wrong say exactly what to fix in plain words.
- Build an automatic self check that performs the main action on a real example of each supported type, compares the result against what is expected, and reports pass or fail per type. Only show a "Works" badge when it passes. When it fails, block the export and say what to fix.
- Export or share the result in the formats this category actually expects, at a quality someone would really use.
THE NEW FEATURE, AND WHAT IT RECORDS
Decide the events your new feature records and the fields on each event, and implement them. Record only what the feature needs. Never store raw personal data such as IP addresses. When something cannot be measured, label it as unavailable instead of inventing it.
INTEGRATIONS AND LOGOS
Where the app refers to another platform, show that platform's official logo unaltered beside its name, and only for platforms the product genuinely touches. Skip this entirely if there are none.
DEMO WORKSPACE, PACKED WITH REAL LOOKING DATA
Seed a demo workspace for the fictional customer you chose.
- Every demo example uses real, working inputs, so the whole app works out of the box, and each one is editable from the dashboard.
- Choose a believable number of items for that business and spread them across the categories that matter in this domain.
- Generate a realistic image for each item with the image integration and use them as thumbnails.
- Ninety days of activity from a fixed random seed, so numbers never change between reloads. Nothing lands on a round number.
- Real patterns: a daily shape and a weekly shape that fit this category, a gentle upward trend, and items that started on different dates, marked on the chart.
- One spike, annotated on the chart with a plain sentence explaining what caused it.
- Realistic splits for whatever segments this domain cares about, adding up correctly.
- A small "Demo workspace" chip in the top bar.
- Real activity lands on top of the seeded data. When I perform the main action for real, it appears in the live feed within five seconds, marked "Just now".
EASY ENTRY TO THE DEMO
The landing hero's main button is "Open the live demo". It opens the demo instantly, no sign-up and no loading wall, and "Live demo" also sits in the top nav. Inside the demo, a slim bar names the fictional customer and offers a "Make your own" button.
PAGES
1. Landing: a headline about the outcome, one line under it, and two buttons, "Open the live demo" and the main action. Beside it, the signature visual and a real working example. Then what it does, a short preview built from real dashboard components, a pricing teaser and a footer. No testimonials, no customer logo wall, no invented user counts.
2. Dashboard overview: a filter row with a date range of seven, thirty and ninety days plus custom, the one filter that matters most in this domain, and compare to previous period. Four headline numbers you choose, each with a delta against the previous period. Then the main over time chart, the signature visual, a breakdown by category, a live feed, and a leaderboard of the main items with thumbnails.
3. A full screen live page, built to look good on camera.
4. A list page and a detail page for the main items, each detail with its own charts and inline editing.
5. A create page: the main action with a live preview, the "Works" check and export.
6. Pricing, not functional yet: every plan button opens the demo. Your three tiers, a monthly and yearly toggle with yearly at two months free, a full comparison table, four questions real customers would ask, and one honest line about what happens when someone cancels.
DESIGN SYSTEM, GO ALL THE WAY
- Dark by default. Start from these tokens and shift the hue to fit the brand you invented, keeping the same contrast relationships: page #120C22, panels and chart surfaces #1B1430, borders and gridlines #2E2447, text #F3EEFF, secondary text #A99FC4.
- One single hue ramp for every magnitude, dim to bright, six steps. Default: #9C470D, #B75F2C, #D17746, #ED905E, #FFAE85, #FED4BE. No decorative gradients anywhere.
- At most four category colors, used only for chart series, never for text. Default order: #E1306C, #4285F4, #059693, #D45701. More than four categories fold into "Other".
- Type: pick one distinctive display face for the hero headline and the headline numbers only, and one clean body face for everything else, both from Google Fonts, chosen to fit this category. Tabular figures in tables. Sentence case everywhere, no all caps labels, no monospace labels, no arrows on buttons.
- Radius hierarchy: 20px panels, 10px controls, full pill chips. No drop shadows. Depth comes from panels on the page color and 1px borders. Glow only on live data.
- Charts: one y axis per chart, 2px lines, bars with 4px rounded ends and 2px gaps, recessive gridlines, a legend when there are two or more series plus direct labels, tooltips with exact numbers on every point, bar and cell, and a table view toggle on every chart.
- Build one signature visual that belongs to this domain and would not fit any other product. A heat grid for scans, a pipeline for sales, a calendar for bookings, a map for deliveries, a funnel for signups, a streak grid for habits, a timeline for projects. On landing it animates once. In the dashboard, when real activity arrives, its matching part flares and the new row slides into the live feed. That is the only motion in the product. Respect reduced motion.
- Fully responsive with visible keyboard focus. The live page must look incredible on a phone.
HONESTY BARS
Demo data is always labeled as a demo. No testimonials, no customer logos and no user counts unless they are real. Pricing is marked as not functional yet. Anything the product cannot measure is labeled, never faked.
