Mar 5, 2025 · 49m · 20vc
Anton Osika, Co-Founder and CEO @ Lovable: Hitting 85% Day 30 Retention - Better than ChatGPT · 20VC with Harry Stebbings
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
This episode of the 20VC podcast features an in-depth interview with Anton Osika, Co-Founder and CEO of Lovable, who shares the strategic decisions, product principles, and hiring frameworks that drove his AI code-generation startup to astronomical growth and exceptional user retention.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Harry holds 33.2% of the talking time here. How this is scored →
Harry as informed peer 3.6 Guest teaching 3.4 Guest disagreement 2.1 Harry pushing back 3.0
speaking balance: gold is Harry, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 37:55 Anton maintains short stance on OpenAI despite host pressure
When Harry forcefully defends OpenAI using consumer brand dominance and front-end reach, Anton refuses the premise and counters that OpenAI lost its key talent to Anthropic.
Hardest push from Harry ▶ 37:28 Harry directly challenges guest's thesis on OpenAIHarry explicitly interrupts and challenges Anton's claim that OpenAI will struggle, pointing to its unequaled consumer brand recognition and front-end product leadership.
Biggest teaching moment ▶ 32:30 Anton refutes 'AI sugar revenue' narrative with retention metricsWhen Harry asks if Lovable's rapid growth is low-quality 'AI sugar revenue' prone to churn, Anton educates him by citing an 85% month-one retention rate that beats ChatGPT.
Harry holds his own ▶ 37:28 Harry uses market distribution arguments to counter guestHarry demonstrates strong market expertise by arguing that brand equity and consumer front-end UI are the two critical moats in this AI wave.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Harry as informed peer | Guest teaching | Guest disagreement | Harry pushing back | Why |
|---|---|---|---|---|---|---|
| Key Lessons and Takeaways from Depict | 4 | 2 | 1 | 2 | Harry references Depict investors and quotes Gmail creator Paul Buchheit regarding product simplicity. Anton agrees with the principle of saying no to features and shares lessons on focus. | |
| Hiring Strategy: Talent over Experience | 4 | 3 | 2 | 3 | Harry cites Creandum investor insights and asks probing questions about hiring inexperienced talent. Anton reframes experience as potentially negative due to fixed habits. | |
| The Genesis of GPT Engineer | 2 | 2 | 1 | 1 | Anton recounts the origin story of GPT Engineer built over a caffeinated weekend. Harry asks standard narrative questions about early product advice. | |
| Recruiting the Perfect Co-Founder | 2 | 2 | 1 | 1 | Anton describes recruiting his co-founder and transitioning out of Depict. Harry asks about waitlist strategy lessons in a collaborative tone. | |
| Conducting Effective User Interviews | 3 | 3 | 2 | 3 | Harry asks how AI alters team structures and questions whether chat prompts are the right default UI. Anton defends the prompt UI while acknowledging future complexity. | |
| Why Lovable Rejected Y Combinator | 4 | 3 | 2 | 3 | Harry questions Anton on rejecting Y Combinator and asks about market trends regarding founder dilution sensitivity. Anton argues focus and speed outweighed YC value. | |
| Going Live and Managing Explosive Growth | 3 | 4 | 1 | 2 | Anton drops massive growth metrics of $2M ARR added per week. Harry reacts with astonishment at the extreme velocity compared to typical venture-backed benchmarks. | |
| Product Simplicity and the 'Build It and They Will Come' Myth | 3 | 3 | 2 | 2 | Harry asks about common product velocity bottlenecks. Anton notes that internal community features were wasted effort during hypergrowth and reframes the 'build it and they will come' myth. | |
| Founder Energy and Management Mistakes | 4 | 4 | 2 | 3 | Harry cites Revolut founder Nick Storonsky on optimal founder age and asks about early management errors. Anton acknowledges making mistakes by trying to hire executives too early. | |
| Maintaining Startup Culture While Scaling | 4 | 3 | 2 | 3 | Harry asks how Anton will handle well-funded US competitors outspending them on talent and marketing. Anton responds that execution is the only true competitive moat. | |
| Building in Europe vs. Silicon Valley | 4 | 4 | 4 | 5 | Harry pushes the common Valley narrative that staying in Europe harms career upside. Anton counters that Europe offers superior raw talent and enjoys playing on hard mode. | |
