A non-technical seller built an Amazon research tool in 48 hours with Cursor and hit $20,000 a month in 90 days.
He had a corporate job and ran 2 Amazon brands on the side. He found Cursor through Twitter and built 10 to 12 projects with it. Most never reached production. The ones that failed were in spaces he didn’t know: AI video generators, automated job applications.
The one that worked came from his own frustration. Every product idea cost him 20 to 30 hours of research, copying data into Google Sheets.
The catch: zero audience, so zero distribution.
Then he remembered a coaching program he had bought 2 years earlier. It had thousands of active Amazon sellers. He wrote to them: give me 48 hours, and if you like what you see, we partner. No strings.
Hours 1 to 4: mapping their workflows. Hours 5 to 12: the core features in Cursor. Hours 13 to 20: testing and bugs. Hours 21 to 30: branding and polish. Hours 31 to 40: edge cases. Then a demo video.
The next day he woke up to a call: quit your job, we’re doing this full-time.
The price was equity. The split isn’t stated.
Day 0: launch, with the partnership. Day 30: $10,000 a month. Day 60: $17,000 to $18,000. Day 90: over $20,000. The coaching customers pay $50 a month.
His cost until it worked: the $200 Cursor plan.
What the story depends on: he knew Amazon from A to Z. On the failed projects he didn’t know what the user wanted.
He had no audience. He traded equity for 1 in 48 hours.
A non-technical founder's job board makes $25,000 a month with creators he pays $20 or less per video.
Roman built Career Hound, a job board that collects listings from company career pages. It is a crowded idea.
He is not a programmer. He sold alarm systems door to door for 5 summers in college and has a psychology degree.
Prices: $7 a week, $20 a month, $50 a year. No free plan.
Revenue: about $28,000 a month, with $12,000 of it recurring.
Customers: 8,000+ users and 779 active subscribers. 31% of them leave, because people cancel when they find a job.
Reach: 11,700,000 social views in the last 30 days, as of September 2026.
Here is what pays for it:
He picks creators in their 50s, because gray hair signals experience to job seekers. He pays $20 or less a video.
Each one gets a sheet of fixed video formats, so they need zero research, and he talks to them on WhatsApp.
1 format is a list of companies hiring remotely worldwide. It can be re-run with new companies every time.
It took him about 2 years of daily marketing tests to get here.
The job board is ordinary. The $20 videos are the business.
Full video: https://t.co/CmZL1zWadr
A quant trader cloned a hit app with AI, changed the niche, and made $5,000 in his first month.
David Attias, France. 8 years as a quant trader, helping traders pick stocks. No background in building apps. Until 6 months before the interview, a regular job.
He watched a podcast with 3 teenage founders whose app hit $200,000 a month after 3 months. He was envious. He downloaded it, loved the onboarding, and decided to copy it.
The twist: a different niche. Women who want to stop sugar cravings.
How he checked it was worth building: the Google Trends curve for “stop sugar” keywords had been rising for 5 years. TikTok and Instagram were full of women influencers talking about quitting sugar.
How he built it: screenshot every screen of the original. Then tell Cursor: make the same screen, same elements, same colors, same images. 2.5 weeks for a full copy. Another week to get Apple’s approval. About 1 month in total.
His numbers, 5 months in: $12,000 a month. About 60,000 downloads. About 900 paying customers. Profit margin 35%.
His customers are mostly Gen Z women, 13 to 25. He worked with French influencers, because nobody in France was doing it.
His monthly bill: Cursor’s top plan $200. Mixpanel about $100. TikTok ads about $100.
The host’s rule: copying an app is fine, copying its niche is not.
The app was a copy. The niche was his.
A 21-year-old college student built his first app in 1 week, and 5 months later it made $25,000 a month.
He had played guitar almost daily for 18 months. He asked ChatGPT how to set his gear to sound like a song, and the answers were wrong.
So he built ToneAdapt. It tells guitarists how to set their own gear to match any song's tone.
Thanksgiving 2025: he builds the first version over the break, in about 1 week, with AI coding tools. It was the first thing he had ever built.
December 8, 2025: he ships it.
About 4 months later: more than 100,000 users.
5 months in: $25,000 a month. He showed a live revenue dashboard during the interview.
The part under the number is the boring part: 3 videos a day.
He posts 3 times a day on Instagram, TikTok, YouTube and Facebook, with his face on camera. His friends mocked the first video. It went viral overnight.
His rule: once 1 video format works, keep posting that format until it stops working. And commit to 3 posts a day for at least a month.
