Investors don't make your product better: lessons from building AssemblyAI

Executive overview

Raising money feels like a milestone, but it changes nothing about your product. Customers won't appear, bugs won't fix themselves, and no investor hands you traction.

Dylan Fox built AssemblyAI into the leading developer speech-AI platform by staying obsessively close to customers and ignoring startup dogma. The core discipline: validate through customer happiness, not external opinion.

Investors don't make your product better — you still have to make everything happen.

From credit-card debt to Y Combinator

• Taught himself machine learning and NLP while going $30k into credit card debt post-college.
• Identified the gap: the only SDK for voice AI required a $10k agreement and a CD-ROM — a completely broken developer experience.
• Submitted a YC application 30 days late with no product, no traction — got in because a YC partner had worked on Siri and recognised the opportunity.
• YC was one of the most stressful periods of his life; the lesson was that getting in doesn't make things happen.

Capital doesn't replace execution

• No investor will hand you customers, fix your product, or create traction.
• Validation comes from one place: are your customers happy with what you're building?
• Ignore external opinion about your company; focus on what customers actually experience.
• Startups are not franchise businesses — every journey is different, and startup dogma can slow you down.

How to stay close to customers

• Spend the majority of your day talking to customers and using your own product.
• Deep subject-matter expertise in your market beats functional expertise — it's undervalued.
• Ask the uncomfortable questions: "What are the top three things you don't like about our product?" or "If you ran our roadmap, what would you prioritise?"
• Skip the flattery; seek the problems.

Competing through focus, not breadth

• AssemblyAI deliberately does not build general-purpose speech AI.
• Specific use cases — voice agents, note-takers, sales intelligence apps — drive every model architecture and training data decision.
• Knowing exactly who you're building for lets you make tradeoffs intelligently and reach product-market fit faster.
• If AssemblyAI's model isn't right for a customer, it's usually because that use case is outside their current focus — and that's intentional.

Speed and iteration

• Optimise for speed as a startup: put up a landing page, run ads, add a "contact us" button, and see who reaches out before building anything.
• A fast iteration loop with customers is more valuable than a perfect plan.
• Usage growing 250% year over year and handling five petabytes of speech data per month — scale follows tight customer focus, not the other way round.