Nov 2025$295k/mo
The plateau. 30% monthly churn caps growth no matter what acquisition does. He stops scaling and spends a month only on retention — the case study Module 9 is built on.
source @andrewxroas · Nov 2025Built from 432 dated posts · every claim cited on screen
Distribution
is the product.
You can build the app in a week — that part is solved. What you don't have is the machine that puts it in front of a million impulse buyers. This course is that machine, reverse-engineered from a founder who documented his climb from $250k to $1.12M a month in public, and taught with the receipts on screen.
38 lessons — all rendered
10 modules
≈5h runtime
source @andrewxroas
Lesson 9 teaches hard paywalls with no free trial. We follow our own syllabus: one lesson free — L20 · THE VOLUME MATH — the rest paid.
The receipts
One founder. One channel. Nine months, in public.
Nothing on this page is a projection. Every number below comes from a dated public post by @andrewxroas, quoted verbatim and cited on the exact frame it's used in the course. This is the run the playbook is built from.
Dec 2025$300k/mo
Through the churn crisis. The funnel holds: ad → long emotional onboarding → hard paywall. The paywall that converts 60% better turns out to be a copy of a 7-figure app's paywall.
source @andrewxroas · Dec 2025Jan 2026$500k/mo
The creative engine compounds. 600 ads in five months, $5k and 100 ads before the first winner. "You can be wrong 99% of the time."
source @andrewxroas · Jan 2026Mar 2026$600k/mo
Four people. One broad CBO. No test campaigns, no scale campaigns — Meta scales winners in place. "$700k/mo with terrible tech debt."
source @andrewxroas · Mar 2026May 2026$1.12M/mo
Profit while scaling. Net margin climbs from 31% to 52% — $325k profit on $722k gross over four weeks, with ad spend, tax, Stripe fees, AI APIs and AWS all modelled in Module 9.
source @andrewxroas · May 2026Summaries above are ours; the course quotes the underlying posts verbatim, dated, on screen. His numbers, not ours — that is both the point and the caveat. More in the FAQ.
The doctrine · Module 1
The whole operating system in one breath
Distribution is the product below $1M a month. Build a B2C AI web app — no Apple tax. Clone demand that is already proven. Sell the transformation behind a hard paywall. Win by out-producing everyone on ad creative through one broad CBO. Product work is churn work, and churn work starts when acquisition plateaus.
- Competition is validation, not a warning. The ads library tells you the money is there.
- $10k MRR is a milestone, not a goal. Impulse-buy consumers scale past any indie ceiling.
- The product is not the business. The machine that sells it is — and it can be built deliberately.
The syllabus
Ten modules. Colour tells you what kind of truth you're getting.
ice — principle
gold — mechanics & numbers
red — failure modes & open problems
M1The thesis
Why B2C AI web apps are the gold rush, why web beats mobile, and the indie-hacker deprogramming.
M2Finding the idea
The burner-IG method, validating the money in the ads library, the hard-paywall filter, the positioning gap.
M3Offer, pricing, paywall
Sell the transformation, not the feature. Hard paywall, no trial. Decoy pricing. The >80% gross-margin rule.
M4Build it in a week
Minimum feature set with Claude Code, where "tech debt is a rounding error" stops being true — and why design still matters.
M5Funnel architecture
The canonical funnel and its benchmarks — 1.2% outbound CTR, 4% landing CR — weekly A/B testing as a moat, event tracking done right.
M6The creative enginecore
Concept vs variant. The volume math: 100 ads before the first winner. Why ugly ads print. Hiring your editor.
M7AI UGC productionversioned
The unlimited-length workflow that beat HeyGen. Voice is the tell, not video. Custom avatars for the audience. Re-recorded when the stack rots.
M8Media buying
One CBO, broad, forever. Min-spend to force learning. Scaling rules. Diagnosing a bad week from which metric actually moved.
M9Retention & the real P&L
Churn is the ceiling — taught as the open problem it still is. Payment recovery. Profit at scale, every line item modelled.
