LLenny's Podcast
← All episodes
Anton Osika (co-founder and CEO)09 March 2025

Building Lovable: $10M ARR in 60 days with 15 people

7Frameworks
15Insights

Episode overview

Lenny Rachitsky interviews Lovable co-founder and CEO Anton Osika about turning natural-language descriptions into working software products. Osika explains how Lovable reached $10 million ARR within two months with roughly 15 people, demonstrates an Airbnb-style application, and describes the product and engineering choices behind that growth. They also discuss AI-era product skills, hiring highly capable generalists, weekly prioritization, and the lesson Osika learned from trying to retrofit personalized AI into an existing educational product at Sana Labs.

Key ideas

  • Lovable aims to let non-programmers create working software by describing what they want to an AI software engineer.
  • Clear, specific prompting, patience, curiosity, and explaining expected versus actual behavior make AI development tools more productive.
  • The team improves reliability by identifying where generated applications get stuck and strengthening critical flows such as authentication, persistence, and payments.
  • Lovable's early growth was driven primarily by product love, organic word of mouth, and public demonstrations of newly shipped capabilities.
  • As software construction becomes easier, product discovery, user understanding, taste, systems thinking, and broad generalist skills become more valuable.
  • Lovable hires people who care intensely, learn across disciplines, possess a distinctive strength, and prove themselves through realistic work simulations.
  • The team prioritizes the largest current bottleneck, plans weekly, keeps its roadmap flexible, and uses in-person collaboration to exchange context quickly.
  • Osika recommends choosing a real outcome and spending a focused week solving it end to end with AI tools.

Transcript available · source text is retained privately and is not published

Frameworks in this episode

Strategy3 steps

Biggest-Bottleneck Prioritization

Solve the single biggest problem, then pick the next — don't dream up a long roadmap.

Read →
Leadership4 steps

Generalist-With-A-Superpower Hiring

Hire people who care obsessively, have a generalist brain, and one absolute superpower — then trial them for a week.

Read →
Innovation3 steps

Minimum Lovable Product Ladder

Replace the MVP with a three-rung ladder: minimum lovable, lovable, absolutely lovable.

Read →
Self-Mastery4 steps

One-Week Top-1% AI Fluency Sprint

Pick a real problem and spend one full week taking it end-to-end with AI — that alone puts you in the global top 1%.

Read →
Productivity3 steps

Precision Prompting for AI Builders

Never tell the AI 'it doesn't work' — state exactly what you expected and which parts do and don't.

Read →
Innovation3 steps

Product-First, Add-AI-Later

Design the end-to-end product experience first, then add AI to solve specific problems — never retrofit AI into a broken flow.

Read →
Innovation3 steps

Stuck-Point Scaling Law

Reliably improve an AI system by hunting where it gets stuck and tuning those spots with a fast feedback loop.

Read →

People & resources mentioned

Attributed to the moment in the episode. Timestamps are approximate.

People · 4

  • Anton OsikaMentionsAnton is co-founder and CEO of lavable

    “today my guest is Anton OC Anton is co-founder and CEO of lavable”

    ▶ Watch at ~1:00approx.
  • Lenny RachitskyMentionstoday we've got another very special compilation episode something I've been pulling on more and more with the podcast and the newsletter

    “it's a pleasure to talk to you Lenny great to be here”

    ▶ Watch at ~5:00approx.
  • FabianMentionsa previous colleague of mine who has also been a Founder Fabian

    “I grabbed a previous colleague of mine who has also been a Founder Fabian”

    ▶ Watch at ~25:00approx.
  • Amjad MasadMentionsamjad is the co-founder of repet an AI powered software development and deployment platform for building and shipping software

    “when we had omj on the podcast from replate he said that this is like the main skill”

    ▶ Watch at ~28:30approx.

Resources · 22

  • Lenny's NewsletterMentionsnewsletter · Lenny Rachitsky

    “if you become a yearly subscriber of my newsletter you now get a yearfree of perplexity and notion and superhuman and linear and granola”

    ▶ Watch at ~2:30approx.
  • Product HuntUseswebsite · Ryan Hoover

    “his company he like launched on product hunt and everything”

    ▶ Watch at ~9:00approx.
  • Lenny's PodcastCoinedpodcast · Lenny Rachitsky

    “I rarely do demos on this podcast”

    ▶ Watch at ~9:30approx.
  • AirbnbMentionscompany · Brian Chesky, Joe Gebbia, and Nathan Blecharczyk

    “okay so we're gonna make our own Airbnb”

    ▶ Watch at ~10:00approx.
  • SquarespaceMentionssoftware

    “you can edit it visually like you can like you do in square space and Wicks and so on”

