Tech brands
30 min read
We broke down exactly what 100 fast-growing AI, consumer and B2B brands did to grow. This study covers 27 deep teardowns, what every other brand did, and the 12 GTM plays they keep repeating.
We went through 100 of the fastest-growing AI, consumer and B2B brands with a public growth number from late 2022 to Nov 2026. For each one we looked at what actually drove growth: where the first users came from, what the launch looked like, which engine scaled it and what came after. Numbers come from founder interviews, podcasts, company posts, press and trackers, and estimates are labeled.
The report has six parts: the headline findings, what winners did at each stage, the 12 plays they repeat, the numbers worth knowing, 27 deep teardowns, and what the other 73 did. Skip to the plays if you want the steps, or to the teardowns if you want the detail.
Headline findings
Where the first users came from
Existing audience (6): Lovable, Bolt.new, Replit, Claude Code, Duolingo, Whop
Niche community (6): Clay, Linear, Granola, ElevenLabs, Midjourney, Partiful
Launch moment (4): Cursor, Manus, Perplexity, Gamma
Paid creators (4): Fyxer, Higgsfield, Cal AI, Coconote
Design partner customer (2): Harvey, Sierra
Marketplace hack (1): HeyGen
Competitor’s bad news (1): Monarch Money
Outbound sales (1): Ramp
Not disclosed (2): Ladder, Opal
What winners did in months 0-6, 6-18 and 18+
Stage | What most winners did | Examples |
|---|---|---|
Before launch: first 100 to 1,000 users | Hand-picked users from a niche they could reach in person or in group chats, onboarded them manually, and used a waitlist to filter for fit. Some tested demand by selling the output by hand first. | Linear picked about 10 users a week; Granola onboarded about 150; Clay ran 8 reverse demos a day; HeyGen sold on Fiverr; Cal AI paid one creator $2,000. |
Months 0-6: the launch | One concentrated moment: a founder tweet, a demo video of a complete job, a Product Hunt launch or a relaunch to an existing audience, ideally the week a new model made the product possible. | Bolt’s single tweet ($4M ARR in 4 weeks); Manus’s demo (2M waitlist in 7 days); Lovable’s relaunch ($1M ARR in week one); Gamma’s AI relaunch (60,000 to 3M users in 3 months). |
Months 6-18: the engine | Turned what worked organically into a repeatable, paid-for engine: creators on retainer, partnership ads, experiments every day, or a community program for power users. | Gamma’s 1,000+ micro-influencers after $10M ARR; Cal AI’s 250 creators then ad-specific creatives; Fyxer’s 514 experiments; Clay’s expert network. |
Months 18+: expansion | Moved up-market (teams, enterprise, seats to workflows), added distribution partners and bundles, and used brand stunts as ‘surround sound’. | Harvey’s seat expansion; ElevenLabs’ enterprise reps and forward-deployed engineers; Perplexity’s Airtel bundle; Ramp’s Super Bowl ads. |
The 12 repeatable GTM plays
These are the moves that did the most work across the 27 deep dives. Most brands ran two or three of them at once.
The artifact loop. Every output carries the brand and pulls in the next user: Gamma’s ‘Made with Gamma’ badge, HeyGen’s watermark, Partiful’s invite pages, CapCut’s template links, Midjourney’s public channels.
Creators paid on views. Pay for results, not posts: Cal AI paid for view thresholds, RIZZ paid $1 to $2 per 1,000 views, and Whop pays clippers per 1,000 verified views.
Own the creator. Build the face in-house so the channel cannot walk away: Ladder’s full-time creators, Opal’s ‘Olivia Unplugged’ account, Duolingo’s tiny TikTok team.
Founder posts the numbers. The founder ships in public and posts users, revenue and outages: Photo AI, Base44, Rork, Lovable, beehiiv. Cluely shows the risk when the numbers are wrong.
Hardest customer first. Win one logo everyone respects, then sell with it: Harvey with Allen & Overy, Legora with Mannheimer Swartling, Sierra with $1B+ consumer brands.
Atomic network wedge. Win one tight group completely before the next: Gas one school at a time, Fizz and Saturn campus by campus, Partiful friend group by friend group, Granola through VCs, Clay through cold email agencies.
Ride a capability jump. Launch or relaunch the week a new model makes the product work: Bolt.new after Claude 3.5 Sonnet, Gamma’s AI relaunch, Replit Agent, Genspark’s Super Agent.
A free tier people recommend. Give away enough that users sell it for you: ElevenLabs’ free tier, Cursor’s expensable Pro plan, OpenEvidence for verified clinicians, MagicSchool for teachers, Opal’s move to freemium.
Do the switching for them. Remove the cost of leaving the old tool: Cursor forked VS Code, Monarch rebuilt its Mint importer, Threads imported your Instagram graph, Ramp switches customers over.
Stunt with activation. A stunt only counts if it gets people using the product: Perplexity’s Super Bowl contest required 5 questions, Hallow’s Super Bowl ad opened a 40-day Lent challenge.