M10Scaling the operation
The 4-person company. Entity and jurisdiction arbitrage. The honest treatment of one-channel concentration risk.
The format
Taught like an engineer's brief. Honest like a lab notebook.
"B2C apps = ecom minus fulfilment."
- Motion graphics, no talking head.
Funnels, CBO trees and P&L waterfalls drawn as diagrams, timed to the sentence that explains them. Dense material wants pictures, not a face.
- A disclosed synthetic voice.
The course itself teaches that voice is the tell in AI content — so we don't pretend ours is live. It's consistent across all 38 lessons and says so.
- Attribution is structural.
Every borrowed claim carries a dated source chip on the same frame. Where the corpus gives a rule but no reason, the extension is marked MY INFERENCE — teaching a doctrine, never laundering one.
Read before buying
What this course does not teach
-
Churn, solved.
He never solved it — 30% churn is what capped him at $295k. Module 9 teaches his instrumentation and the honest state of the problem, not a fix that doesn't exist.
-
Organic growth.
His TikTok, Google and X experiments were all inconclusive. This is a paid-Meta playbook, and the concentration risk of that is Lesson 38, not a footnote.
-
Our own $1M screenshot.
We built the machine he described — we haven't run it to $1M. The receipts on this page are his, dated and cited. If that made you pause, good instinct. It's exactly why every claim on screen carries its source.
The paywall · per lessons 9 & 10
Hard paywall. No trial. Its own doctrine, applied.
The playbook · one-time
$199 USD
- All 38 lessons, ≈5 hours — rendered, not promised
- Lifetime updates, including Module 7 re-versions as the AI stack rots
- The source archive — the 60 long-form posts, organised and dated
- The benchmark sheet — funnel numbers to hit at each stage
- Captions and transcripts for every lesson
Why no free trial? Lesson 9: AI products have real unit costs, and free trials bleed you at traffic volume. A course is no different. One complete lesson — L20 · THE VOLUME MATH — is free instead, and it's the most quoted lesson in the syllabus.
Why one price? Lesson 10's decoy pricing is for subscriptions. This isn't one. You buy it once; updates are included because stale lessons are the one thing we re-record on principle.
Refunds: 14 days, no questions. A hard paywall you can walk back out of.
Questions
Asked before you could ask
Isn't this just his tweets, repackaged?
No — and the difference is structural, not cosmetic. The corpus is our syllabus research: 432 posts read, dated, and organised into an operating system he never wrote down in one place. The structure, diagrams, worked examples and extensions are original. Every borrowed number is attributed on the frame it appears, and everywhere we extend the doctrine it's labelled MY INFERENCE. The alternative — teaching his system without his name — is the thing we refused to build.
Who is @andrewxroas?
A B2C app founder who documented his run from $250k to $1.12M/mo — Nov 2025 to Jul 2026 — in public, with real account structures, real benchmarks and real P&L. We have no affiliation with him; he doesn't sell a course and is on record saying so. His posts are public, quoted under attribution, and linked from inside the course.
Have you run this playbook yourselves?
We run a portfolio of products in public and apply this doctrine to it — that's why the course exists. But we won't dress our numbers up as his: the $1.12M run is his, cited; our own results appear only where they're real. No fabricated screenshots, no fabricated students. That rule is in the production notes and it doesn't bend.
Will the AI tooling module go stale?
Yes — his own stack changed three times in nine months, and the course dates every change. Module 7 teaches the selection criteria rather than worshipping tool names, and it's versioned: when the stack rots, we re-record it, archive the old version with a date, and you get the update free.
Do I need a big budget to use this?
You need what the math says: his documented cost of finding a first winning ad was about $5k and 100 creatives. Lesson 20 walks the numbers honestly, and lesson 11's margin rule tells you when running paid ads is a mistake at any budget. If your product can't pass those two lessons, the course will tell you to keep your money — that's a feature.