Build & launch your own software.
Get your first customers.

Who’s teaching you this. 25+ products, ~$330k ARR, zero funding raised — all public on TrustMRR. Why I stopped spending months learning to code, and why I’m finally documenting the new way of building.

The old way was six months learning to code. The new way ships from day one. What vibe coding actually is — you own how the app should work, AI handles the code — and why we build real revenue products, not toy projects or portfolios.

What you get: recorded modules from idea to shipped product, my Claude Code setup + prompt library, payments & deployment, a private community, lifetime updates, and walkthroughs of my own products. Who this is for — and who should skip it.

Total beginner? Perfect. Zero to a working Claude Code setup — subscription, installing Node.js the next-next-next way, CLI running in your terminal. I promise you won’t learn a single word of code. You need this before Module 3.

My own full walkthrough on a Mac. Google “Claude Code”, grab the one-line CLI command, paste it into Terminal, then run “claude” and sign in with /login. Why I still prefer the terminal over the desktop app, and why I pick a Claude subscription (Pro/Max) over the pay-as-you-go API. Start to finish, no code.

This is the piece that lets Claude actually click inside your browser for you. Open Chrome, add the Claude extension from the Chrome Web Store and pin it, then run “claude” and “/chrome” to reconnect it. I show it live, plus the allow/deny prompts and how to stop them. (Recorded for the Meta Ads course, so the live demo happens in Ads Manager — the setup is identical for everything we build here.)

Get started the cheap way. Why I use BuildCheap — you add money and see the exact ₹ each generation costs, instead of shaky “credits.” Then we build a real portfolio: search a GitHub portfolio site, import & clone it for ~₹0, and start making it yours. No VS Code, no setup, no code.

Make it yours and ship it. Drop in your LinkedIn info, swap in your photo — and watch the per-message cost fall to ₹0 by picking the right model (Kimi vs Claude). Then deploy to a free buildcheap.host domain and rename your URL. Post your live link in the course community when it’s up.

Before we build anything, how this actually started. In 2021 I made a Chrome extension that pulled the question-answers out of Facebook group member requests — and apps are still charging monthly for that exact thing today. The console hacking I used to test it, and why I'm happily handing you the product that made me thousands of dollars.

Open the terminal — that's where everything starts. I name the app MagicGroup, have Claude create the folder inside software/, and let it scaffold a Chrome extension from a plain-English description. Also: install Whisper Flow. I talk to Claude instead of typing, and it changes how fast you move.

The scope: pull membership-request questions, push them into a Google Sheet through an Apps Script call. Why I cut every extra feature people asked me for — an MVP is not a smaller version of everything. Then a plain-English explanation of what an API and JSON actually are. No code.

The extension exists. Now we point it at a real group with real pending requests — fixing the group settings so questions actually get asked, sending a join request from another account, and watching what the code does and doesn't see.

Getting the backend: the Apps Script code I paste into script.new, the API URL, and the sheet it writes to. Also why I let AI pick Cursor vs VS Code for me — AI should reduce the number of things you have to think about, so you can spend that on the right thing.

The Chrome Web Store developer fee is $5, one time, unlimited extensions. Then the Apps Script side: paste the spreadsheet ID, name the sheet, deploy as a web app, execute as me, access to anyone — and what that last setting actually means.

chrome://extensions, developer mode, load unpacked, pin it. It's live in my browser. I change the button colours so a user can tell it's working, and switch to asking for the Google Sheet URL instead of the ID — because nobody knows what an ID is.

It assumed every question ends with a question mark. They don't. This is the most important lesson in the module: when you vibe code, go and check. I send it back to look at the real page, it finds its own flaw and fixes it. Plus replicating Facebook's own UI so the extension doesn't look bolted on.

Grok, Cursor's agent, the desktop apps. What I actually use and why: the terminal is a direct line to the operating system, everything else puts a layer between you and the machine. Try them all — this is why I stay.

Default the sheet name to Sheet1 instead of teaching people to rename it. Take a URL, not an ID. Every time you're about to educate your user, ask whether you could just handle it instead. Then redeploy — and how versioning works in Apps Script.

People run multiple groups and don't want the leads mixed together. Adding a Google Sheet URL per group, then running the whole thing for real: request comes in, I approve, and the lead is in the sheet before Facebook even finishes.

I've built Google, Apple and Microsoft login. Email OTP is still the simplest thing that works. A second Apps Script for sending mail, the 100-emails-a-day Gmail limit, and why a Workspace account lifts that to thousands.