| Tackling Retention and the AI Churn Problem | 5 | 6 | 4 | 6 | Harry directly challenges Lovable's retention asking if it is temporary 'AI sugar revenue'. Anton schools Harry with data showing 85% month-1 retention, superior to ChatGPT's initial retention. | |
| Lovable's North Star Metric and Paying User Scale | 6 | 5 | 4 | 6 | Anton states he would short OpenAI and buy xAI/Grok. Harry aggressively pushes back citing OpenAI's dominant consumer brand and UI lead, but Anton holds his ground on talent drain and Anthropic enterprise growth. | |
| Navigating Megacorps and Shifting to Enterprise | 4 | 3 | 1 | 3 | Harry asks about megacorp distribution risks and shifting from prosumer to enterprise. Anton explains delaying enterprise to focus on empowering individual builders. | |
| Building Y Combinator into Lovable | 3 | 4 | 2 | 2 | Anton shares his ambitious product vision to build Y Combinator functionality directly into Lovable software. Harry acknowledges the feasibility via modern APIs. | |
| 20VC Quickfire Round | 3 | 3 | 3 | 3 | In the quickfire round, Anton delivers concise answers on model commoditization, leadership, and public markets. Harry shares a personal anecdote about early failure. |
Statements from this episode (32)
Osika: Early startups must say yes broadly, but scaling requires focus
“What works in the beginning Is to say yes onto a lot of opportunities and try out what works. Once you become more people and you have to follow up and maintain everything that you start, you have to be much more focused.”
Osika: Say no to most features to maintain Apple-like quality
“You should say no to as many things as possible and make it more of like an apple feeling. The things you do them superb.”
Osika: Domain experience can be a negative signal when hiring
“So I think experience can be a negative thing in some cases. You often want people who are super ambitious, they have a lot to prove, and they are more open-minded towards how, how you should work together in the team.”
Osika: Hire Top Young Talent Before They Become Founders
“The best people that you can hire at a young age, they would go on and become founders, and then you can't hire them anymore, right? So, so that's why junior people are often quite good.”
Stebbings: Successful tech founders typically made money early and excelled at gaming
“You fit, like, all the characteristics of, like, successful founders that I have on the show. It's like, number one, made money early, number two, excelling gaming both very, very clear archetypes.”
Osika: Putting LLMs inside a for loop enables autonomous AI agents
“Hook them up and make you basically put the large language model in a for loop and then you can have it do a lot of agentic things.”
Osika: Users are building revenue-generating SaaS companies entirely via Lovable prompts
“When we went on to iterate over the coming year, we packaged together all of these things so that you can ask today lovable, I want to build a basically a SaaS business, and then people have built their entire SaaS companies and get made money by just promptin…”
Osika: Product waitlists exist to control onboarding pace and conduct user research
“I think the waitlists are useful in that you can control exactly how many people you want to get on board and take user interviews with them.”
Osika: Product user research requires separating UX observation from pain-point discovery
“So they, I think for us, there are two different types of user interviews. We have this type where we just see them use the product and see, like, where do they understand the product? And so that's more of a user experience interview. And the other one is to …”
Osika: Having a team slows down founders shipping V1 SaaS products
“Hey, if you want to ship the first version of your SaaS and start making money, it's all you. It's not a team. Like a team just slows you down.”
Osika: AI can easily destroy existing codebases without engineer oversight
“But once you have existing software, With users, and then you want to iterate and change that software, AI might absolutely bring down, kind of mess up your entire codebase.”
Osika: Prompting Remains the Right Default UI for AI
“I think yes. Prompting can, you can do almost anything with prompting and explaining your thoughts in written form.”