1 problem he had. 3 posts a day. $25,000 a month.
Full video: https://t.co/q1XDo5C858
After 400 days of daily videos, a first-time app maker earned $10,000 on launch day and $400,000 a month a year later.
Yoni Smolyar built Brainrot, a screen-time app with a digital pet whose brain "rots" as you scroll.
It was his first iPhone app, after many failed projects and businesses.
About 400 days before launch: he starts posting daily videos about his journey.
Before launch: 100,000 to 150,000 followers on Instagram.
Launch day, summer 2025: $10,000.
30 days in: $26,000, with about 300 downloads a day, mostly from people searching the App Store.
About 1 year later: $400,000 a month, about 2,000,000 downloads and about 150,000 users.
Nobody screenshots the 400 days.
He says the "get users now, charge later" plan did not work in the App Store. He puts the paywall in from the start, and he tests onboarding text, trial length and pricing without waiting for Apple's review.
He also says the founders he admires launched every product to an audience that already existed.
400 days of videos. 1 launch day. $10,000.
Full video: https://t.co/4QXEkN8yAI
A broke founder who wasn’t a developer built an AI app in 8 weeks, and it now makes almost $100,000 a month.
Dustin Stout grew up in small-town Pennsylvania and moved to California to be an actor. He hated waiting tables, so he started a business. He taught himself web design and marketing, then built digital products. 2014: a WordPress plug-in company with partners. He left it. Two more products crashed and burned.
Then ChatGPT showed up. You couldn’t search your chats. You couldn’t put them in folders. And he figured people would soon need 5 or 6 AI subscriptions to get the best of everything. What if 1 app gave them all of it?
He had no money to hire developers. He found an online course and learned as fast as he could. 8 weeks later he had a working first version, built on a no-code tool plus a little code he understood. He launched it thinking nobody would buy.
Month 1: $3,000. Month 2: $4,000. Month 3: $10,000. After 1 year: $30,000 a month.
2.5 years in: over $1,000,000 in total revenue and almost $100,000 a month. 1 chat window, with GPT-5, Claude and the top image and video models inside it. $20 a month solo, $40 for a team of 5, no free plan.
The part the title leaves out: for 10 years he had run a personal blog. It had more than 100,000 email subscribers. He did no advertising. He also built an affiliate program that pays on every month a referred customer pays, so 12 payments means 12 commissions.
He built the app in 8 weeks. The audience took 10 years.
A developer with no audience got 1,800 signups before building anything, then hit $42,000 a month in 90 days.
CJ Zafir moved to Canada in 2018. His first 9-to-5 was a horrible experience. He knew he couldn't do any 9-to-5.
2022: he saw that AI coding was blowing up unnoticed. To learn it, he coded with AI tools 10 to 12 hours a day. He announced publicly that he would build 12 startups in 12 months. He built 11. 10 collapsed.
The idea came from his own work. He ran a small agency, and each project took 9 to 10 hours of chatting with ChatGPT to produce the technical documents AI coding tools need. He chained his prompts into 1 sequence and cut 7 of those 9 hours.
30 minutes: a landing page with Bolt. 2 weeks: 1,800 waitlist signups, off a screen recording of his ideal workflow. He messaged every person who commented and asked what they used to fix AI mistakes. Nobody had a better answer.
Only then did he build it. 10 to 12 hours planning the screens. 3 days designing the front end. A friend took the back end. 2 weeks from raw design to a live web app. 100 users on day 1, December 24, 2024.
3 months later: $42,000 a month, with a community of more than 4,800 AI developers. No paid marketing. All traffic from X. Price: $29 a month, no free trial, because the AI models cost real money.
His monthly bill: about $3,500. $2,800 of it is OpenAI.
His marketing: 4 threads and 3 long posts a week. Every one opens on a problem, shows the fix, then places his product as part of it. People bookmark them.
He calls it an overnight success. It came after 7 years.
Ten products collapsed. One had 1,800 people waiting.
A man who couldn't code built an app on his commute with AI, and it made $30,000 in 7 months.
At 18 he needed money for college. The only job he could get was door-to-door sales. One day he knocked on the wrong door and a stranger pulled a knife on him. He quit.
He tried everything else: web design, copywriting, sales funnels. At university he studied game development and was bad at code. He was visual, and he couldn't make sense of it.
By late 2024 he had tried several business models. None got him out of a 9-to-5. So he built a database of influencers, because it was something he would use himself.
How: Perplexity for the step-by-step plan. Bolt for a working first version. Cursor to connect the back end. Clerk for sign-in. A free Framer template for the design. Hosting and database on free plans. Margins about 90%. All of it on his daily commute.
