Attio is the CRM that transforms how top teams grow. Ask Attio how to win this deal, what to say in your follow-up, where your pipeline actually stands, and get answers with full context, instantly. With Universal Context, Attioās intelligence layer, every signal across your tools becomes instantly actionable.
The engine behind every top 1% startup is a great CRM.
One of the tools that fascinates me most is Lovable, both because of how fast their product ships (daily?) but also for how spectacular it is to enable creativity (i.e. meet my friend Arnold) and give you what genuinely feels like superpowers.
So I decided to deep dive into how this European outlier grows.
Hereās what actually worked, stage by stage:
š Quick note on methodology: 40+ sources including podcasts, company blogs, investor analyses, and third-party teardowns. All revenue figures from TechCrunch or public statements. This company moves so fast that data from 3 months ago is already outdated, I used the latest confirmed source in every case. Treat all numbers as directional.
Act 1, The Spark: $0ā$10M
Founder brand as a permissionless acquisition channel
Anton Osika studied engineering physics at KTH, worked at CERN, was the first engineer at Sana Labs, and co-founded Depict.ai through YC. In 2023 he built GPT Engineer as a side project, an open-source experiment to see what LLMs could do with code.
It hit 50K GitHub stars which was the fastest-growing repository on GitHub that year. GitHubās systems temporarily shut it down thinking it was a DDoS attack š , because it was spawning 15,000 new repositories a day.
He tried to productize it twice.
Launch 1 (early 2024): user growth stalled almost immediately.
Launch 2 (mid-2024): retention collapsed, the AI got stuck on anything complex, and the pricing model was bleeding money on power users.
Most people read that as a no, but Anton read it as a spec.
He and co-founder Fabian Hedin migrated the entire backend from Python to Go, built new scaling laws for AI self-debugging so the model could get itself unstuck, and made the call that changed everything: stop targeting developers.
Their hypothesis was that the real opportunity was giving non-technical people the ability to build software from scratch (superpowers).
āLovableā as a rebrand was spot on, as it told you how it should feel. Elena Verna, who joined later as Head of Growth, has a name for this: the Minimum Lovable Product, software with personality, software you trust because it feels alive.
The name has also become somewhat of a cultural enforcement mechanism inside the company, when someone says āthis is not lovableā in a meeting, everyone stops what theyāre doing.
Third launch was in November 2024, which became #1 on Product Hunt, #1 on Hacker News and drove $10M ARR in 2 months with 15 people.
The levers that drove it:
Founder brand as distribution: Anton had already been cultivating an audience on socials built over years of posting raw numbers, failed experiments, hiring photos, half-formed ideas a-la Pieter Levels. When hey launched people already trusted the person before they trusted the product. This deserves a deep-dive on its own, worth having a trip down memory lane on Antonās timeline.
GPT Engineerās 50K GitHub stars were a pre-built trust base: Thousands of developers already knew the name and the credibility likely transferred directly to the new product, again part of a long-term founder brand investment.
Short video demos as an ad: Nobody knew what vibe coding was in late 2024 but you watch 10 seconds of someone building a real app by typing a sentence and you think āI had no idea this was possible.ā / what kind of sorcery is this. I think Twitter, Tiktok and YouTube drove real curiosity at scale due to the productās inherent āmagic momentsā.
Co-marketing with Supabase, Replicate and Resend: gave them reach into audiences that were already product-curious and technically adjacent.
Product Hunt and Hacker News as coordinated launch amplifiers: as usual, but also with leveraged via points 1 and 2 above.
User-generated content from day one: Every post someone shared of their Lovable build was basically an organic proof point (PLG mechanics).
One mistake Anton calls out directly that caught my eye and is worth highlighting as it is a perpetual debate I had both working at Meta on product discussions but also with many of my founder friends (almost weekly):
On trying to keep the GPT Engineer open-source community alive alongside Lovable: āA bad idea to do two things that were like a bit too tangentially related.ā