Borrow distribution. Ride an audience someone else built: Perplexity’s Airtel bundle, Meta AI inside WhatsApp and Instagram, Supabase inside Lovable and Bolt, Kalshi through Robinhood, Abridge inside Epic.
Price on outcomes. Charge for the result the buyer wants: Sierra per resolved conversation, Decagon per conversation, Whop brands per verified view.
Ramps worth posting
Brand | The ramp | Source |
|---|---|---|
Gamma | $0 to $10M ARR nearly 100% word of mouth; $10M to $50M over 50% word of mouth, the rest influencers and referral; $100M profitably with about 50 people. | Grant Lee |
Lovable | $1M ARR in week one, $4M in 4 weeks, $10M in 2 months with 15 people, $100M in 8 months, $600M run rate by Sep 2026. | Lenny’s, TechCrunch |
Replit | About $2.83M ARR for years, then $10M (end 2024) to $100M (Jun 2025) after Agent. | Amjad Masad |
Opal | Hard paywall to $5M ARR; freemium cut conversion from 20% to 9%; then 1M DAU and $10M ARR. | Kenneth Schlenker |
Growth levers across all 100
By category: Consumer subscription apps split between short-form/UGC (6) and paid UA (5). Vertical AI is almost all sales-led (5 of 8). AI coding is the most product-led category (6 of 11) and the most founder-posted (3 of 11).
Consumer vs work tools
Consumer means consumer apps, social, AI creative and commerce (45 brands). Work tools means AI coding, B2B SaaS and vertical AI (32 brands). Even so, work tools borrow consumer tactics: Gamma and Fyxer both ran creators on LinkedIn.
Time to milestone
Fastest consumer launches
Growth speed
Deep teardowns: 27 brands
Each teardown covers what the brand did, the lever that did the most work, and where the numbers come from. (M) marks a lever call or detail with medium confidence, inferred from coverage rather than stated by the company.
1. Cursor: Zero switching cost, then a year of product
Forked VS Code so developers kept every extension, theme and shortcut and could switch in minutes. An early social push in 2023 filled a waitlist, then the founders went ‘monk mode’ for about a year and let developer word of mouth and a busy Discord do the work. Custom models for tab completion and code edits sat under frontier models, and a $20 a month Pro plan was cheap enough to expense, so individuals adopted first and teams standardized later.
Main lever: PLG/viral loop
Sources: forbes.com, businessinsider.com
2. Lovable: Open-source audience, relaunch, everyone posts
Started as GPT Engineer, an open-source project launched with one tweeted video that passed 50,000 GitHub stars, then relaunched as Lovable on Nov 21 2024 to that audience. It hit $1M ARR in its first week and $10M in 2 months with 15 people. Founder and employees all post on X and LinkedIn, users are pushed to share what they build, and it co-marketed with Supabase, Replicate and Resend.
Main lever: Founder-led social (M)
Sources: techcrunch.com, lennysnewsletter.com
3. Bolt.new: A dying company relaunched with one tweet
StackBlitz was stuck at $0.7M ARR after 7 years and weeks from shutting down. It hard-pivoted on Jul 1 2024 after Claude 3.5 Sonnet made its shelved prototype work, and launched Bolt.new on Oct 3 2024 with a single founder tweet. Its in-browser WebContainers ran full-stack apps with no setup, so the first prompt produced a live app people shared.
Main lever: PR/launch (M)
Sources: growthunhinged.com
4. Replit: Eight years of users, then an agent that finishes the job
Replit had millions of users since 2016 but revenue sat around $2.83M ARR for years. Replit Agent (early access Sep 2024) writes the code, sets up the database and deploys, so non-coders can ship. It repositioned from developers to PMs, designers and founders, and subscribers grew 45% a month.
Main lever: PLG/viral loop (M)
Sources: growthunhinged.com
5. Claude Code: Dogfooding to a bottom-up enterprise wedge
A side project by Boris Cherny that 20% of Anthropic engineers used on its first internal day and 50% by day five. Released as a research preview in Feb 2025, it was not an instant hit; general availability in May 2025 with new models and inclusion in Claude subscriptions made it take off. Engineers tried it on weekends and pulled it into their companies.
Main lever: PLG/viral loop
Sources: anthropic.com, newsletter.pragmaticengineer.com
6. Manus: One demo video, an invite wall, resold codes
Launched Mar 6 2025 as invite-only with a demo video of the agent doing real tasks; 2M people joined the waitlist in 7 days and invite codes were resold for the equivalent of thousands of dollars. Shareable task replays and a team plan followed. It grew 20%+ a month after Manus 1.5 and hit $100M ARR with 105 people.
Main lever: PR/launch (M)
Sources: manus.im, businessinsider.com
7. Perplexity: Stunts plus other people’s distribution
No sign-up needed to ask, and cited answers with suggested follow-ups got people to value fast. The CEO posts on X daily, runs stunts (a TikTok US bid, a Chrome offer, a Super Bowl contest that required 5 questions) and an @AskPerplexity reply bot answers tagged posts in public. Bundles like Airtel’s free year of Pro for 360M customers bought habit at scale.
