Tanmay Jain
Where to hear and read about Tanmay Jain1 episode or test
Tanmay Jain works on ElevenLabs' consumer mobile apps, after a stint at Canva and a spell as a founder. The thing most worth learning from him is to launch from a story you wrote first as a Twitter thread, and to treat each earned-media launch as an asset that compounds.
If you don't start with a very clear storyline yourself, then you're not going to be able to rely on anybody else.
How ElevenLabs Ships So Fast With Small Teams
Tanmay Jain: the story and the lessons
1Tanmay Jain's path from Canva to ElevenLabs
Jain ran a slow, single-date launch at Canva from 2019 to 2022. The Canva video product had a 12-month run-up and three months spent on messaging. The launch hinged on one day, with news articles lined up for that exact date.
This was before GPT-3.5 made AI a mainstream topic.
He left big-company review layers for autonomy without process. At ElevenLabs he found a place where the only rate limiter is your own capability, not PR and legal sign-offs.
His view is that one quarter without iterating could be fatal, so legal and PR review matter but are less existential than moving.
His role is whatever it takes, including shipping code and experiments. Jain merges PRs, reviews Figma designs and ships paywall experiments himself.
He ran his first launch about six weeks after joining. He was handed a doc describing the launch process in his first weeks, with no long onboarding.
The creative team reviews scripts and the head of growth reviews threads, so the craft is learned through repeated feedback.
2Jain's launch playbook: thread first, story first
Jain writes the launch Twitter thread before anything else. The team writes the core value promise, turns it into a thread with a headline and hooks to the next tweet, then turns the thread into a video.
"If you're working on a feature and it doesn't make for a good Twitter thread, like you better have like a really good reason to be."
He spends one to two weeks refining the story before assets. A new state-of-the-art AI benchmark seems to appear every week, so the team cannot count on Apple, Google or press to tell its story.
ElevenLabs keeps an in-house creator to film short-form content and a motion design team. Relationships with App Store and Play Store reps are good, but featuring is not the main lever.
Launch fast, launch often replaced the long single-date launch. Teams launch roughly monthly, then learn what resonates, instead of the Canva-style 12-month build to one date.
A campaign that looks like a flash in the pan in hindsight can still guide later decisions about market, users and pricing.
He avoids X-for-Y features and starts from what only ElevenLabs can do. Jain says the tools that failed tended to come from applying ideas from somewhere else.
The CEO passed on a Brian Armstrong line that action gets information.
He looks for a natural shareable moment instead of copying Spotify Wrapped. Wrapped worked because it turned passive listening into an identity people wanted to show in a compressed window. Jain's working examples are the GitHub skyline, a YouTube milestone and ElevenLabs giving agent builders their total hours of agent interaction.
The best shareable moments are rarely designed, so teams keep testing until one sticks. Only about one in ten unhinged ideas makes it out.
3How Jain pairs earned media with paid growth
Jain treats earned media as a compounding asset. Each launch earns more attention, so the next launch lands bigger. It also fills branded search, the first place paid starts, and warms people who later see ads.
He admits they have not run the numbers but is confident it lowers CPA and CPI. "Every launch we do earns more eyeballs, which means that the next launch earns more eyeballs."
He runs launches and steady paid as two parallel tracks. Launches chase viral upside and are targeted by audience, for example LinkedIn may not suit a short-form creator app. Paid runs on Meta, TikTok, Apple Search Ads and Google as a forecastable monthly engine.
For each paid channel the team defines the product and message, then builds models to keep it sustainable. "Both things have to be true at once."
Brand from launches lets paid tolerate slower payback. The team does not have to optimize for day zero, day six or week-one conversion, so it can test new creatives and accept a higher CAC while working out payback.
Each channel behaves differently, so creative is learned channel by channel and the channels layer up together.
4How ElevenLabs organizes and hires for speed
ElevenLabs runs like 10 to 12 startups in small pods. Each pod is self-contained with development, design, growth and product. The mobile team is about 12 people, and the company is about 300 going to 400.
Shared roles for operations, creative and brand, SEO and web sit across pods, so pods take standards off the shelf instead of rebuilding them.
Jain's team screens engineers by public GitHub side projects. They look for people who build on their own and obsess over a user pain point, and for people from companies known for strong product cultures.
Referrals form a flywheel, but the team probes how well the referrer really knows the person. Former founders fail in a hierarchical culture, so they are hired only into one built for speed and autonomy.
Speed is the guiding principle, even when it costs polish. Model blips will happen and the support team absorbs them. The trade-off is a 105% fine-tuned model versus moving quicker and pushing the edges.