Osika: Lovable rejected Y Combinator due to equity dilution and distraction
“We felt that at best YC would be a lot of dilution and some acceleration with a lot of dilution and some acceleration. And at worst it would be a distraction to go through SF and go through all of these, like, fun things that happen when you go to YC. And we j…”
Osika: Lovable expanded a $3M preemptive offer into an $8M pre-seed
“So we took three million and then we added on I got the advice to say like, yeah, just to get a lot of cash because you never know what happens in the markets. So we raised a quite a large pre-seed round with an up to almost eight million dollars.”
Osika: Founders should prioritize minimizing equity dilution over growing the pie
“Here again, like I have a very wise person who was like, no, dilution doesn't matter so much. It's all about the size of the pie. And I think that was, I was both affected by that advice as well as my, like, no, minimize dilution. This is my life, life's work.…”
Osika: Lovable wasted effort building early in-product community features
“I think at Lovable we thought a lot about the community, and like community features inside of the product, and the, that, I think that could have made a lot of sense if we had seen slower growth. But now that it's just like growth is not something we, not a c…”
Osika: Adding management layers at 40 employees was my biggest mistake
“Thinking that we should change the culture and become a more of a scale up something slow moving or like more management layers when we became 40 people. That's, that was the biggest mistake I did.”
Osika: Europe has more raw available tech talent than the US
“The most important thing is talent and culture, and it's, More raw, there's more raw available talent in Europe.”
Osika: Lovable's 85% month-one retention outperforms ChatGPT
“We have month one retention that's better than ChatGPT's month one retention on paying customers, and so it's about 85%.”
Osika: Lovable's North Star metric is apps hosted with active users
“It's the number of users that go all the way to getting, like, users on what they built. Getting, getting something hosted with users on what they built. That's what we focus on.”
Osika: Lovable has reached nearly 40,000 paying users
“So we have almost 40,000 paying users, and then that's the proxy for that.”
Osika: Anthropic's Claude is Lovable's main workhorse for code generation
“We use all of OpenAI's models, Google, Gemini, and the main workhorse is Anthropic's cloud model for writing the code.”
Osika: Buy Grok for business execution, short OpenAI on talent drain
“So I would buy Grok. I think they're also very ruthless in like finding business opportunities. We'll, we'll see about that, but I feel like they could be good at that. Even though Anthropic is my favorite, I love the culture and I love the, like the leadershi…”
Osika: Anthropic's enterprise revenue has almost caught up to OpenAI's
“If you look at the enterprise revenue, like, Anthropik is almost caught up to open AI from nothing, from being absolutely dominated”
Osika: OpenAI has lost all of its best talent to Anthropic
“OpenAI lost all their best talent to Anthropic”
Osika: Tech megacorps won't build the best products for years
“The big megacorps, they move very slowly in many domains. So they're never going to, in many years, they're not going to have the best product on the market.”
Osika: Giving AI models long-term memory and context will take years
“What you need to ramp is to decide what, like, the process for storing memory, I think is what you have to ramp and it has to store all of these things. Like my conversation with Harry has to be stored somehow into the system and all like my childhood almost h…”
Osika: Tesla is my 10-year public stock buy-and-hold pick
“Some talent play. What comes to mind is Tesla. They're also interestingly positioned outside of software.”
Osika: No single foundation model provider will win the AI market
“You don't need to be attached to one foundation model provider. They're all going to be amazing. You don't need to There's not going to be one winner here.”
Osika: Per-seat SaaS companies will collapse as AI replaces human employees
“I'm not so read, well read on the public market. It would be a SaaS company. It would be like a SaaS company that has a per seat pricing where those, their ICP is going to be replaced by AI SaaS that just replaces the employees. So the seat, the number of seat…”
Shorts cut from this episode
▶ The Scaling Nightmare Behind €2M ARR per Week 📈 · 20VC with (@19:52)
▶ Does the EU or US have better talent in startups? 🇪🇺🇺🇲 · (@0:16)
▶ Why the fastest growing EU startup turned down YC 😱 · 20VC (@16:04)
▶ "The only thing that matters is execution" 🎯 · 20VC with Ha (@0:22)
▶ The FASTEST growing company in Europe 📈 · 20VC with Harry S (@0:00)