Anton Osika: How He Grew Lovable to $400M ARR in 15 Months

Lovable went from a 50,000-star open-source repo and two failed launches to $1M ARR in 8 days — and $400M ARR in 15 months. The exact playbook, and the three preconditions you can't copy.

TL;DR

• Lovable hit $1M ARR in 8 days, $100M in 8 months, and $400M in 15 months — the fastest path to $100M ARR ever publicly recorded for a software company.
• The launch that detonated was the third launch. The same core product had already failed twice in 2024 under the name "GPT Engineer App." What changed was the name, not the code.
• The latent phase was pre-paid: founder Anton Osika's open-source project gpt-engineer hit ~40,000 GitHub stars in roughly three months in 2023 — building a ~27,000-person audience before Lovable the company existed.

The Numbers

At the July 2025 Series A snapshot Lovable reported 2.3M users against roughly 180K paying — about a 2.25% free-to-paid conversion rate. By March 2026 the company ran on 146 employees, which works out to $2.77M ARR per employee — above every public AI-unicorn benchmark. Funding followed the revenue: a $7.5M pre-seed, then a $200M Series A at $1.8B, then a $330M Series B at $6.6B.

What they did differently

Move 1: Build the audience before you build the company

Most startups spend 18–24 quiet months manufacturing demand. Lovable skipped that phase because it had already happened — under a different name.

In June 2023, while still CTO at Depict.ai, Anton Osika spent a weekend building gpt-engineer, a CLI tool that turned text prompts into codebases, and pushed it to GitHub. The repo crossed 40,000 stars in about three months and eventually passed 52,000 — one of the fastest-growing repos of 2023. By the time Lovable AB incorporated in November 2023, the company already had three assets almost no pre-seed startup has: public proof that demand for the category existed, a watching audience of ~27,000 people who had starred or installed the tool, and a founder who didn't need to introduce himself in developer circles.

On Lenny's Podcast, Osika later framed Lovable's capital efficiency as roughly $2M of capital deployed by the time the company hit $1M ARR. That number is only possible because the expensive part — earning an audience's attention — was paid for in GitHub commits a year earlier.

Move 2: When the brand attracts the wrong buyer, rename — don't out-market it

Here is the part most coverage skips: Lovable launched commercially twice in 2024 and failed both times, under the name "GPT Engineer App."

The product was sound. The name was the problem. "GPT Engineer" said developer tool, OpenAI-dependent, command-line. The product Lovable actually wanted to sell said anyone with an idea, multi-model, point-and-click. Non-technical users self-deselected on the name alone, and every channel that had worked for the open-source project (Hacker News, dev-Twitter) reached exactly the wrong audience for a consumer product.

The team's fix wasn't a bigger marketing budget. It was a rebrand to "Lovable" plus multi-provider model support that decoupled the product from the GPT name. The relaunched product on November 21, 2024 was substantially the same software that had failed twice. This time it produced the steepest launch curve any SaaS has publicly disclosed — and Osika narrated it in real time on X:

$4M ARR in 4 weeks. The fastest-growing startup in Europe.

— Anton Osika, X, December 2024

Move 3: Never let a milestone fire alone

From early 2025 on, Lovable ran one repeatable play: bundle funding + an ARR milestone + an enterprise customer name into a single news cycle, then run the milestone again as a standalone story a week later.

The Series A is the textbook example. TechCrunch leaked the round two days early. On announcement day, the $200M raise shipped alongside $75M ARR, 2.3M users, Klarna and HubSpot as named customers, and an angel list that included the CEOs of Klarna, Slack, and HubSpot. Six days later, TechCrunch ran the $100M ARR milestone as its own article — a one-two punch that kept Lovable in the press for ten straight days.

The category timing made all of it land harder. When Andrej Karpathy posted his now-famous note coining the term in February 2025 — a post that drew 4.5M+ views — Lovable was the most consumer-positioned tool in the exact category he had just named:

There's a new kind of coding I call "vibe coding," where you fully give in to the vibes and forget that the code even exists.

— Andrej Karpathy, X, February 2025

Five funding-and-milestone bundles in 13 months, each on the same template: leak early, bundle on the day, follow up with the standalone milestone. Same announcement budget, three times the coverage surface.

What you can copy

1. Ship a radically free wedge before you have a paid product. An open-source repo, a free tool, a public dataset — anything that earns an audience while you have nothing to sell. The audience is the asset; the product is downstream of it.
2. If your brand attracts the wrong buyer, rename it. Two failed launches under the wrong name cost Lovable less than three years of marketing spend would have. A name that excludes your real customer cannot be fixed with budget.
3. Never announce funding alone. Bundle every raise with a revenue number, a named customer, and ideally a product update. One announcement, several headlines.
4. Leak, then bundle, then follow up. Give one outlet the round two days early to build anticipation; bundle the news on the day; run the ARR milestone as a separate story 5–10 days later.
5. Put the founder on one long-form podcast — with real numbers. Osika's Lenny's episode ("$10M ARR in 60 days with 15 people") did more than fifty short posts. Long-form rewards founders who bring concrete figures and punishes those who don't.

What probably won't work for you

The playbook is real, but three preconditions made it available to Lovable and not to most teams.

You cannot manufacture a 50,000-star repo on demand. gpt-engineer's stars are the load-bearing asset in this entire story. Without a pre-built audience, the post-launch curve looks nothing like Lovable's. If you don't already have organic distribution, the lesson is what to build toward — not a tactic you can run next quarter.

The category timing was a gift, not a skill. Karpathy's "vibe coding" post created a consumer-grade frame for the whole category. Launch six months earlier and the wave isn't there; launch six months later and a competitor owns the consumer slot first. You can position to catch a wave, but you can't schedule one.

The headline numbers hide unknowns. A 2.25% free-to-paid conversion rate is healthy but not extraordinary, and vibe-coding products typically carry 40–60% gross-margin drag from LLM tokens. A $400M ARR figure says nothing about retention or unit economics. Copy the distribution moves; don't assume the underlying business is as clean as the press cycle.

Sources & references

• gpt-engineer on GitHub — the open-source project that pre-built the audience
• Rebranding GPT Engineer to Lovable — Lovable's official rebrand post
• Anton Osika's "$4M ARR in 4 weeks" thread — X, December 2024
• Building Lovable — Anton Osika on Lenny's Podcast
• Andrej Karpathy's "vibe coding" post — X, February 2025
• Lovable becomes a unicorn with $200M Series A — TechCrunch
• Lovable crosses the $100M ARR milestone — TechCrunch
• Lovable added $100M in revenue in a single month — TechCrunch
• Full timeline: The Lovable Growth Story — every funding round, product launch, and milestone, charted

This case study is part of GrowthHunt's growth teardown series. See the Cursor teardown for the other end of the vibe-coding curve, or track the fastest-growing AI repos and founders live on GrowthHunt Velocity.

More from the growth playbook.

See it run in the wild.