The sources mention the costs but do not weigh them.
5How Jain thinks about consumer AI products and pricing
Jain prices consumer apps in hours, not credits or tokens. ElevenReader dropped a whole pricing tier and sells hours of listening. Jain calls it the biggest win of the year for that app.
The flagship creator app keeps credits because its users were trained on them. Real compute costs exist but should not shape how a consumer is spoken to.
Mobile gets one price while web offers upwards of five plans. The mobile app distills what works on web instead of squeezing the dashboard onto a small screen, and tests incentives around a single price.
One price also holds the team accountable when picking the next thing to build. Apple and Play Store limits were one reason the web plan variety did not carry over.
ElevenReader's aha is bringing your own file and picking a voice. Users upload an ebook, EPUB or PDF and choose the voice to hear it in. The team found this by reading mentions on Reddit, Discord and X, where users share screenshots.
The team skipped goal-setting quizzes and focused on getting users to a voice quickly.
A 600-voice catalog works because preferences are a long tail. No single voice has more than 10 to 12% of use, and the median voice gets about 1 to 2%. Most apps offer no choice or two or three options.
Selection itself is a value driver and a shareable moment.
Jain sees the moat as speed, ecosystem and brand, not model quality. Large labs will ship good-enough voice models, so ElevenLabs keeps doubling down on voice and adds platform depth such as custom evals, workflows and integrations.
Advantages accrue at the application layer and in brand: when people think of voice, ElevenLabs comes to mind first. He admits brand sounds too simple for a moat.
6Jain's mistakes and what he changed
He over-optimized onboarding without testing a plain base case. The team spent too many cycles on onboarding quizzes and funnels, without comparing against a sign-up form straight into the zero-state product.
His fix is to strip onboarding out for 10% of users for 5 days and size the opportunity. Simple, direct onboarding with fewer screens to the aha won out.
His Canva Stories spin-out won awards but never found an audience. The app aimed at Instagram stories won Apple awards, yet never cracked acquisition, retention or brand awareness and was folded back into Canva.
Users wanted Canva and saw Stories as a subset. ElevenLabs is a horizontal platform, so its two apps with separate brands but a shared design system and tech stack fit better.
As a founder he learned price cuts do not fix indifference. Raising prices and then dropping them after churn may keep users one extra month, but people who do not care about the product still leave.
Users who value a product will pay a lot to use it. He calls the trap getting academic about price in the moment.
How ElevenLabs Ships So Fast With Small Teams
ElevenLabs runs small autonomous pods, prices consumer apps in terms users understand, and treats earned-media launches as a compounding asset that also makes paid cheaper. The mobile leads share what worked, such as dropping a pricing tier for hours of listening, and what failed, such as over-optimizing onboarding instead of testing a stripped-back base case.
Jain ran a slow, single-date launch at Canva from 2019 to 2022. The Canva video product had a 12-month run-up and three months spent on messaging. The launch hinged on one day, with news articles lined up for that exact date.
This was before GPT-3.5 made AI a mainstream topic.
He left big-company review layers for autonomy without process. At ElevenLabs he found a place where the only rate limiter is your own capability, not PR and legal sign-offs.
His view is that one quarter without iterating could be fatal, so legal and PR review matter but are less existential than moving.
His role is whatever it takes, including shipping code and experiments. Jain merges PRs, reviews Figma designs and ships paywall experiments himself.
He ran his first launch about six weeks after joining. He was handed a doc describing the launch process in his first weeks, with no long onboarding.
The creative team reviews scripts and the head of growth reviews threads, so the craft is learned through repeated feedback.
Jain writes the launch Twitter thread before anything else. The team writes the core value promise, turns it into a thread with a headline and hooks to the next tweet, then turns the thread into a video.
"If you're working on a feature and it doesn't make for a good Twitter thread, like you better have like a really good reason to be."
He spends one to two weeks refining the story before assets. A new state-of-the-art AI benchmark seems to appear every week, so the team cannot count on Apple, Google or press to tell its story.
ElevenLabs keeps an in-house creator to film short-form content and a motion design team. Relationships with App Store and Play Store reps are good, but featuring is not the main lever.
Launch fast, launch often replaced the long single-date launch. Teams launch roughly monthly, then learn what resonates, instead of the Canva-style 12-month build to one date.
A campaign that looks like a flash in the pan in hindsight can still guide later decisions about market, users and pricing.