Main lever: Founder-led social (M)
Sources: sacra.com, techcrunch.com
8. Harvey: Win the hardest customer, sell with lawyers
Founders cold-emailed Sam Altman with proof (attorneys would use 86 of 100 answers unedited) and got OpenAI Startup Fund money. Four people then signed Allen & Overy, one of the world’s largest firms, and every later deal pointed to that logo. Sellers are paired with ex-Big Law ‘legal engineers’.
Main lever: Sales-led (M)
Sources: lawnext.com, originnotes.substack.com
9. Sierra: Enterprise only, priced per resolution
Sold only to large consumer brands from day one, opened by founders Bret Taylor and Clay Bavor. It charges per resolved conversation, so the buyer pays for results, and starts on one channel before expanding. Half its customers have $1B+ revenue.
Main lever: Sales-led
Sources: sierra.ai
10. Clay: Seven years searching, then agencies as the wedge
Seven years in, Clay narrowed to cold email agencies found in WhatsApp and Slack groups, ran ‘reverse demos’ where the prospect drove, and moved all support into a Slack community. Agencies posted Clay workflows on LinkedIn to win clients, and Clay built an expert program, 60 Clay Clubs and a new job title around them.
Main lever: Community (M)
Sources: reuters.com, review.firstround.com
11. Linear: Hand-picked waitlist and a public changelog
Launched an invite-only beta in Apr 2019 with a blog post, a tweet and a waitlist of about 10,000 emails, then hand-picked about 10 users a week. It kept the waitlist for about a year, published a changelog constantly and never ran paid marketing. Startups adopted it, then their alumni took it to new companies.
Main lever: PLG/viral loop
Sources: linear.app, review.firstround.com
12. Gamma: Magic first 30 seconds, then 1,000 micro-influencers
Relaunched with AI in Mar 2023 so one prompt made a finished deck, and grew from 60,000 to 3M users in 3 months. Every free deck carries a ‘Made with Gamma’ badge that you pay to remove. After word of mouth carried it to $10M ARR, it added 1,000+ micro-influencers.
Main lever: PLG/viral loop (M)
Sources: techcrunch.com, linkedin.com
13. Granola: VCs as the wedge, founders as the spreaders
Spent a year in stealth with about 150 hand-onboarded users, then launched May 22 2024 with 4 people. It targeted VCs first because they have many similar meetings and talk to founders all day, then switched focus to founders on launch day. It skipped the meeting bot, so rivals’ bots became its referral prompt.
Main lever: PLG/viral loop (M)
Sources: techcrunch.com, startupriders.com
14. Fyxer AI: LinkedIn creators first, then an experiment machine
A human executive-assistant agency from 2016 to 2022 that rebuilt itself as an AI email assistant. LinkedIn creators and customers posting real experiences got early traction, then high-volume performance marketing and 514 experiments in 2025 scaled it. A $50 credit per invited teammate pushes it into whole teams.
Main lever: Paid UA (M)
Sources: growthunhinged.com
15. ElevenLabs: Free creators on one side, enterprise on the other
Launched a free tier in Jan 2023 that spread through Discord, Reddit, Hacker News and AI newsletters, reaching 1M+ users in 6 months. Faceless YouTube channels and viral parodies used its voices, and a Voice Library pays creators when their voices are used. Then it built enterprise sales and forward-deployed engineers.
Main lever: PLG/viral loop (M)
Sources: techcrunch.com, posts.adamgtm.com
16. HeyGen: Fiverr validation and a watermark
Validated demand by selling spokesperson videos on Fiverr before revealing they were AI (first customer paid $5). Launched on Product Hunt in Jul 2022 with a free tier and a visible watermark on every video, then fixed onboarding and shipped weekly to $1M ARR in 178 days. Viral translation clips in late 2023 took it further.
Main lever: Short-form/UGC (M)
Sources: linkedin.com, heygen.com
17. Midjourney: A product that lives in public on Discord
The product lived inside Discord: new users generated images in public ‘newbie’ channels where everyone saw every prompt and result. That made learning social and every image a demo, and it became the largest server on Discord. Free trials ended in Mar 2023 and the $10 a month plan carried it.
Main lever: Community
Sources: the-decoder.com, discord.com
18. Higgsfield: Presets plus paid creator volume, with a trust cost
Built around 400 one-click presets for camera moves and effects that look good in a feed. It paid creators to post clips (one offer Forbes saw was $200 to share a post with a pre-made clip) and leaned into controversial content to get attention on X. Growth was huge, and so was the backlash over unlabeled paid posts.
Main lever: Influencer/creator (M)
Sources: dealroom.co, forbes.com
19. Cal AI: One $2,000 story, then 250 creators on retainer
First test: about $2,000 for one story from a weight-loss influencer brought about 1,000 downloads in 2 days and about 100 payers. They then recruited influencers as fast as possible, reached about 250 on monthly retainers, and paid for view thresholds rather than single posts. Monthly revenue went from $30K to $1M in 6 months.
