Marc Torres

Stripe offered to acquire us for $1.2 billion when we had $2M in revenue. Today, we've raised $330M at an $8B valuation and reached $1B ARR. We could've died three times during this journey. This is the story I've never told anyone before:

In a world where everyone has access to the same AI models, one might think returns begin to converge across firms. But it’s clear that for now a real advantages you can have as a company, team, or individual is your ability to make AI successful. A big part of that is willingness to change processes and workflows to leverage AI agents. As well as your team or company’s ability to give agents the right context to work with. But getting them this context remains one of the biggest challenges for most companies. Data is fragmented, it’s in people’s heads, decisions and workflows aren’t written down, and so on. It’s a huge advantage if you’re starting from scratch and you can design your team or organization in a way that assumes you have to give agents this context and adapt your workflows to make agents work, instead of the other way around. But short of that, you just have to do the work to change your processes and have the right systems that work well for AI. Anyone that doesn’t figure this out will just be stuck with lower returns on AI.

Git Merge vs. Rebase vs. Squash Commit!

Last week, we shut down our startup. This is how AI hype pushed us to $100k ARR, and then killed us 6 months later. In 4,5 intense years, we raised $200k from incredible investors, went through pivot hell, and built Luca, an AI Running Coach on Whatsapp, hitting $100k ARR in a few months. Then we set out to build the platform for Athletes/Gyms/Influencers to launch their branded AI Coaches. We even had a deal with a sports legend. But soon after, we realized our AI bet was fundamentally wrong. Here’s what we learned 👇 1️⃣ If your product must be right 100% of the time, you'll have a hard time Even ChatGPT isn’t right 95% of the time - and it doesn’t matter. It still delivers huge value. We had a 95% success rate in all interactions... but that meant one wrong workout per month. By the second one, trust was gone. 👉 Build for use cases where an intern could do it - or 95% success feels great. 2️⃣ The market is everything (Marc Andreessen was right) We built complex workflows, guardrails, and prompt evaluations for 700 demanding users paying €10/month. 70k monthly interactions wtih 95% success rate. Still not enough. I bet most AI B2B SaaS products aren’t much different. B2B just pays 10× better and sticks around. 3️⃣ Don’t make “AI” your marketing engine Our hook worked while AI was new. Then hype faded… and it stopped working. 👉 Sell the solution, hide the AI Leading with “AI” creates unrealistic expectations you’ll never meet. Most users assume it’s perfect and that works against you. 4️⃣ Design for trust (reduce the black-box feeling) When AI generates images, video, or code, you see results instantly, no trust needed. With text-based products, show context, data, or sources to build credibility. 5️⃣ Don’t replace sympathetic jobs (this hardly applies to business owners cutting costs) Users and microinfluencers with personal trainer friends resisted trying Luca. 6️⃣ Go international early The US and English-speaking markets write the playbooks. By the time something reaches Europe, it’s already crowded. Our TikTok/IG microinfluencer strategy worked great in Spain ($4 CAC). But abroad, the channel was owned by Runna. They’d started years earlier and had 35% of interesting microinfluencers on their payroll. 7️⃣ Don’t build proactive products on WhatsApp AI on WhatsApp works great reactively (answering questions, receiving info). But if your product is very proactive... you’ll drown in templates, costs, and translations. We had 200+ templates and paid 2–3€/user/month. A nightmare for internationalization. - Bonus: Short-form marketing really works. Design your product with a marketing angle in mind (a visual hook, etc.). It’s been 4,5 relentless years, mostly just Linos (CTO) and me, doing everything. The result is hard to swallow: years of effort, savings, and belief that ended up in nothing. Even harder knowing friends and such great investors trusted us with their time and money. But we’re proud of what we built and learned. And if this helps anyone, it was worth it. For now, Linos and I are exploring new opportunities. Each of us wants to dive into new problems, new industries, and contribute our skills to other projects. 👇 If you’re building in AI or want to chat, explore opportunities or collaborate (AI product / GTM / growth), or just exchange notes, happy to chat.