BUILD ORDER
1. The core feature working end to end, with the self check and export.
2. The new feature and everything it records.
3. The dashboard and the live page.
4. Landing, create and pricing.
WHEN YOU ARE DONE
Run the self check on a real example of each supported type and show me each result beside what was expected. Then perform one real action and confirm it reached the live feed. Publish, send me the live link, and finish with a short table of every decision you made: name, new feature, audience, events recorded, demo customer, colors, fonts, signature visual and pricing tiers.If you would rather pick the feature yourself
Send this first and the agent comes back with a list of spin-off ideas, each one a real feature the original is missing. Pick the one you would actually use, tell it which, then paste the build prompt above.
Worth adding while you are there: ask it to build a few design directions to choose from before it starts. Designs are cheap to change now and expensive to change later.
Give me a spin off idea of [PASTE THE WEBSITE LINK HERE] adding one new valuable feature.Then build the free tool that markets it
This is the part that turns a live link into traffic. You take the new feature, strip it to its smallest useful version, and give it away on its own page with no sign-up. People arrive for the free thing and meet the paid thing on the way out.
Same conversation, straight after the build.
Now turn the new feature into a free tool on its own page at /free, with no sign-up.
You decide the smallest version of it that is genuinely useful on its own, and that still shows off what the paid product does. A visitor should get a real, finished result in under a minute, not a teaser, and it has to pass the same automatic check as the paid product.
Give them a private link to watch their result update live for seven days. Under the result, one button: "Get the full dashboard". Add the free tool to the landing nav under a name you choose, and make the page fast, shareable and indexable with its own title and description.The free tool pages that bring the traffic in
In the video I went one step further and asked for a wall of small free tools, the kind people search for every day. Plenty of companies grow on exactly this, dozens of tiny utilities feeding one paid product.
The agent picks the tools, so this works whatever your category turned out to be.
Now build a set of free tool pages that people in this category already search for, the small utilities they open, use once and bookmark.
You choose which ones, between eight and twelve, based on what someone who needs the main product would search for on the way to needing it. Each tool gets its own page under /tools, works fully in the browser with no sign-up, and actually does the job rather than demonstrating it. Every page carries a short how-to in plain words, its own title and description, and one quiet link to the main product where it is genuinely relevant. Add a /tools index that lists them all, and link it from the landing nav and the footer.What I built with it
Heatscan, a dark QR analytics app. Codes for Instagram, TikTok, Google Reviews and WiFi, each one decoded automatically so a broken code never gets exported, plus a live feed where a scan from my phone shows up within five seconds.
The demo workspace is a made up cafe with twelve codes and about 38,000 scans over ninety days, seeded from a fixed random seed so the numbers never move. The pricing page is built and marked as not functional yet. Nothing on that page pretends to be a real customer.
Yours will not look like mine. Different link in, different category, different feature, different visual.
Keep it honest
Three rules I keep in every build of this kind, and they are written into the prompt so the agent holds them too.
Original everything. Name, logo, copy and images are yours. The only thing you take from the original is the problem it solves.
Label the demo. Seeded data is marked as a demo workspace. No testimonials you did not get, no customer logos you do not have, no user counts you made up.
Mark pricing honestly. If checkout is not wired up yet, the page says so and the buttons open the demo.
For the design side once it works, see the Claude Code design skills.
Get the next one first
New prompts every week.
Free. The new drops and the tools behind them, before they hit the feed.
No spam · New issues Sunday · Unsubscribe anytime
Need it custom?
Want this built for you?
Tell me the idea and I’ll build it. An app, a tool, an automation. You don’t need to be technical.
Related guides
Money
Rebuild Any App That Already Makes Money (Replit + Arrfounder)
Find a product with verified revenue, hand its URL to one Replit prompt, and get a better version back. Then the prompt that finds what the original is missing and builds it. All three prompts are inside.
Money
Clone a $46K/Month App With Replit and Flip It (My Exact Prompt)
Find an app already doing $46K a month, hand Replit one prompt with its URL, and get a real working rebuild plus comparison pages built to outrank the original. The exact prompt from the video.
Guide
Build a Viral Peekaboo Login Page With AI (Replit, No Code)
The cute login form where the characters cover their eyes when you type your password. Here are the exact prompts to build it in one shot.
Guide
The 5 Free Claude Code Skills That Make Your Front End Stop Looking Vibe Coded
Anthropic's own design skill, Hallmark's 57 slop checks, Vercel's guidelines, 10,000 pre-built components and a browser Claude can see through. Every install command on one page.
Frequently asked questions
Yes. The prompt tells the agent to study the site and decide the feature, the name, the data model, the demo customer, the colors, the fonts and the pricing tiers itself, then report what it chose. You only step in if you disagree with something on that list.
No. You paste the prompt and review what comes back. The part you bring is the choice of company, and the judgement about whether the result is any good.
Anything the agent can build and publish on the web. The prompt never names a category, it infers it from the link you give it, so a booking tool, a sales dashboard and a habit tracker all come out of the same prompt.
Yes. Send the short message in the section above first, read the ideas it comes back with, tell it which one you want, then paste the build prompt.
Building a better version of a product category is how most companies start. Do not copy their code, their brand, their copy or their images. Your name, your design, your build.
A free page with no sign-up gets found and shared, and everyone who uses it meets the paid product on the way out. It is the cheapest marketing a new app has.
