Why This YC Startup Is Betting That AI Will Get Things Wrong

How Alessia and Elia did it: Why This YC Startup Is Betting That AI Will Get Things Wrong

Before VibeFlow, Alessia and Elia were both separately tinkering with ideas and going through co-founder matching. Elia had joined EF, Entrepreneur First, in Paris — a program to do co-founder matching where they "celebrate break up" so you don't lose time and figure things out fast. He met around 60 people in three days and at one point realized he couldn't find anyone in the same mentality as him: super hardcore, wanting to work seven days a week, super ambitious. Alessia was working full time for a company, always enjoying her side projects, but none of the people she worked with were willing to go full time on a startup — so she took a leap of faith and quit her job, then tried the YC co-founder matching platform — described as "like a dating platform... like a tinder or Bumble."

They went through this process together — Elia in Paris, Alessia in Sri Lanka — doing the exact same thing and calling every day as friends, complaining about not finding the right match. At one point they just said: but why are we not brainstorming together at least. They decided to brainstorm and directly ended up on an idea. It wasn't VibeFlow yet, but they saw the team was working well. Their lesson from that period: the co-founder is the most important thing, especially in an early-stage startup, because the idea is going to change — the right way in is personal fit, the way of working, and the values, and the rest will come together.

Both are technical: Elia has a background in computer science and AI specialized toward biotech, Alessia in computer graphics and computer vision with more than four years of software engineering experience. They started in health, looking at a very niche problem around the microbiome — deep tech, biotech — got some initial traction, then came back to what they know best: AI and data tools. At the moment Lovable exploded, they saw so many complaints online — everyone was super mad on social — and they tried to understand what was happening. They built the first version of VibeFlow within a month, with no AI: people could import their UIs from tools like Lovable and build the backend manually with a drag-and-drop interface, all in one solution instead of the "Frankenstein approach" of stitching three different platforms and subscriptions together. They added the AI element, launched on Product Hunt, won product of the day, and hit 6,000 users in three weeks.

The product thesis is reliability. All the existing solutions are pretty good at the first generation, but they don't tackle maintaining and scaling these applications — the need non-technical people have. VibeFlow's AI doesn't touch the raw code directly; you iterate with the AI on the logic with a visual feedback of what's being built, and once you agree on the logic the code is generated deterministically. That avoids the loop where you let the AI generate code, it doesn't work, you go around for four days, lose time and money and get super frustrated. They focus on the backend — a very hard problem a lot of people aren't tackling — and on deterministic code for workflows like connecting multiple apps, where you want something to always happen, not have an AI decide whether to do it.

They built and launched continuously — the internal YC platform, then YC, then Product Hunt, then Hacker News and Reddit — treating launching as a continuous feedback loop rather than one huge launch. When they got into YC they moved from Zurich and Paris all the way to SF within days, started living together and working 24/7 at a crazy pace. For them YC's biggest value was the network over the badge or the mentorship, and SF itself — where everyone believes you can do anything — accelerated the startup ten times compared to Europe. Their most-repeated lesson: launch fast and often, do things that don't scale, solve a problem for one person who comes back every day and loves your product, and keep being more ambitious. As Elia puts it, if you're not talking to users or building, what are you doing.

What you'll hear

• Co-founder dating hell — both tried co-founder matching separately, complained to each other daily, then asked why they weren't just building together
• The co-founder is the most important thing — prioritize personal fit, way of working and values over the idea, because the idea will change
• From microbiome to AI tools — starting in health and deep-tech biotech, then coming back to what they know best
• Built in a month, no AI — the first version let people import UIs and build the backend manually before any AI was involved
• Product Hunt and the continuous launch — winning product of the day, hitting 6,000 users in three weeks, and launching again and again across platforms
• Why AI shouldn't touch the raw code — iterating on logic visually, then generating deterministic code so it actually works
• Network over badge — why YC's network beat the logo and the mentorship, and how SF made them ten times more ambitious

Key claims from this episode

Chapters

Quotes from this episode

Themes Alessia returns to

• Co-founder first — the co-founder is the most important thing in an early-stage startup; pick on personal fit, way of working and values, not the idea
• Launch fast and often — a continuous feedback loop across platforms beats one huge launch; you're never gonna feel ready anyway
• Reliability over generation — go beyond the first generation by generating deterministic code instead of letting AI rewrite everything
• Do things that don't scale — solve a problem for one person who comes back every day and loves your product, then expand
• Environment matters — San Francisco makes you believe you can do anything and accelerates the startup ten times; Europe is hard mode
• Stay hungry — keep pushing, do a roadmap for two weeks not a year, and be way more ambitious on the goals