Main lever: Influencer/creator
Sources: techcrunch.com, netinfluencer.com
20. Ladder: Creator partnership ads at industrial volume
Ladder hired full-time creators in 2023 and runs most of its Meta ads as creator partnership ads. A third-party ad tracker counts 7,318 Meta ads since Jul 2025, 6,847 of them partnership ads and 97% video.
Main lever: Paid UA
Sources: linkedin.com, admanage.ai
21. Opal: Loosen the paywall, own a creator account
At $5M ARR with a hard paywall growth stalled, so Opal went freemium; conversion fell from 20% to 9% but DAU passed 1M and ARR hit $10M. It runs ‘Olivia Unplugged’, a person-led educational account by its social media manager.
Main lever: Short-form/UGC
Sources: revenuecat.com, linkinbio.news
22. Coconote: 25 part-time creators instead of ads
No paid ads. About 25 part-time content creators with 5,000 to 10,000 followers, found through the personal Gmail in their bio, posted for students on TikTok and Reels. $1M ARR in 4 months, then onboarding and cancellation-flow tweaks.
Main lever: Short-form/UGC
Sources: subclub.com, revenuecat.com
23. Duolingo: A retention model plus an unhinged owl
Two engines: a retention machine (streaks, leaderboards, tuned notifications) that grew DAU 4.5x over four years, and a TikTok account run by a tiny team that treats comments as the brief. The unhinged owl made the brand a meme.
Main lever: Short-form/UGC
Sources: investors.duolingo.com, lennysnewsletter.com
24. Partiful: Every invite is an ad
Every party is a link sent by text, and guests RSVP on a page that shows off the product. It spread friend group by friend group from New York, and a 3-person marketing team amplifies organic moments like lookalike contests hosted on Partiful.
Main lever: PLG/viral loop
Sources: techcrunch.com, marketingbrew.com
25. Ramp: Free product, sold hard
A free corporate card paid for by interchange, sold by an outbound team from a month after launch. Ramp does the switching for customers, pitches savings, built an accountant partner program and uses stunts and Super Bowl ads as ‘surround sound’ for sales.
Main lever: Sales-led (M)
Sources: prnewswire.com, originnotes.substack.com
26. Monarch Money: A competitor’s bad day as a channel
Founded by Mint’s first product manager and one of the only apps that could import Mint data. When Intuit announced Mint’s shutdown in Nov 2023, Monarch published a guide, offered 50% off the first year with code MINT50 and rebuilt its importer for tens of millions of transactions a day. Sign-ups ran 20x to 35x normal.
Main lever: SEO/ASO (M)
Sources: fastcompany.com, monarch.com
27. Whop: Pay for views, not posts
Whop is a marketplace for digital products and communities. Its Content Rewards product lets any brand or creator fund a budget and pay clippers per 1,000 verified views; Whop says it hit 700,000+ submissions and 4B views in 8 weeks.
Main lever: Influencer/creator
Sources: sourcery.vc, whop.com
The other 73, and what each did
Grouped by the lever that did the most work. (M) marks medium confidence.
PLG/viral loop (17)
OpusClip (M): OpusClip cuts long videos into short clips and gives each a ‘virality score’. The score became the hook creators and agencies talk about, and the clips themselves flood TikTok, Shorts and Reels. It sells self-serve plans to creators and team plans to agencies.
CapCut (M): ByteDance made CapCut the default editor for TikTok trends: templates match trending formats and export straight to TikTok. Posts made from a template show a CapCut template link on TikTok, so each viral edit sends viewers into the app. Free editing with paid Pro features.
Sora app (faded): OpenAI launched Sora as an invite-only app where each user got codes to share. ‘Cameos’ let you put yourself and friends into videos, so every video recruited new people. It passed 1M downloads in under 5 days, but usage did not hold and the app was discontinued in Mar 2026.
Hevy: A workout logger with a social feed: users post workouts and follow friends, which pulls friends in. A generous free tier keeps the feed full, and Pro unlocks unlimited routines and analytics. Its founder says it reached 2M downloads with no paid marketing.
Wispr Flow (M): CEO Tanay Kothari personally onboarded about 500 early users on Google Meet to find what made people rave. Its founder said more than half of new users came from Reddit and X with no paid ads. The referral program gives a friend a 30-day Pro trial and the referrer a free month once the friend dictates 2,000 words or subscribes.
ChatGPT: OpenAI released it on Nov 30 2022 as a free ‘research preview’ with no waitlist and no marketing. People posted screenshots of answers on X, which became the ad, and Sam Altman said it passed 1M users in 5 days. A $20 Plus plan (Feb 2023) and mobile apps (May 2023) sat on top of the free tier as a huge funnel.