Months of building before launch, with no idea it would pay.
Launch night: dinner with his girlfriend. He pulls out his phone and sees sale after sale. 20 to 30 that night.
7 months in: over 1,000 users, over 350 paid, $30,000 in revenue. On December 24 he quit his job.
What got the attention: 1 post on X with close to 500,000 impressions. He tied it to the debate everyone was having that week: is AI coding good enough to build a full product? A real demo, a raw reaction.
Copycats keep appearing, he says, and none take off. AI tools let anyone copy a product, so he says you need an edge a copy can't match.
He couldn't write the code. He knew what he would pay for.
Code is free, says a Google Labs director, and 8 photos became a walkable 3D Paris for $48.97.
Jaclyn Konzelmann is a product director at Google Labs. Before that she worked on Gemini API, AI Studio and Project Mariner.
In October 2026, Google shipped Gemini 4 Argon and then Gemini 4 Carbon, 2 weeks apart. Google says Carbon ranks #1 on Artificial Analysis' v5.0 index.
Her own blog shows what that means. A builder gave Argon 8 photos. The next morning he had a 3D model of the area around the Eiffel Tower that you can walk through in your browser, with day turning to night.
It cost him $48.97 and about 3.5 hours of AI helper time.
When building costs that little, she says the question changes. It stops being "can you build it?" and becomes "should it even exist?"
2 rules she keeps:
Rule 1: reading a doc should never take longer than writing it. AI can help you think, but you stay in charge, and every PM on her team must defend each statement without the AI.
Rule 2: Google Labs never treats a killed experiment as a failure. The team keeps the lessons, not the prototype.
For a beginner, that is 1 sentence to write before you build anything: why should this exist?
Building got cheap. Choosing what to build did not.
A father of 2 with a full-time job shipped 28 apps in 8 months and went from $200 to $10,000 a month.
He is an iOS engineer by day. About 8 years into his career, he had 1 pet project he had spent years trying to grow. Nothing worked.
February 2025: he watched a YouTube video that changed his plan. Build 1 simple app with 1 feature. Ship it. Forget it. Start the next one.
8 months later: 28 apps. $10,000 a month. Over 1,000 subscribers. 4,000 to 5,000 daily users across all apps.
Where the money sits: 4 apps make about $1,500 each. The other 24 make small amounts.
His filter: he builds only for a search term people type into the App Store, where the top apps already make at least $100 to $200 a month.
How he ships in hours: he asks ChatGPT or Gemini for a plan, drags in ready-made screens for settings, onboarding and paywall, and for some apps copies about 90% of the code from earlier ones. Cursor writes the rest. The record: 2 hours from idea to App Store review. Typical: a few hours to a week.
His monthly bill: OpenAI $200, Gemini $50, Cursor $20, a keyword tool $10, Firebase $5 to $10.
After launch, every app gets an App Store boost that fades. The ones that don’t sink get polish and ads. He lets the rest go.
He grew 1 app for years. Then he shipped 28 in 8 months.
2 college track athletes with no coding experience built an app that made $70,000 in 6 months.
They are 19 and 20, and they met in high school.
Before the app, they traded for a while and lost about $3,000.
The app helps sprinters train. It has workouts, nutrition, a community, and an AI that scores your running form from a video.
Built in about 2 weeks.
Month 1: $10,000.
Month 6: about $70,000 in total, 40,000 users, 4.7 stars in the App Store.
Today: over $20,000 a month.
It costs $10 a month or $30 a year. Their users are young kids, so they keep the price low.
The 2 weeks make the headline. The followers make the number.
They had built hundreds of thousands of followers in track and field with their videos, with no marketing. Before launch, they put the app's name in their videos, on their store's website and in their bio. Then they dropped the app.
It was not their only app. For the launch of their football app, they did not sleep for 4 days.
Trading lost them $3,000. Their audience made them $70,000.
Full video: https://t.co/XZq6VL0drw
A 19-year-old college dropout built a wrestling weight-cut app with AI, and it made over $20,000 in 1 month.
Before the app, he was a provincial judo champion and a national wrestling champion. Every competition meant cutting a lot of weight in a short time. He only learned to do it properly when 1 coach gave him a plan.
Summer 2024: he builds his first app before starting a computer engineering degree. It took 6 months and automated part of his mom's workflow at her job. Then a second app, built only with AI tools. That's when he decided to leave university.