He avoids X-for-Y features and starts from what only ElevenLabs can do. Jain says the tools that failed tended to come from applying ideas from somewhere else.
The CEO passed on a Brian Armstrong line that action gets information.
He looks for a natural shareable moment instead of copying Spotify Wrapped. Wrapped worked because it turned passive listening into an identity people wanted to show in a compressed window. Jain's working examples are the GitHub skyline, a YouTube milestone and ElevenLabs giving agent builders their total hours of agent interaction.
The best shareable moments are rarely designed, so teams keep testing until one sticks. Only about one in ten unhinged ideas makes it out.
Jain treats earned media as a compounding asset. Each launch earns more attention, so the next launch lands bigger. It also fills branded search, the first place paid starts, and warms people who later see ads.
He admits they have not run the numbers but is confident it lowers CPA and CPI. "Every launch we do earns more eyeballs, which means that the next launch earns more eyeballs."
He runs launches and steady paid as two parallel tracks. Launches chase viral upside and are targeted by audience, for example LinkedIn may not suit a short-form creator app. Paid runs on Meta, TikTok, Apple Search Ads and Google as a forecastable monthly engine.
For each paid channel the team defines the product and message, then builds models to keep it sustainable. "Both things have to be true at once."
Brand from launches lets paid tolerate slower payback. The team does not have to optimize for day zero, day six or week-one conversion, so it can test new creatives and accept a higher CAC while working out payback.
Each channel behaves differently, so creative is learned channel by channel and the channels layer up together.
ElevenLabs runs like 10 to 12 startups in small pods. Each pod is self-contained with development, design, growth and product. The mobile team is about 12 people, and the company is about 300 going to 400.
Shared roles for operations, creative and brand, SEO and web sit across pods, so pods take standards off the shelf instead of rebuilding them.
Jain's team screens engineers by public GitHub side projects. They look for people who build on their own and obsess over a user pain point, and for people from companies known for strong product cultures.
Referrals form a flywheel, but the team probes how well the referrer really knows the person. Former founders fail in a hierarchical culture, so they are hired only into one built for speed and autonomy.
Speed is the guiding principle, even when it costs polish. Model blips will happen and the support team absorbs them. The trade-off is a 105% fine-tuned model versus moving quicker and pushing the edges.
The sources mention the costs but do not weigh them.
Jain prices consumer apps in hours, not credits or tokens. ElevenReader dropped a whole pricing tier and sells hours of listening. Jain calls it the biggest win of the year for that app.
The flagship creator app keeps credits because its users were trained on them. Real compute costs exist but should not shape how a consumer is spoken to.
Mobile gets one price while web offers upwards of five plans. The mobile app distills what works on web instead of squeezing the dashboard onto a small screen, and tests incentives around a single price.
One price also holds the team accountable when picking the next thing to build. Apple and Play Store limits were one reason the web plan variety did not carry over.
ElevenReader's aha is bringing your own file and picking a voice. Users upload an ebook, EPUB or PDF and choose the voice to hear it in. The team found this by reading mentions on Reddit, Discord and X, where users share screenshots.
The team skipped goal-setting quizzes and focused on getting users to a voice quickly.
A 600-voice catalog works because preferences are a long tail. No single voice has more than 10 to 12% of use, and the median voice gets about 1 to 2%. Most apps offer no choice or two or three options.
Selection itself is a value driver and a shareable moment.
Jain sees the moat as speed, ecosystem and brand, not model quality. Large labs will ship good-enough voice models, so ElevenLabs keeps doubling down on voice and adds platform depth such as custom evals, workflows and integrations.
Advantages accrue at the application layer and in brand: when people think of voice, ElevenLabs comes to mind first. He admits brand sounds too simple for a moat.
He over-optimized onboarding without testing a plain base case. The team spent too many cycles on onboarding quizzes and funnels, without comparing against a sign-up form straight into the zero-state product.
His fix is to strip onboarding out for 10% of users for 5 days and size the opportunity. Simple, direct onboarding with fewer screens to the aha won out.
His Canva Stories spin-out won awards but never found an audience. The app aimed at Instagram stories won Apple awards, yet never cracked acquisition, retention or brand awareness and was folded back into Canva.
Users wanted Canva and saw Stories as a subset. ElevenLabs is a horizontal platform, so its two apps with separate brands but a shared design system and tech stack fit better.
As a founder he learned price cuts do not fix indifference. Raising prices and then dropping them after churn may keep users one extra month, but people who do not care about the product still leave.
Users who value a product will pay a lot to use it. He calls the trap getting academic about price in the moment.