Gemini app: Google placed Gemini inside Android, Chrome, Workspace and Search. The breakout was the Nano Banana image model in Aug 2025: users turned photos into 3D figurine images and posted them everywhere, which pulled in younger users. MAU rose from 450M to 650M+ in one quarter.
Genspark (M): Genspark spent 2024 as an AI search engine with modest traction, then launched Super Agent on Apr 2 2025. Its COO says it reached 2M users and $36M ARR in 45 days, mostly from shareable outputs such as slides, sheets and reports and word of mouth. The team shipped new agents every few weeks and added creators and paid later.
Character.ai: Founded by ex-Google LaMDA researchers, it opened a web beta in Sep 2022 and launched apps in May 2023. Users create characters and share links to them, so each viral character pulled in its fans and the catalog grew without the company making content. The apps got 1.7M installs in week one.
Emergent (M): Aimed at non-technical founders and small business owners rather than developers. Growth ran on niche creators on Instagram, TikTok, X and LinkedIn filming real apps being built, with each creator tracked for conversions, plus user success stories reposted by the company. The team sold mostly to the US and Europe from India, and its founder says 8 engineers built it to $100M in 8 months.
v0 by Vercel (M): Vercel launched v0 to its existing Next.js audience in Oct 2023 behind a waitlist, then opened it up. Generated UI drops straight into Next.js and deploys on Vercel, so every v0 project is also a hosting lead for the parent company. Sacra estimates v0 at about $42M ARR by Feb 2025.
Windsurf (M): As Codeium it gave individual developers free autocomplete, building a large bottom-up base, and sold enterprise plans with self-hosting and security controls. In Nov 2024 it launched the Windsurf editor with an agent to compete with Cursor head on. After Google hired its founders in a licensing deal, Cognition bought the rest with $82M ARR.
Gas: Nikita Bier’s team launched one school at a time. They made a private Instagram account per school, followed students who listed it, then accepted every follow request at once after classes ended; the bio said to download Gas to ‘see who likes you’. Compliment polls about classmates gave everyone a reason to invite the rest of the school.
Lapse (faded): A disposable-camera photo app that made every new user invite 5 friends before getting in. The invite wall drove it to #1 on the US App Store in Sep 2023. Downloads then fell 70%, because forced invites brought people who never chose the app.
OpenEvidence: Free for clinicians who verify with their NPI number, so doctors adopt it without procurement. Content licensing deals with NEJM and JAMA made answers trustworthy and became a selling point. Revenue comes from pharma and device ads, not doctors, and it reached about 40% of US physicians by Jul 2025 by word of mouth.
pump.fun (M): Anyone can launch a token with a name, ticker and image; since Aug 2024 creation is free and the first buyer pays the fee. Tokens trade on a bonding curve and graduate to a DEX, and every creator promotes their token, which promotes pump.fun. It passed 4M tokens by late 2024.
Phantom (M): Phantom became the default Solana wallet through simple onboarding in a browser extension and mobile app. When memecoin trading boomed, every new trader needed a wallet and most picked the one their friends used. In-app swaps earn fees, so trading volume became revenue.
Community (11)
Krea (M): Krea launched real-time generation where the image updates as you draw, and designers posted screen recordings of it. It then put many image and video models in one interface, so users did not need several subscriptions. Designer word of mouth drove most of its 20M+ users.
Leonardo.ai (M): Leonardo grew in a Discord community and gave free daily credits, so people came back every day to generate. Community fine-tuned models and game-asset use cases drew designers and game studios. Canva bought it once it had 19M+ users.
Retell AI (M): Retell sells voice agents self-serve at $0.07 to $0.31 a minute with no contract. Its real channel is AI automation agencies that build agents for clients in separate workspaces, set their own prices and teach Retell on YouTube. A partner program and an affiliate deal paying 15% of referred revenue for 12 months formalize that.
Framer (M): Framer moved from a prototyping tool to a no-code website builder for designers. Its marketplace lets creators sell templates, so each template seller is also a marketer, and designer YouTubers teach it for their own audiences. AI site generation and an expert network for agencies came later.
n8n (M): Fair-code and self-hostable, so technical users can run it free on their own servers. When AI agents arrived, its community published templates and YouTube tutorials on building agents in n8n, and many creators sell templates. The company says 80%+ of workflows now involve AI agents.
Fizz: Launched at Stanford in 2021 as an anonymous campus feed and reached 95% of undergrads. Each new campus got student ambassadors, launch events with free food and merch, and trained student moderators who kept the feed local. It expanded one college at a time to 800+.
Saturn: A high school calendar that works alone (your schedule) and gets better with friends (see theirs). Student ambassadors ran ‘launch night’ at each school, and it spread to nearly 20,000 schools without a single school partnership.
Bluesky: Bluesky surged after the Nov 2024 US election. ‘Starter packs’, curated lists that let anyone follow many accounts in one tap, let communities move together, and a Lancaster University study found they played a central role. It added 1M sign-ups in a single day.
