AI Makes the First 10% of Every Project Free: Notion's Max Schoening on Agency and Malleable Software

May 3, 2026 · Lenny's Podcast
🎧 PodShort 86 min squeezed to 2 AI SprinklerAS Sales Tech New
Episode artwork
Max Schoening
Head of Product at Notion
Lenny's Podcast
86 min squeezed to 2
Full episode from Lenny's Podcast
Quotable Moments

I think 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.

Key Insights
  • In the AI era, even with skills readily available, the ability to drive change and act (agency) is crucial for success.
  • The best way to cultivate agency and realize you can change things is by actively 'making things' rather than just contemplating them.
  • Malleable software empowers users by allowing them to adapt and rearrange their digital environment, aligning with their interests over corporate dictates.
  • AI makes the first 10% of any project (e.g., prototyping, initial versions) effectively 'free' or effortless, shifting the focus to iterative refinement.
  • For designers and PMs, understanding and 'thinking in the medium' (code) is vital, even if they don't ship production code, as it deepens their understanding of how systems like agent loops function.
  • While roles are merging in product development, it's critical to avoid losing specialists who excel in areas like engineering optimization and design craft.
  • SaaS will not disappear but will evolve, with general-purpose tools becoming more prominent, while the 'as a service' model will persist for maintenance and specialized solutions.
  • Despite AI's ability to generate vast amounts of software, the overall quality hasn't significantly improved, highlighting a lack of focus on robust engineering over mere feature quantity.
Metrics Mentioned
  • First 10% of every project are now free (AI reduces the effort needed for initial project phases.)
  • First 90% of a project are now done with AI (AI streamlines the majority of project completion, leaving 10% as the hardest part.)
  • 10% of GitHub's top contributors were designers (pre-LLMs) (Illustrates designers' historical involvement in coding, even before AI.)
  • $3,000 a day (The potential cost of using very fast AI models for inference.)
  • $1,000 off Vanta (Promotional discount for podcast listeners.)
  • 35 security and privacy frameworks (Vanta automates compliance across numerous security and privacy standards.)
  • 50% of engineers in the world writing 100% AI code (A prediction for within a year, highlighting the rapid shift towards AI-generated code.)

RevBots.ai View:

  • AI Sprinkler stage teams can leverage AI for rapid prototyping and initial project phases.
  • Malleable software aligns with ARM's focus on user empowerment and adaptability.
  • Specialized roles remain vital despite AI's role in merging functions.
  • SaaS evolution hints at ARM's orchestration replacing legacy tools.
🎧Full Episode:Lenny's Podcast →