FIELD NOTES / 2024–2026 RESEARCHED OCT 15, 2026

Which viral AI play is actually copyable?

Seven revenue-backed cases point to the same answer: the model is rented, the product is narrow, and distribution is built into the output. Here’s the machinery—not the hype.

01 / The signal

Small teams, very large outcomes.

These are public or founder-reported figures, not a promise. Source quality varies, so every number keeps its caveat.

$30M+
Cal AI annual revenue in under two years; 15M+ downloads.
TECHCRUNCH / HIGH CONFIDENCE
117d
Chatbase to $1M ARR, entirely organic, according to its published story.
SUPABASE STORY
$132–138K
Photo AI monthly revenue in late 2025, from founder revenue posts.
FOUNDER-REPORTED
$10–15M
Arcads approximate ARR by 2026 with eight people.
SECONDARY / APPROXIMATE
~$200

The leanest launch proof: RizzGPT reportedly paid two TikTok creators about $50–100 each. Both videos went viral; reports cite 45K downloads on day one and 500K in a month.

MEDIUM-CONFIDENCE
CASE, NOT A GUARANTEE
02 / Route selector

Start with your distribution advantage.

Choose the channel you can realistically feed every week. The product format follows the channel—not the other way around.

What can you consistently do?

Tap the closest fit. This is a decision aid built from the three repeated patterns in the research.

PLAYBOOK 01

Consumer demo loop

Best fit for a watchable result people understand in under 15 seconds: a reply, score, scan, or visible transformation.

  • Thin mobile wrapper + API + paywall
  • Low-friction weekly or annual subscription
  • Seed small niche creators before paid ads
03 / Case files

Reverse-engineer the engine.

Filter by the part you want to study. “Wrapper” here means the business uses existing frontier models rather than training its own.

Consumer / mobile

Cal AI

$30M+
annual revenue

Photograph a meal; vision models and the phone’s depth sensor estimate calories and macros.

Price$30/year headline price; 3-day trial
GrowthTikTok fitness creators, systemized at scale
BuildMedium wrapper; no custom model training
Team7 people at acquisition

TechCrunch revenue report ↗ · RevenueCat creator playbook ↗

HIGH CONFIDENCE ON REVENUE
Consumer / mobile

RizzGPT + Umax

~$80K →
~$200K MRR

Dating-reply suggestions from chat screenshots; the companion Umax product scores selfies and suggests improvements.

PriceRizzGPT tiers unverified; Umax about $4/week
GrowthTwo paid TikTok creators; demo is the content
BuildSimple wrapper, built by a non-coder in ~1 month
SeedReportedly about $50–100 per creator

Growth case ↗ · Founder podcast ↗

MEDIUM CONFIDENCE / FOUNDER-REPORTED
B2B / SaaS

Chatbase

~$863K MRR
mid-2026

Upload documents or a site URL, then publish a trained support chatbot and embed it anywhere.

PriceFree; $40 / $150 / $500 monthly tiers
GrowthX demos and build-in-public; no paid marketing early
BuildSimple retrieval wrapper; 2 weeks to launch
Proof12,424 active subscriptions; Stripe-verified

TrustMRR metrics ↗ · Supabase launch story ↗

STRONGEST VERIFIED SAAS CASE
Consumer / transformation

Photo AI

$132–138K
MRR, late 2025

Upload 15–20 selfies and receive photorealistic photos in new settings from a personalized image model.

PriceOne-time photo packs / credits
GrowthFounder’s X audience + shared outputs
BuildVanilla web stack, one VPS, external GPU API
EconomicsAbout 87% margins; founder-reported

Indie Hackers deep dive ↗ · Case study ↗

FOUNDER-REPORTED REVENUE
Consumer / SEO

HeadshotPro

~$300K MRR
estimated

Turns a few selfies into professional LinkedIn and résumé headshots.

PriceOne-time $29–$59 packages
GrowthSEO/content + founder’s X audience
BuildSimple third-party image-model wrapper
TeamSolo, zero paid ads in year one

Single-source profile ↗

LOWER CONFIDENCE / SINGLE ESTIMATE
B2B / video

Arcads

~$10–15M
ARR by 2026

More than 1,000 AI actors deliver ad scripts in 30+ languages for e-commerce brands.