MagicSchool: Adeel Khan, a former principal, made a free tool that saves teachers time on lesson plans, rubrics and differentiation. Teachers shared it with colleagues and became champions who asked their districts to buy it. District sales then converted that demand into paid contracts, often starting with teacher pilots.
Hyperliquid: An 11-person team with zero VC built a fast on-chain perps exchange that feels like a centralized one. A points program from Nov 2023 rewarded trading and referrals, and the Nov 2024 HYPE token gave 31% of supply to users with no allocation to VCs, turning users into owners and promoters.
Nothing: Carl Pei built hype with teaser drops, an invite-only first phone launch and a distinctive transparent design with Glyph lights. Fans could invest through community rounds: $8M from nearly 8,000 people, which made them promoters. Revenue doubled in 2024.
Short-form/UGC (7)
Captions (M): Started as an auto-caption app for talking-head creators, whose videos showed its caption styles to millions. It added AI dubbing, eye-contact correction and AI avatars, each a feature creators demo on camera. The company later renamed its research and studio arm Mirage.
Suno (M): Suno launched as a plugin inside Microsoft Copilot in Dec 2023, then its v3 model in Mar 2024 made full songs from a text prompt. Songs made in Suno spread on TikTok and X as the ad, and meme songs drove waves of sign-ups. It reached $200M revenue despite label lawsuits.
Kling AI (M): Kuaishou launched Kling in Jun 2024, in China first and then globally. Its realistic output clips spread on X and TikTok within days of each model release, and new versions shipped every few months to restart the cycle. Revenue comes from credit packs and subscriptions plus an API.
Umax (M): Built by Blake Anderson, who later co-founded Cal AI, around the looksmaxxing trend on TikTok. The team ran in-house UGC accounts, faceless slideshows and face-rating demo videos, and tested hooks fast. Anderson says influencers were the most profitable channel and paid ads more scalable but less efficient.
RIZZ (M): Roman Khaves copied a proven TikTok format, ‘texting story’ videos, and wrote the app into the story. He paid faceless creators $1 to $2 per 1,000 views with a cap per post, and creators ran fleets of TikTok and Instagram accounts. He says he paid creators over $2M for 300,000+ videos and 5B+ views.
Finch (M): Finch ties self-care goals to a virtual pet bird that grows when you do. Fans post their birds and routines on TikTok, which drives most installs. Bootstrapped with a generous free tier and an optional Plus subscription.
Tea (faded, M): A women-only app for anonymously reviewing men you date. It went viral on TikTok and Reddit in mid-Jul 2025, with daily downloads going from about 15,000 to a peak of 142,000 and a waitlist of about 900,000. A data breach, then removal from the App Store, ended the run.
Influencer/creator (6)
Quittr: Launched Aug 2024 with $3,000 to pay creators; a Christian TikToker with 700,000+ followers was an early promoter, and the founders say it made $30,000 in its first month. It charges $30 a year up front and gamifies streaks from ‘Sprout’ to ‘Nirvana’ at 90 days. It now recruits creators through a paid UGC program.
Function Health (M): Function sells a $499 a year membership for 100+ lab tests. Co-founder Dr. Mark Hyman’s podcast, books and social following were the first channel, then a creator program where creators get tested and post their own results. The company scaled it into a multi-tier creator program.
Lemon8 (faded): ByteDance paid creators to post on Lemon8 before its US push and promoted it with ads, reaching the US top 10 in Mar 2023. Paid seeding filled the feed but did not create a reason to return. It had about 2.6M US downloads by Nov 2023 and stalled.
Kick (M): Kick bought supply: it signed star streamers such as xQc to a reported 2-year deal worth $70M, up to $100M with incentives, and offered creators a 95/5 subscription revenue split versus Twitch’s 50/50 standard. Viewers followed the streamers. Hours watched roughly doubled in 2024.
Oura (M): Oura seeded the ring with athletes and celebrities (the NBA used it in its 2020 bubble) and positioned sleep scores as status. The Gen3 ring (2021) added a monthly membership, turning hardware into recurring revenue. Retail and partner deals widened distribution.
TikTok Shop: TikTok subsidized discounts and shipping and recruited affiliate creators who earn commission on products shown in their videos and lives, often with free samples. Affiliates drove about 60% of US GMV. Shopping happens inside the feed, so a viral video is also the checkout.
PR/launch (7)
Runway (M): Runway marketed to filmmakers and studios: it runs an annual AI Film Festival and signed a model deal with Lionsgate in 2024. Each model (Gen-2, Gen-3, Gen-4) launched with polished demo reels. It grew more slowly than consumer-viral peers, a useful counterexample.
DeepSeek: DeepSeek released R1 on Jan 20 2025 with open weights and a free app. Its claim that the V3 base model cost about $5.6M to train spread through press and X and hit AI stocks. There was no paid marketing; price and openness were the news.