    ▶ Watch at ~15:00approx.
  • WixMentionscompany

    “you can edit it visually like you can like you do in square space and Wicks and so on”

    ▶ Watch at ~15:00approx.
  • CloudflareUsescompany

    “it's hosted by like a cloud vendor which is hosting I think a huge chunk of the Internet it's called Cloud flare”

    ▶ Watch at ~16:30approx.
  • ChatGPTMentionssoftware · OpenAI

    “when CH gbt came out they were starting to get really good at taking a human instruction and spitting out code”

    ▶ Watch at ~22:00approx.
  • GPT EngineerCoinedsoftware · Anton Osika

    “I created a open source tool called GPT engineer”

    ▶ Watch at ~22:30approx.
  • GitHubMentionssoftware · GitHub

    “we started creating 15,000 projects per day”

    ▶ Watch at ~23:30approx.
  • Microsoft CopilotMentionssoftware · Microsoft

    “there's a lot of companies using a to make Engineers more productive Microsoft you build co-pilot and so on”

    ▶ Watch at ~25:00approx.
  • ReplitMentionssoftware · Replit

    “when we had omj on the podcast from replate”

    ▶ Watch at ~28:30approx.
  • StripeUsescompany · Patrick Collison and John Collison

    “specifically adding login creating data persistence adding payment with stripe”

    ▶ Watch at ~29:00approx.
  • BoltMentionssoftware · StackBlitz

    “what's the simplest way to understand maybe how Lev B might be different from say bolt and ret”

    ▶ Watch at ~34:30approx.
  • GitHubUsessoftware · GitHub

    “something that's different from I think the other other tools is that it's it is synchronized with GitHub”

    ▶ Watch at ~35:00approx.
  • OpenAIUsescompany

    “already today you can just ask add a chat with open Ai and then you get a chat with open AI”

    ▶ Watch at ~37:00approx.
  • FigJamUsessoftware · Figma, Inc.

    “we have like a fig Jam board where we have all the main problems”

    ▶ Watch at ~50:00approx.
  • LinearUsessoftware · Linear

    “we do so many things in our company in linear because it's just such amazing product”

    ▶ Watch at ~51:30approx.
  • Sana LabsMentionscompany

    “I was the first employee at an AI startup here in stockolm called sonal labs”

    ▶ Watch at ~1:01:30approx.
  • DuolingoMentionssoftware · Luis von Ahn and Severin Hacker

    “some some way to learn where be it like English thing du lingo”

    ▶ Watch at ~1:02:30approx.
  • ClaudeRecommendssoftware · Anthropic

    “ask that tool or ask Claude B about how how how that topic that domain works”

    ▶ Watch at ~1:07:30approx.
  • DiscordUsessoftware · Discord Inc.

    “we have a Discord where you can share like oh this is how I use loveable”

    ▶ Watch at ~1:08:30approx.

Spot an error or want something removed? Request a correction or removal.

Insights & moments

The myth-busts, hot takes, explainers, and tools worth keeping.

Hot Take· 2

Hot Take38:00

What Gets More Valuable: Taste and Being a Generalist

As building gets automated, Anton argues the scarce skills shift to figuring out what to build and having the taste to judge whether it's good. Engineering knowledge stays valuable for understanding constraints, but people should abstract up a level and see themselves as translators of problems into technical solutions. Being a generalist matters much more than it used to.

  • Figuring out what to build and taste become the scarce skills
  • Engineers should reframe as translators of problems to solutions
  • Technical skill still matters for understanding constraints
  • Being a generalist is much more important than before

“taste and like refine tasting what is what is good is even more of the important part”

Anton Osika · 39:30

“doing a bit of everything being in generalist is I think much more important than it used to be”

Anton Osika · 40:30
#skills#future-of-work#hiring#ai
Hot Take53:30

The Underrated Leverage: Eating Lunch Together

Asked what else lets a tiny team move so fast, Anton points to working from the office and, specifically, eating lunch together as a highly productive hour of cross-pollination. It's a strikingly human answer from the CEO of a cutting-edge AI company, balancing focused solo work with high-bandwidth unstructured communication.

  • Team works from the office most of the time
  • Shared lunch is a productive hour of cross-pollination
  • Balances focused work with high-bandwidth informal communication

“eting lunch together is a pretty productive hour uh where you're cross-pollinating”

Anton Osika · 54:00
#team#culture#productivity

Explainer· 3

Explainer06:00

What Lovable Is: Your Personal AI Software Engineer

Anton frames Lovable as an AI software engineer that turns an English prompt into a fully working product, aimed at the 99% of people who don't write code. He describes the long-term goal as building 'the last piece of software' anyone will ever need to write, because it can create all future products.

