Key Takeaways

  • AI effectively makes the initial stages of any product or feature development free, with Notion's Max Schoening stating, “The first 10% of every project are now free.” This drastically lowers the barrier for entrepreneurs.
  • Designers and Product Managers need to shift from traditional wireframing in tools like Figma to "thinking in code" when building AI products, prototyping directly in LLM-friendly environments.
  • At Notion, this means using a "playground concept" for LLM prototyping, which enables faster iteration and blurs conventional role boundaries between design and engineering.
  • The point isn't for designers to ship production code, but to design within the actual medium of the product. This forces a deeper understanding of AI's capabilities and limitations before engineering takes over.
  • This approach moves beyond simply "demos not memos" by providing functional product versions for feedback, leading to more concrete and actionable user reactions.

AI Makes the First 10% of Any Project Free

Max Schoening, Head of Product at Notion, has a blunt take on AI's impact for ambitious builders: "The first 10% of every project are now free." This isn't just a marketing slogan; it's a fundamental shift in how early-stage product development works. Schoening points out that it now “takes almost no effort to now build the first version of a startup.” For founders in their 20s and 30s, this means the historical barriers to entry — the time, money, and engineering muscle required to simply get started — have dropped through the floor.

Think about the implications. What used to take weeks of backend setup, front-end scaffolding, and database integration can now be spun up in days, or even hours, with AI's generative capabilities. This frees up precious resources, not just capital, but the mental bandwidth of early teams. It means founders can test more ideas, fail faster, and pivot with far less sunk cost. The speed at which an idea can move from concept to rudimentary working prototype is the real game-changer here.

Why Designers and PMs Must 'Think in Code'

The biggest mental shift AI demands from product teams, according to Schoening, is for designers and product managers to start "thinking in code." He argues it's not about them pushing code to production, but about "them thinking and designing in the medium that will actually end up being the real thing once engineering takes it over." At Notion, they embrace a "playground concept" for LLM-friendly prototyping. This allows teams to manipulate and experiment with AI prompts and outputs directly, rather than designing abstract mockups.

This hands-on coding isn't a throwback; it's a necessity. Schoening states that “the only way that you can actually get to understanding agent loops is if you build them in the material that they're made of, which is currently code.” Trying to design complex AI interactions solely in Figma is like trying to design a car by only drawing pictures of it. You miss the physics, the mechanics, the actual user experience of how it works. Building in code, even for a prototype, forces a deeper, more realistic understanding of the AI's behavior and limitations. It's about designing with the material, not just for it.

Beyond Demos, to Working Product

The shift to "thinking in code" also fundamentally changes how teams gather feedback. Schoening recalls the mantra from his days at GitHub: "demos not memos." The idea was always to show something tangible rather than just writing about it. With AI, that idea takes on new power. "Now it's much easier to give people something to react to as in yeah here's the version of the product," Schoening explains.

This means you're no longer presenting static screenshots or click-through mockups; you're giving users a functional piece of the product, however rudimentary. The feedback you get is inherently more valuable because it's based on actual interaction with the AI, not just theoretical understanding. Users react to how the AI responds, how it integrates, and where it fails, providing insights that no amount of design review or written specification could ever capture. It's about getting real reactions to real functionality, much earlier in the cycle.

What to Do With This

Stop creating pixel-perfect mockups for your next AI-driven feature. This week, pick one new AI product or feature idea your team is considering and commit to building its absolute simplest, coded prototype. Use a basic LLM API (like OpenAI's Playground or Google's Vertex AI), build a prompt chain, and get a rough UI working. Then, put that functional, if unpolished, prototype in front of real users and gather feedback.