PriceUsage-based; some six-figure ACVs
GrowthOne viral tweet, then influencer virality
BuildMulti-model wrapper, automated ops
Team8 people; bootstrapped and profitable

Podcast notes ↗ · Agent-ops notes ↗

APPROXIMATE / SECONDARY SOURCES
Bonus / app builder

Base44

$1M ARR
in 3 weeks

A no-code app builder whose solo founder reportedly used Cursor to write about 90% of the code.

Price$20–$200/month subscriptions
GrowthLinkedIn, WhatsApp groups, hackathons, Product Hunt
BuildAI-built product; 95% reported margins
Outcome$80M cash acquisition by Wix, June 2025

Founder-story compilation ↗

SECONDARY RESEARCH COMPILATION
04 / Three repeatable plays

Copy the mechanics, not the product.

Each route begins with a channel-native promise, uses existing model APIs, and gets paid before technical complexity becomes the project.

01

Consumer demo loop

PAINFUL MOMENT → 15-SECOND DEMO → MICRO-CREATOR → LOW-FRICTION SUBSCRIPTION
  1. Pick a result that is visually obvious: reply, score, scan, or transformation.
  2. Build a mobile shell, API call, and paywall in 2–4 weeks; no custom model.
  3. Seed 5–20 small niche creators with a flat fee or revenue share before paid ads.
  4. Track revenue per 1,000 views by creator; scale only what converts.
  5. Optimize the first “magic moment” and paywall before adding reach.
02

B2B wow-demo

EXPENSIVE WORKFLOW → 60-SECOND PROOF → BUILD IN PUBLIC → FREEMIUM SAAS
  1. Choose a business pain where one demo makes the time saving undeniable.
  2. Build in 2–3 weeks on managed infrastructure; sell access to model output.
  3. Post demos, screenshots, failures, and revenue progress on X or LinkedIn.
  4. Start freemium or usage-based; add enterprise tiers only when customers pull for them.
  5. Automate repetitive operations before adding headcount.
03

Paid transformation

$100–500 SERVICE → INSTANT AI OUTPUT → ONE-TIME PACK → SHAREABLE RESULT + SEO
  1. Target an output people already pay for: headshots, résumés, designs, or mockups.
  2. Build upload → model API → gallery → download in 1–2 weeks.
  3. Use one-time $29–$59 packages or credits to match the tangible output.
  4. Launch with before/after proof; the transformation becomes the ad.
  5. Publish durable “[thing] AI generator” search pages while social spikes demand.
05 / The verdict

The leanest first bet is a demoable consumer utility.

Not because consumer apps are easy—they are not—but because this route has the clearest low-capital proof: a narrow wrapper, a watchable outcome, and creator seeding measured by conversion.

  1. Choose the output first. If someone cannot understand the result without narration, it is a weak short-form product.
  2. Choose one distribution loop. Short-form, build-in-public, or search. Don’t launch into all three.
  3. Charge for the first useful result. Subscription for repeat use; one-time payment for a tangible deliverable.
  4. Measure proof, not views. Cal AI’s system priced creator deals around revenue per 1,000 views.
06 / Source desk

Follow the receipts.

Sources were read through the web index on October 15, 2026. Where a figure comes from a founder, podcast note, or single secondary source, the page says so.

  1. TechCrunch — Cal AI revenue and acquisition
  2. RevenueCat SubClub — Cal AI creator system
  3. Alpha Leaders — TikTok-grown apps
  4. The Startup Series — Blake Anderson
  5. TrustMRR — Chatbase Stripe-verified metrics
  6. Supabase customer story — Chatbase launch
  7. Indie Hackers — Photo AI deep dive
  8. aituts — Photo AI case study
  9. Supabird — HeadshotPro profile
  10. Everand — Arcads founder podcast notes
  11. Everand — Arcads agent operations notes
  12. Founder-story compilation — Base44

Open questions retained: exact RizzGPT tiers; Cal AI retention and paywall metrics; independently verified HeadshotPro revenue; independently verified 2026 Arcads revenue; and current Photo AI revenue after late 2025.

Keep going