  • You describe an idea and get a fully working product back
  • Built for the 99% of the population who don't write code
  • Entrepreneurs, designers and PMs use it to ship first versions
  • Framed as 'the last piece of software' ever written

“I'd say lovable is your personal AI software engineer you describe an idea and then you get a fully working product”

Anton Osika · 06:00

“we say we say we're building the last piece of software”

Anton Osika · 07:00
#ai#product#no-code#vision
Explainer27:00

The Scaling Law: Making AI Not Get Stuck

Anton describes a scaling law where putting in more work reliably makes the product better. AI systems tend to start strong then get stuck, so the team painstakingly identifies the places it fails, tunes the whole system quantitatively, and builds a fast feedback loop. They prioritized never getting stuck on the highest-value flows like login, data persistence and Stripe payments.

  • More work reliably yields a better product (their scaling law)
  • AI starts strong then gets stuck; they find and fix those points
  • System tuned quantitatively with a fast feedback loop
  • Prioritized flows: login, data persistence, Stripe payments
  • Getting AI unstuck is a fading problem, not a permanent one

“when you put in more work this the pro the product reliably gets better and better”

Anton Osika · 27:00
#ai#engineering#scaling#reliability
Explainer41:30

How Lovable Hires: Care, a Superpower, and a Week-Long Work Trial

Anton says the most important hiring signal is that people genuinely care about the product, users and team rather than treating it as a job. He looks for a generalist brain paired with one absolute superpower dimension. Beyond a hard unorthodox problem, he almost always runs a work simulation of at least a day, often a full week.

  • Care or obsession about product, users and team is the top filter
  • Wants a generalist who is also world-class in one dimension
  • Asks candidates about work they cared deeply about
  • Runs a work-simulation trial of at least a day, often a full week

“everyone should really care about the product the users and Care a ton about the team”

Anton Osika · 42:00

“I pretty much always have people join the work simulation for at least a day often a full week”

Anton Osika · 44:00
#hiring#team#culture

Story· 5

Story07:30

The Numbers: 300K Users and $10M ARR With 15 People

Anton shares the scale Lovable reached in under three months: 300,000 monthly active users, 30,000 of them paying, and revenue growth of roughly a million ARR per week. The company hit $4M ARR in the first four weeks and $10M in two months with just 15 people, growing almost entirely through organic word of mouth.

  • Launched less than three months before the interview
  • 300,000 monthly active users, 30,000 paying
  • $4M ARR in first four weeks, $10M ARR in two months
  • Growth almost entirely organic word of mouth
  • Had to rewrite the entire codebase as it scaled

“we launched lavable less than three months ago and now we have 300,000 monthly active users and 30 of those 30,000 of those are actually…”

Anton Osika · 07:30
#growth#revenue#metrics#startup
Story10:00

Building an Airbnb Clone Live in 30 Seconds

In a live demo, Anton builds an Airbnb clone from the two-word prompt 'Airbnb clone.' Lovable generates a fully interactive UI with categories, listings and login buttons. He notes the first prompt takes about 30 seconds, then iterates by adding a purchase flow and connecting a Supabase backend for real data.

  • Two-word prompt 'Airbnb clone' generates a full interactive UI
  • First prompt renders in about 30 seconds
  • Output is a functioning site, not just a static design
  • Backend (Supabase) connects with roughly one click for real data
  • Deploys to Cloudflare hosting

“the the first prom takes 30 seconds 30 seconds”

Anton Osika · 11:30
#demo#product#no-code#prompting
Story22:00

Origin Story: From GPT Engineer to Lovable

Anton recounts building the open-source tool GPT Engineer to prove that large language models could turn instructions into working code, after colleagues doubted him. It became the most popular open-source demo of LLM app generation with 50,000+ GitHub stars. GitHub even shut Lovable down once, mistaking 15,000 projects created per day for an attack.

  • GPT Engineer built to prove LLMs could turn instructions into code
  • Reached 50,000+ GitHub stars and dozens of academic references
  • Motivation: enable non-coders, not just make engineers faster
  • GitHub shut them down once, thinking 15,000 projects/day was an attack

“I created a open source tool called GPT engineer where you you write something like create a snake game and then it spits out a…”

Anton Osika · 22:30
#origin-story#open-source#ai#founding
Story45:00

The Shackleton-Style Job Posting as a Filter

Lenny reads from Lovable's job description, which echoes Ernest Shackleton's famous ad: long hours, high pace, thriving under AGI-timeline urgency, honor in success, and no comfort-seekers. Anton confirms he wrote most of it, and explains that a deliberately intense posting is a filter that repels the wrong people and attracts exactly the ambitious ones he wants.