First months of 2025: several apps, none marketed. He calls that a mistake. Then he finds Cursor and shifts to learning sales and marketing.
June 2025: Cut Coach. Built with Cursor and ChatGPT. An MVP in about 1 month, tested by his wrestling club.
The first version failed. The coach was supposed to hand out the plans, and the club didn't use it. In July and August he flipped the concept: the app gives the plan straight to the athlete. He tested the weight cuts on himself.
September 2025: launch, right as wrestling season started. 6 months later: $0 to over $60,000 in total revenue, about 39,000 downloads, and 1 month above $20,000.
The marketing: organic posts with 200 to 500 views brought 10 to 15 downloads a day. Then small creators with 1,000 to 10,000 views per video, later 20,000+. Then the best creator videos became paid ads.
His rule: solve a problem inside your own hobby.
Big apps serve everyone. His serves wrestlers cutting weight.
A teacher of 100,000+ AI builders makes his AI assistants prove their work in 4 steps.
Micky Shimeles has taught 100,000+ people to build with AI. Everything he ships to real users goes through the same 4 steps.
He says most setups where AI builds software nonstop turn out sloppy work.
The whole thing is 4 steps.
Step 1, isolate. Give each AI assistant its own copy of the project, so you can run several at once and none breaks another's work.
Step 2, build. Give the assistant a fixed structure to build on: clear inputs going in, clear results coming out. It writes to that shape every time.
Step 3, prove. Make the assistant record a video of itself testing the app. You compare the before and after of every change and watch the proof instead of taking its word.
Step 4, ship. A second AI, called Greptile, reviews the work. It keeps testing and improving until the score is 5 out of 5 with 0 open comments. Then Micky usually reviews it himself.
Each step is a saved instruction in his free collection, so he runs the same routine every time.
Before: 1 assistant says "done" and you take its word.
After: 1 video shows the change, and 1 reviewer scores it 5 out of 5.
The AI says it works. The video shows it.
A 23-year-old with no computer science background built an app in under 2 weeks with AI and grew it to $20,000 a month.
Before apps, he was a kid who played too many video games and stayed up too late. One night he woke up with an app idea and wrote it in his notes. He had never coded.
He had no MacBook and no money. He sold random items from his room until he got an old one. His first app, a social one, took 6 to 8 months and kind of failed.
He moved to utility apps, which are easier to scale alone. About 3 years later he says his apps have done over $1,000,000 in revenue.
August 2025: he launches Payout, an app that lists class action settlements you may qualify for and prepares the claim form to mail in. Per the host, it went from $0 to $20,000 a month in 50 days. It passed 12,000 downloads. It won first place and the $65,000 grand prize in a RevenueCat hackathon with over 55,000 entrants, for the most growth over 2 months.
How he built it with Claude: download about 20 apps in the niche. Screenshot every onboarding screen. Lay them in 1 long row in a design file. Pick the best parts and redesign them. Write a text document describing the app's data, with sample data attached. Drop the screenshots into Claude, which writes the code for each full screen. His rule: don't be picky. Get a version out.
Why he spends as long on onboarding as on the main features: if an app has a paywall, he says, about 90% of users will only ever see the onboarding.
How it spread: 1 creator in the niche brought thousands of downloads. He copied the format with other creators, then turned the best videos into ads.
He sold things from his room for a MacBook. Now Claude writes the code.
A founder who sold his last startup for $4,000,000 now builds 5 products alone with AI agents.
Josh Pigford has been a solo builder for 25 years.
2026: 5 products at once, no team, AI agents doing the building.
His first rule is to ship early. He says months of work before anyone can use the product is a bad idea.
Launching still scares him every time. After 25 years, the fear is still "what if zero people care?"
The $4,000,000 is the headline. The checking is the work.
He builds with 1 AI model, then has a different one review the result. GPT finds 3 to 5 mistakes that Opus missed.
He also keeps a saved instruction called "but for real." It tells the AI to check its own work again before it changes anything. Each run turns up 3 to 5 more mistakes the AI missed.
A third saved instruction runs at the end of each stage. It turns every "no, that did not work" into a new rule, so the AI stops repeating the same mistake.
The AI writes the code. 25 years of shipping checks it.
Full video: https://t.co/BbOfjyNtKS
Exa grew about 10x a year for 3 years to $100,000,000 in annual revenue by selling search to AI agents, not people.
Marcus Holm, its chief revenue officer, joined about 6 months ago from LaunchDarkly, where he was president.
His one-line pitch: what Google is for humans, Exa is for AI agents. Agents don't open Google. They need a connection built for software.
