Taste is the last moat
When the cost of building is zero and competition is everywhere, taste is everything
I was talking to a startup founder yesterday and he said AI has made startup development the hardest it’s ever been.
Because now the cost of developing a new feature is zero for startups AND the complexity of developing one even for massive codebases is far less onerous AND the difficulty for DIYers who are fine with building single player utilities has never been lower. So you’re now competing with other startups, the incumbents, and DIYers all at the same time at a breakneck pace.
So how do you survive when competition comes from everywhere all at once?
Taste.
Taste isn’t just how your software looks. It’s every single part of your unique approach to software development from the design system to the onboarding to the pricing to the functional mandate to the back-end to logo to UX to operating system deployment. Taste is form, and increasingly, function will follow form, not the other way around.
Let me make this tangible with an example - AI assistants. There seems to be one popping up every day - Viktor, Ollie, Town, Valet, Zo, Spence. In theory it’s simple to build (with AI). Give a model access to tool use, provide context on your writing style and give it memory so it can self-improve. Everyone has the same models and same infra, so we’re all left with the same product right? Wrong.
Instead, the bar is raised. The ability to simply answer questions correctly isn’t enough. Look further and you’ll see that taste is what differentiates the copycats from the true builders, the burnouts from the billion dollar businesses:
Harness: How much orchestration do you build into your harness? Do you allow tool use organically or ask for users to connect proactively? To what extent does memory help with self-improvement?
Functional philosophy: Is your app built to automate processes? Is it designed to improve productivity? Is it built to organize your life? Is it built to be whatever the user wants it to be?
Privacy: Do you orient towards giving users more access at the cost of privacy or towards more privacy at the cost of less context?
Proactivity: Does the AI figure out what’s important and start doing or do you direct it towards problems? Does it suggest ideas or wait for yours?
Pricing: Will you charge by input or output? Subscription or credits? Do you offer a free trial to see value or ask for money upfront to prove people are willing to pay?
Persona: Are you utilitarian or warm/cozy? Is your assistant a bot or a being? How much personality are you embedding into your app?
Structure: What does an interaction look like? Is it a session? Is it a thread? Is it a task? Is it a process automated?
Embedding: Do you design for desktop and mobile or choose a platform? Will this work cross-platform or is it Mac only? Are you embedded in the menubar, do you use a universal shortcut, or can you only be accessed in app?
Connectors: Do you rely on a catalog of one-click connectors that the user adds? Do you have the AI connect as it operates? Are you using external services or building direct connections?
AI can’t answer these questions for you because there isn’t a clear right answer yet. Your experience, philosophy and focus is what helps you make the right decisions. The founders that stand out are the ones who think about all the right questions and have strong views, weakly held about which decisions they’re making and why.
Contrary to the premise, it has never been a better time to build. Development costs should never have been the hard part. Coming up with great ideas and thinking through every element of making them real should have been. And finally, we’re here.
Let’s build something amazing with great taste as the harness.