Starting the landing page while the code runs. Five logo options, pick one. Then payments: Gumroad to start, PayPal if you're selling internationally. And personalize everything — my font, my name, my face in the footer.

Ask people to reply — a reply is the strongest signal there is. Offer a free month for feedback shared on social. Say plainly that this isn't a recurring email so nobody reaches for the spam button. Small things, but this is the difference between reaching your customers and not.

Supabase, Cloudflare, or the thing I actually used for years: Google Sheets. Every Chrome extension I've built used a sheet as the database. I create the users sheet, paste the ID, redeploy, and login has somewhere to store people.

A checkpoint. Extension works, login works, user profiles work. What's left is the landing page — I build it in Next.js inside a project I already had. Also why a free plan makes selling easier than going paid-only.

Remove a member, send a fresh join request, answer the questions, approve — and watch the lead land. Then the store listing: the tabs you fill in, and the privacy policy and terms of service you can't skip.

The page needs a centered title, a short line under it, and a visual mockup immediately — nobody should scroll to find out what it does. Then the comparison section: the copy-pasting they do today, next to the one click they'd do instead.

The title has to land for a cold lead who will never watch your video. What I've taken from Alex Becker, Russell Brunson's DotCom Secrets and Hormozi on offers — and why targeting only the people already doing this shrinks your market to almost nothing.

It stopped doing a good job — that's the context filling up, not the model getting dumber. I copy the state into a fresh terminal and carry on. Don't grind one session forever; it eats your limits and your output quality with them.

One primary CTA, one secondary link under it. Paste screenshots into Claude with Ctrl+V, not Cmd+V. I draw a rectangle straight onto the screenshot to show exactly where the mockup goes — same shape sizes, arrows, a glow moving through the flow. Understandable without reading.

We're not selling a scraper. We're selling time, reliability and consistency. The moment the page put "hours" next to "two minutes", it started selling. People shouldn't have to read the whole thing to get it.

A navbar just for this page, with its sections on it — when you're selling, help people jump to what they care about. Then an about section with my picture and why I built it: I grew groups to thousands of members, and doing this by hand was slow and unreliable.

The standalone HTML landing page came out better than the one inside my existing project — build your landing page separately so it doesn't inherit another page's style. Then getting it live on Netlify for free, start to finish, sign-in detour included.

The user sheet ID was blank in one place. Fixing it, saving the Apps Script files, and watching the site go live. Also my domain-buying addiction, 50+ domains in, and why you shouldn't copy that one.

The deployment said authentication required when it should have been anyone. Checking doPost, running the function once to clear the permission prompt, then logging in for real — and reading the console when Facebook starts throwing errors at you.

Paste the code, verify, and the login is done. Then testing the popup properly: free plan, 25 leads, the user landing in the sheet. The plan didn't sync — and why the upgrade path is the one thing in an extension you can't afford to make people wait for.

The unglamorous part nobody films: waiting. What I do while it's thinking, why watching the output matters even though you can't read code, and the honest cost of opening a new terminal instead of staying in the one you've got.

A free tier complicates everything downstream. You pay, you use; you don't, you don't. I change my mind on camera, simplify it, push the new Apps Script version, and redeploy.

Auth works, payment works, the pro plan works. Now getting it store-ready: minifying (47% smaller — I used to do this by hand on uglifyjs), and the privacy policy page the Web Store won't let you skip.

My site has too many payment options, which makes it a bad example. So here's the simple version: a PayPal button. Why it's the fastest way to take money internationally, and the doc trick I used for years to publish policy pages without building any.

The developer dashboard, version 1.0.0, and the assets it demands: 512×512 icon, screenshots, the lot. This is the stage I used to get stuck on for hours, missing one requirement at a time.

A waitlist that doesn't save is worse than no waitlist. I check whether the emails are really landing in the sheet, then test it live with a real address and watch the row appear.

Wiring PayPal IDs in as script properties instead of pasting them into code. Then generating the store images — I genuinely didn't expect what came back.

Short description, long description, category, single-purpose statement, and a justification for every permission you request. Reviewers read these literally. This is where submissions die, and it's the part every tutorial skips.

The promo tile I've never bothered with on extensions that made thousands — and why that's changing. Then adding your own domain: the DNS records on Namecheap, and confirming it resolves.