Devin (M): Cognition launched Devin in Mar 2024 with one demo video calling it ‘the first AI software engineer’ and a waitlist, which earned global press. It first sold to teams at $500 a month, then cut entry pricing to $20 with Devin 2.0 in Apr 2025 to widen the funnel. It later bought Windsurf to add an IDE and an enterprise sales team.
noplace (M): A Gen Z social app with customizable, Myspace-style profiles. It ran an invite-only beta to build a queue, then opened publicly in Jul 2024 and hit #1 on the US App Store as teens shared their profiles. Sustaining usage after launch week was the hard part.
Polymarket (M): Polymarket made its odds the content: shareable charts and screenshots spread across X, TikTok and Substack. Its odds appeared in Perplexity answers (Aug 2024) and in news embeds, and the founder posts almost only Polymarket content. The 2024 election market became a daily news source.
Plaud (M): Plaud launched its card-sized AI recorder on Kickstarter in 2023 around one pain point: recording iPhone calls. It then moved to Amazon, where reviews and creator videos on TikTok and YouTube with discount codes drove sales. It shipped 1M+ devices and is profitable without heavy fundraising.
Rabbit r1 (faded): Rabbit unveiled the r1 with a keynote video at CES in Jan 2024 at $199 with no subscription, and preorders sold out batch after batch. Reviews after shipping called it underpowered, and the hype did not survive contact with the product.
Sales-led (6)
Synthesia: Synthesia sold AI avatar video to enterprise learning and development and internal comms teams rather than creators. A free self-serve tier lets buyers make a sample video, then sales handles security, languages and seat expansion. It crossed $100M ARR in Apr 2025 with Adobe investing.
Glean: Glean sold enterprise search for years, then repositioned as the AI assistant and agent platform for work once ChatGPT made buyers want one. Because it indexes every app a company uses, it is sold top-down with pilots that expand to company-wide rollouts. ARR doubled in 9 months.
Anthropic: Anthropic sold the Claude API to developers and enterprises first and won coding, as Claude became the default model inside Cursor, Bolt, Lovable and Replit. Enterprises could buy it through AWS Bedrock and Google Vertex AI, where they already had budgets. Claude Code then turned the model lead into its own product line.
Legora (M): Legora started as a small team working inside Mannheimer Swartling’s Stockholm office, its first law firm partner, building with its lawyers. It then repeated the play market by market: win one elite local firm, build with them, sell to the rest using that reference. In the US it went after firms such as Cleary Gottlieb and Goodwin.
Decagon (M): The founders ran about 100 discovery conversations before building and used warm intros to land Eventbrite, then Bilt. They sold a labor-cost ROI customers could prove, and let support teams configure agents themselves with ‘agent operating procedures’ written in plain language. Pricing is mostly per conversation, with per-resolution available.
Mercor (M): Mercor pivoted from general hiring to supplying vetted domain experts to AI labs for training data and evals. Supply grew through referral bonuses for bringing in a successful expert. The founders sold directly to a small number of labs with very large budgets.
Paid UA (6)
Bible Chat (M): Built by a Romanian team and scaled on high-volume paid acquisition: hundreds of Meta ad variations (UGC, narrative, animation, advice formats) plus Google and TikTok ads. Each creative angle targets a different reason to open scripture. Forbes Romania reports revenue tripled in 2025 to 72.16M lei.
Hallow (M): Hallow bought a Super Bowl spot in Feb 2024, days before Ash Wednesday, with Mark Wahlberg and Jonathan Roumie inviting viewers into its ‘Pray40’ Lent challenge. The ad itself was a short prayer, and the challenge turned one night of attention into 40 days of habit. It reached #1 on the US App Store.
Speak (M): Speak launched in South Korea first, where adults spend heavily on English speaking practice. It built a Korean team and ran brand campaigns with celebrity TV ads, Apple Search Ads and influencers. Its CEO says 10% of South Koreans have tried Speak and two thirds know the brand.
Gauth (M): ByteDance’s homework helper: snap a photo of a problem and get a worked answer. It scaled with paid user acquisition timed to back-to-school season. By late Sep 2024 it was the #2 US iOS education app by downloads.
Talkie (M): MiniMax’s companion app scaled mainly with paid user acquisition on social ad networks in the US and other markets. Characters are user-created, and collectible card features give users reasons to return daily. It was the top AI companion app by 2024 downloads.
Temu: Temu outspent everyone on Meta and Google ads, ran ‘Shop like a billionaire’ Super Bowl spots in 2023, then 5 spots with $15M+ in giveaways in 2024. Referral games gave free items and coupons for inviting friends. Ultra-low prices from shipping direct from factories made the ads convert.
Founder-led social (5)
Photo AI: Pieter Levels built it alone and shipped live on X, posting features, bugs and revenue to a large following. Users posted their AI headshots, and he kept a public revenue counter. No team, no ads.
beehiiv (M): CEO Tyler Denk builds in public on X and LinkedIn, shipping weekly. The product spreads through a native referral program for newsletters, a ‘made with beehiiv’ footer and Boosts, where newsletters pay each other to recommend them. It took 0% of creators’ paid subscription revenue to win switchers from Substack.