The company had about 60 employees when he arrived. Now it's almost 200. It was almost all engineers. Sales and marketing are now bigger than engineering. A $250,000,000 round at a $2,200,000,000 valuation, led by a16z, came in May 2026.
His claim on the record: in 2026, agents passed humans in search volume. Over half of web traffic is now agents and bots, he says, citing Cloudflare. Exa's CEO had predicted it. Marcus: he nailed it.
Underneath the numbers: ads. Search makes its money from ads, and websites fight for rankings. For an agent, he says, that is clutter and clickbait that can lead it to make things up. Exa charges for usage instead.
What agents ask for keeps shifting. 6 months ago: coding agents looking up documentation. Then agents researching people and companies for sales teams. Then analysts doing due diligence. In the last month or two: personal agents booking trips and restaurants.
A product launched about a month ago: price history. What a product cost at Target 6 months ago, or 2 years ago. Hedge funds love it, he says.
Google earns from ads. Exa earns from every search an agent runs.
Full episode and transcript: https://t.co/WetXE27y3L
Brex's CEO runs a $5,000,000,000 business and tells builders to stop making AI agents.
Pedro Franceschi, co-founder of Brex, says "agent" is the wrong word. He wants virtual employees, not "open-ended agents that could do a million things."
Every AI employee gets 4 things:
1A job: 1 clear outcome to own.
2Skills: written instructions for how to do it.
3A manager: a person it goes to when things are unclear.
4A budget: spending tied to what the outcome is worth.
His example is Jim, Brex's AI recruiter.
Jim finds job candidates, screens
applicants, and scores them against set criteria.
Brex has engineers build with AI during job interviews. Pedro says he spends half his time reviewing work.
The risk is real. AI employees can leak company data, and his team built a tool called CrabTrap to stop that. He also warns that a big bill for AI usage is coming, and shows how to control it.
For a beginner, that is 4 sentences written before the first prompt.
An agent can do a million things. An employee does 1 job.
Full video: https://t.co/AN7uNxtyfM
Ethan Bloch, who co-founded Digit and now leads ChatGPT Finances at OpenAI, shows 6 ways to save money.
Digit, his last company, helped users save more than $9,000,000,000.
ChatGPT Finances connects to your bank and investment accounts. As of October 2, 2026, OpenAI is rolling it out to free users in the US.
His 6 uses, in the order he shows them:
Use 1: cancel or negotiate subscriptions.
Use 2: find tax savings before tax time.
Use 3: cut the fees on investment funds.
Use 4: get more out of credit card rewards.
Use 5: model a pay cut or time off.
Use 6: decide whether to rent or buy.
Use 1 starts with a plain-English request.
Look back 90 days and flag subscriptions, price hikes, double charges, fees you can get back, and bills that look too high.
Not everyone is sold. In July, AI consultant Allie K. Miller wrote that most people do not need it in its current state.
She said it answers questions about your money more than it acts on them.
The bank shows the balance. The AI shows the leaks.
Full video: https://t.co/jRNt2JDyx1
A man from Daytona Beach who had never heard of SaaS watched a company go from $800,000 to $30,000,000 in 100 days.
Chris McDonough grew up in a small town known for NASCAR and spring break. His family owned a printing company. The plan was that he'd take it over.
2017: he texted a college acquaintance on Snapchat who had moved to San Francisco. A recruiter set up interviews. A couple of weeks later he had an offer from Motive, then called KeepTruckin. Employee number 40. Revenue: $800,000.
Two start dates: in 2 weeks or in 6. He picked 2. That weekend he signed a lease on a living room with a partition and a loft bed. Everyone told him he was insane.
Truckers were about to be legally required to buy the product. For months nobody did. They believed the new administration would kill the mandate. May to September: almost nothing sold.
Then the deadline got real. 70+ demos a week, calendar blocks just to eat. About 800 to 900 demos in 3 months. $800,000 to $30,000,000 in about 100 days. About $100,000,000 by the time he left.
His hiring class was promoted in 30 days. The class that started 4 weeks later waited 18 months. In those 18 months he went from SDR to interim director running managers in Nashville.
This year he left Rippling, 5,000 employees. His reason: he feared he wasn't building AI skills fast enough to stay employable. At that size, he says, you can't switch AI on for 1,200 sellers at once without burning $50,000,000 in tokens by accident.
He joined Campfire, an AI-native ERP. His bet: AI widens the gap between reps. A core 30% worth triple everyone else, and a bottom 20% who become hard to employ.
He took the two-week start date. Then he took the scarier job.