The privacy policy URL 404s, which means the site isn't actually deployed. Fixing it, and how Netlify redeploys off a connected GitHub repo on every change.

What git actually is, explained the way I understand it: I used to copy my code somewhere before changing it so I could get back. Then merging, a green build, and the landing page live.

Everything's in: auth, payment, landing page, listing. Now the submit button, and the fear around it. Worst case is a rejection and you try again — after three to four hours, the extension is out the door.

The whole build as it actually happened, 11 PM to 2:18 AM: idea research to a real iPhone app submitted to the App Store — "Waiting for Review" on camera. Watch this when you want the honest, unedited rhythm of a build night; the numbered lessons below are the same footage cut by topic.

Name the app so people find it. Ahrefs Keyword Explorer, keyword difficulty vs search volume, spotting branded vs unbranded terms — and the programmatic-SEO idea (recipe categories in your sitemap) that falls out of the research.

Check the domain the smart way (and when NOT to buy one). Then one prompt creates both projects — a Next.js landing page and a Flutter app — with zero code written.

Describe the product like a PM: two pages, what each shows, how the Instagram share flow works — and why Apify handles the reel transcription so you never touch a scraper.

Why both projects open in Cursor, why Git matters even if you never code (my CA friend uses it), and starting Claude from a clean page.

Install Xcode from the App Store, let Claude set up the emulator, hit shift+tab for plan mode — and the exact logo prompt that returns five options rendered in Chrome.

Firebase vs Supabase vs Cloudflare, honestly — from someone with millions of users through Supabase. Why Cloudflare wins for agents, plus the Wrangler CLI and Node install.

The four decisions that shape the app: cloud storage over on-device (the programmatic-SEO reason), OTP login over Google/Apple sign-in, Apps Script over a paid email service, and where the backend lives.

Five logo versions rendered in Chrome, one mixture request, and a decision. Favicon to app icon from a single prompt — work people used to pay lakhs for.

A waitlist landing page while the app builds: Apps Script API into a Google Sheet with email, source and date — so you can start collecting leads today.

The 2-minute foundation: what an API actually is, what JSON looks like, and what a curl command does. Explained for someone who has never coded.

The anti-vibe-code design pass: Outfit font, no gradients, no orange — a shade of green for a food brand. Watch the same page go from AI-made to real.

Reading Claude’s plan like a founder: the database as a Google Sheet, why it added a request queue I never built for apps with 1M installs — then the site deploys live with Wrangler.

Before showing anything, Claude tests its own backend: health APIs, a real creator’s reel through Apify, screenshots to verify. Plus the two keys you need — Anthropic and Apify.

Deploy the OTP email API from your own Gmail, connect an iPhone by cable, and watch the app run on real hardware — plus the $99 Apple developer account.

The honest part: builds fail, Xcode stalls, two builds collide. How to rename terminals, when to close them to save cost, and why an hour still beats the months this used to take.

“Can’t reach servers” on login — debugged across two terminals. Why a fresh terminal with fresh context finds the misconfigured secret the long-context one misses.

Claude flags the security hole itself: anyone with the URL could send email as you. Script properties, a shared secret, and Wrangler-managed secrets — without writing code.

Buying easyyrecipes.com on Namecheap (with GST), pasting the whole checkout as a screenshot for a sanity check, and Claude reading the confirmation from Gmail.

It’s 1 AM — so Claude logs into Namecheap in Chrome and moves the domain to Cloudflare itself, while the login bug gets fixed in parallel.

Pricing thinking out loud (margin vs cost, import limits), a 3-day free trial — and why you push to GitHub the moment something works.

Claude creates the App Store listing: bundle ID, name, keywords, description, copyright — and turns two raw phone screenshots into store-ready marketing images.

One prompt, three parallel agents: screenshot design, the rest of App Store Connect, and a pricing/cost audit of the codebase. Toggling between sub-agents like a team lead.

Hands literally off the keyboard while Claude answers Apple’s entire age-rating and privacy questionnaire — then the marketing plan: daily reels, and free subs for influencers who post.

A secret full-access account for Apple’s reviewers (Claude’s idea), Pro Monthly and Pro Yearly — and why yearly gets a real discount: churn.

The unglamorous last mile: builds processing on Apple’s side, JavaScript form-fills failing, and Claude switching to typing into App Store Connect key by key.

The paywall reveal, “Waiting for Review” at 2:18 AM — and the full recap: idea → research → app → landing page → waitlist → domain → paid plans → submitted.