Cluely (faded): Roy Lee launched with a deliberately provocative video and a ‘cheat on everything’ manifesto after Columbia suspended him over an interview-cheating tool. Controversy on X and LinkedIn was the whole channel. In 2026 he admitted the ARR he had posted was overstated, which shows the risk of revenue-as-content.
Base44 (M): Maor Shlomo built it alone and posted the whole journey on LinkedIn and X: user counts, revenue, outages and lessons. Users earned free credits for posting what they built, a near-free sharing loop. He has said paid ads and influencers did not work for him. Wix bought it for $80M cash about 6 months after launch.
Rork (M): Rork turns a prompt into a native mobile app. The founder posted each launch and revenue milestone on X, and each launch produced a visible revenue spike. Four people reached $1M ARR in about 3 months.
Partnerships (7)
Supabase: An open-source ‘Firebase alternative’ that runs a ‘Launch Week’ several times a year. It became the backend that Lovable, Bolt, v0 and Figma Make create for their users, so every vibe-coded app is a new Supabase project. Developers grew from 1M to 4.5M+ in under a year, and 55% of a recent YC batch used it.
Meta AI: Meta put Meta AI in the search bars and chats of WhatsApp, Instagram, Facebook and Messenger in 2024, so billions met it without downloading anything. A standalone app followed in Apr 2025. The 1B MAU figure is almost entirely placement inside apps people already use daily.
Doubao (M): ByteDance promoted Doubao inside its own apps such as Douyin, so it did not have to buy most of its reach. It shipped voice, image and video features quickly and priced its models low for developers. It became the first Chinese standalone AI app past 100M DAU.
Threads: Meta launched Threads in Jul 2023 on top of Instagram: one tap to create an account, import your bio and follow everyone you already follow. It launched during user frustration with X. 100M sign-ups in 5 days came almost entirely from Instagram’s graph.
Abridge (M): Abridge built its note-taking AI into Epic, the records system doctors already use, through Epic’s partner program. That made system-wide deals possible: Kaiser Permanente rolled it out to 24,600 physicians across 40 hospitals and 600+ clinics, and UPMC to 12,000 clinicians. It sells to health system leadership, then champions drive usage.
Kalshi (M): Kalshi won its court fight with the CFTC over election contracts in 2024 (the CFTC dropped its appeal in May 2025), turning regulation into a moat. It then listed sports contracts nationwide and distributed through brokers such as Robinhood, where millions already trade. Sports became the volume engine.
Ray-Ban Meta: Meta partnered with EssilorLuxottica so the glasses look like normal Ray-Bans and sell through eyewear stores and optical retailers. Meta AI with the camera became the hook in 2024. 7M+ sold in 2025.
SEO/ASO (1)
Photoroom (M): Photoroom won App Store search for background removal and product photos, so small sellers found it at the moment of need. Free removal hooks them, and batch editing and pro templates convert sellers who list many items. It grew mostly through app store search rather than ads.
Flagged numbers, checked
Ladder, confirmed: 6,847 creator partnership ads. The AdManage tracker of Meta’s Ad Library shows 7,318 Ladder ads since Jul 2025, 1,833 live and 6,847 partnership ads as of Nov 2026. It is a third-party count, so cite it as ‘per AdManage’ (AdManage).
Whop, confirmed: 4B views and 700K submissions in 8 weeks, a company claim in Whop’s own update for the week of May 2025: ‘over 700k submissions and 4B views in only 8 weeks’ (Whop).
Cursor, confirmed: acquisition by SpaceX valued at $60B, effective Aug 14 2026 (TechCrunch).
How we built this, and its limits
Selection: 100 brands with a public growth number from late 2022 to Nov 2026, weighted toward consumer apps and AI products, plus notable B2B and prosumer names. A brand needed a sourced metric to make the list.
Depth: 27 brands got a full teardown because founders or operators have described their playbook in interviews, podcasts or first-party posts. The other 73 got a 3 to 5 sentence summary from public reporting.
Sources: founder interviews and podcasts (Lenny’s, Sub Club, Invest Like the Best, Training Data), First Round Review, Growth Unhinged, company blogs, TechCrunch, Reuters, Forbes, Fortune, Sacra, Appfigures and ad-library trackers. Sacra, Appfigures and tracker numbers are third-party estimates.
Teardown analyses: newsletters such as Origin Notes and Startup Riders were used for structure; key numbers were checked against the interviews and company posts they cite where those were reachable.
Judgment calls: primary lever and play tags are our call from public reporting. (M) marks medium confidence. Where a first-user channel was not public it says ‘Not disclosed’ rather than a guess.
ARR is not one thing: run rate, annualized revenue, GMV-based run rate and contracted ARR all appear, and most numbers are self-reported. Cluely’s CEO later admitted overstating his.
Survivorship bias: this is a list of winners plus a few visible flameouts. Many companies ran the same plays and failed.

© Nexgen Media 2026
