JOSS SCHOLTEN

The Idea

After Yonomi I wanted to build something in a completely different space. A co-founder and I kept coming back to the same observation: generative AI was about to fundamentally change the way people cook — not just recipes, but the entire relationship between a person, their pantry, their dietary needs, and their kitchen appliances.

The existing recipe apps were static. Content was generic. Personalization was fake. Nobody had built a truly AI-native cooking experience yet. Ladle was our attempt to do it right before anyone else did.

What We Shipped

The core of Ladle was a generative AI personalization engine that could adapt any recipe in real time — swapping ingredients based on what you have, adjusting for dietary restrictions, scaling servings, translating across languages, even adjusting cooking times for altitude or equipment.

• Generative AI engine adapting recipes across dietary needs, restrictions, and preferences
• API-first architecture supporting both consumer apps and enterprise brand integrations
• Smart kitchen appliance integrations — recipes that talk directly to your oven
• Enterprise APIs for food brands to embed AI-powered recipes in their own products
• Full concept-to-MVP in 4 months; full product within the first year

Building Fast

Going from blank document to working product in 4 months required making fast, high-quality architecture decisions under uncertainty. I led engineering and product simultaneously — designing the AI systems, building the team, managing cloud infrastructure, and keeping investors and partners aligned.

The hardest calls were the product ones: which personalization dimensions mattered most to users, how to handle AI-generated content users didn't trust, when to let the AI surprise people vs. when to ask first.

What I Learned

Building AI-native products is a fundamentally different exercise from traditional SaaS. The hardest problems aren't engineering — they're product. How do you make an AI feel like a collaborator rather than a black box? How do you build trust with users who are skeptical of AI-generated content? How do you define quality when output is generative by nature?

I came out of Ladle with much sharper answers to those questions — and a clearer sense of what it takes to build AI products people actually rely on.