  • Job posting deliberately modeled on Shackleton's intense recruiting ad
  • Emphasizes long hours, high urgency and AGI timelines
  • 'Those seeking comfortable work need not apply'
  • Intensity acts as a filter to attract the right, ambitious people

“long hours High Pace candidates must Thrive under high urgency under AGI timelines approaching”

Lenny Rachitsky · 45:00
#hiring#culture#recruiting
Story01:01:30

Failure Corner: Don't Retrofit AI Onto an Existing Product

Anton shares a product lesson from his first job at Sana Labs, where they built a personalized-learning AI API that other companies had to bolt onto their existing products. Retrofitting AI, like swapping out an engine, worked poorly. The lesson: start from how the product works end to end, then decide where AI solves a specific problem.

  • Sana Labs sold a personalized-learning AI API to bolt onto other products
  • Retrofitting AI onto existing products worked poorly
  • Start from the end-to-end product and user experience first
  • Add AI to solve specific problems, not as the starting point

“you have to start with like how is this product working end to end and then add AI”

Anton Osika · 01:03:00
#failure#product#ai#lessons

Q&A· 1

Q&A34:30

How Lovable Differs From Bolt and Replit

Asked how Lovable differs from Bolt and Replit, Anton points to packaging for non-technical people: you can edit text and colors instantly without opening a code editor or waiting 30 seconds for the AI. It also syncs with GitHub so technical teammates can drop into Cursor. Above all, he argues Lovable is ranked most reliable at not getting stuck.

  • Packaging aimed squarely at non-technical users
  • Edit text and colors visually and instantly, no code editor needed
  • GitHub sync lets technical teammates work in Cursor
  • Reliability (not getting stuck) is the core differentiator

“no other tool out there lets you generate code from an AI engineer and then actually just like change a small element of it”

Lenny Rachitsky · 15:30

“not getting stuck is I I think the most important thing for people”

Anton Osika · 35:30
#competition#product#positioning

Takeaway· 4

Takeaway19:00

The Pro Tip: Be Specific, Don't Just Say 'It Doesn't Work'

Anton's core tip for getting value from AI tools is precise communication. Instead of vague complaints like 'it doesn't work,' users should state exactly what they expected and which parts are and aren't working. He notes this clarity matters even more with AI than with humans, and it's a skill most people don't do naturally.

  • Explain exactly what you expected, not just that something failed
  • Say which parts work and which parts don't
  • Clear communication matters more with AI than with humans
  • This is a strong product-manager skill many people lack

“don't say it doesn't work just explain exactly what you're expecting and which parts are working and which parts are not working”

Anton Osika · 20:00
#prompting#skills#product#ai
Takeaway21:00

Minimum Lovable Product: The New MVP

Anton explains the company's name comes from his belief that the best word for a great product is 'lovable.' He uses the jargon of building a minimum lovable product, then a lovable product, then an absolutely lovable product, reframing the classic MVP around whether users actually love what you built.

  • 'Lovable' is the best single word for a great product
  • Progression: minimum lovable to absolutely lovable product
  • Reframes the MVP concept around user love, not just viability

“the best word for a great product is that it's lovable”

Anton Osika · 21:00

“building a minimum lovable product and then building lovable product”

Anton Osika · 21:00
#product#branding#philosophy
Takeaway48:00

Prioritization: Just Solve the Biggest Bottleneck

Anton's prioritization approach is a deliberately simple algorithm: identify the single biggest bottleneck or product problem, solve it well, then move to the next, without over-planning a long roadmap. He notes the hard part is figuring out what the biggest problem actually is, which comes from talking to users and reading requests, and that the process is engineering-led.

  • Find the biggest bottleneck, solve it, then pick the next one
  • Avoid dreaming out a long roadmap
  • Hardest part is identifying the real biggest problem
  • Process is engineering-led rather than PM-led

“identifying what is the biggest ball neck what the biggest produ problem and iterating fast”

Anton Osika · 48:00
#prioritization#product#process
Takeaway01:05:30

How to Reach the Top 1% at Using AI Tools

Anton's closing advice is that the best thing you can do for your career is to be in the top 1% at using AI tools. The heuristic: spend a full week taking one real problem end to end, from idea to something someone actually uses, asking the AI whenever you don't understand. Surround yourself with obsessed friends and you reach the top 0.1%.

  • Being top 1% at using AI tools is the best career move
  • Spend a full week taking one problem from idea to real usage
  • Ask the AI whenever you don't understand something
  • Obsessed peers accelerate you to the top 0.1%

“the best thing you can do for your current profession or if you want to have a new job is to be in the top…”

Anton Osika · 01:05:30

“if you spend a full week on trying to reach an outcome the best way to learn is like I want to do this thing…”

Anton Osika · 01:06:00
#skills#ai#career#advice