All podcasts / Lenny's Podcast / Summary

AI era skills: Why cultivating agency matters more than job titles | Max Schoening (Notion)

2026-05-02 - 87 min - source - Read full transcript
Lenny Rachitsky (host)Max Schoening

Key insights

The old excuse of 'I don't have the skill' no longer holds, so what separates people now is agency, not ability.
Max argues that AI has removed the skills bottleneck for most knowledge work, which means the deciding factor becomes whether someone treats the world as changeable and acts on it. He ties this explicitly to founder mindset: people with true agency who understand their environment is malleable will do well, while people who stay attached to a fixed job description will struggle.
ai-era-agency
Notion's designers-write-code culture began as a narrow prototyping hack, not a company-wide mandate.
When Max joined Notion, two designers built a deliberately minimal, LLM-friendly 'playground' codebase just for chat-interface prototyping, inspired by Bret Victor's 'Stop Drawing Dead Fish.' As model capabilities improved, the same designers started contributing to Notion's real production codebase, but the original goal was never shipping code, it was forcing designers to think in the actual medium.
product-building-fundamentals
Malleable software means software should adapt to users, not lock them into decisions made by the company that built it.
Max frames most apps today as tightly glued stacks (UI, data ownership, behavior) that users can't rearrange, unlike a house where you can move furniture. He wants software, and Notion specifically, to move back toward general, adaptable tools that individuals and teams can reshape, without forcing everyone to run their own Linux distribution to get that control.
malleable-software
AI has made the first 10% of any project free, but the last 10% is still 90% of the real work.
Cheap early exploration means teams can now demo a rough version instead of writing a PRD, and can send multiple agents to explore different directions in parallel. But Max is clear that finishing, polishing, and making something reliable for real users has not gotten proportionally easier, so the hardest work has shifted rather than disappeared.
ai-era-agency
Every great product has one exceptionally good tiny core, and adding more features to a mediocre core never fixes it.
Max cites GitHub's pull request, Heroku's one-line 'git push heroku master' deploy, Dropbox's menu-bar sync icon, and Notion's blocks/slash-commands as examples. He warns against the common trap of believing 'one more feature' will finally make a product great, calling that loop one of the biggest pitfalls in product work, and shares that his own 2014 Notion-adjacent startup failed by polishing document editing instead of finding its core.
product-building-fundamentals
Taste is a trainable skill, not an innate gift, built the same way a model is trained: through reps and feedback.
Max describes taste as running a mental 'virtual machine' that predicts whether a specific in-group will like an idea. He says you don't need universal appeal, you need to correctly model your chosen audience, and that this capability compounds with iteration count in a specific domain before it transfers to other domains.
taste-and-quality
Deliberately building for a small, high-status in-group can be a winning strategy rather than a limitation.
Max argues the world is effectively run by 'group chats of eight people or fewer,' and that trying to serve all 8 billion people (or all of Notion's user base at once) dilutes the product for the users who actually push the frontier. He cites TBPN's small, influential listener base as proof that exclusivity paired with genuine value for that audience can still produce a massive outcome.
taste-and-quality
Notion currently treats AI token spend as unlimited on purpose, but expects a hard ROI reckoning within six to twelve months.
Max says he deliberately doesn't track individual token spend right now because it's the wrong thing to optimize for during an exploration phase, and that boasting about token counts is as hollow as boasting about lines of code written. He expects that within roughly six to twelve months, companies broadly will start asking uncomfortable ROI questions about AI spend.
token-economics
AI's biggest near-term impact is accelerating software eating the world, not directly transforming non-coding domains.
Max is skeptical that models have meaningfully improved at non-coding tasks like writing, calling AI writing still noticeably 'sloppy.' Instead, he argues that as the cost of writing software drops toward zero, more of every other domain (HR, marketing, ops) gets encoded as software, meaning the transformation happens by domains absorbing coding capability rather than by AI directly getting better at those domains.
ai-era-agency
A useful litmus test for how 'agent-filled' an engineering team is: any manual code intervention should feel like a bug.
Citing Simon Willison, Max says that outside of code review (which he still wants humans doing more of, not less), any time a human has to manually intervene in written code, it signals something went wrong in the verification loop of the software factory, not that intervention itself is inherently useful.
ai-era-agency

Books referenced

Media referenced

Companies

Techniques and frameworks

Summary

Max Schoening, Head of Product at Notion and a two-time founder with stints leading design and product at GitHub and Heroku, joins Lenny to argue that the scarce resource in the AI era is not skill but agency. His central claim: AI has quietly removed most people's excuse of "I can't do this because I lack the skill," which means the real differentiator becomes whether someone treats their environment as something they can reshape. He roots this in a version of the Steve Jobs line about realizing the world was built by people no smarter than you, and in his own practice of "driving Notion like it's stolen" even though he wasn't a founder there.

Much of the conversation traces how this plays out concretely at Notion. Max describes the origin of the company's now-famous "designers ship code" culture as a narrow prototyping hack: two designers built a deliberately minimal, LLM-friendly playground codebase for chat-interface prototyping (inspired by Bret Victor's "Stop Drawing Dead Fish"), not a company-wide mandate. He is explicit that he doesn't care whether designer-written code ever reaches production; the value is in forcing people to think in the real medium software is made of. This connects to his broader thesis of "malleable software," the idea that software should bend to the interests of the people using it rather than the interests of the company that built it, much like a house that a family gradually reshapes to fit their life rather than one dictated entirely by an architect.

On what actually makes products succeed, Max returns repeatedly to the idea that every great product has one small, exceptionally good core, GitHub's pull request, Heroku's single-line deploy command, Dropbox's reliable menu-bar sync icon, Notion's blocks and slash commands, and that piling on features to a mediocre core never works. He shares his own cautionary tale: a 2014 Notion-adjacent startup that lost out to Notion itself partly because his team polished document editing (markdown folding, block selection) while Notion's much rougher early editor won anyway, because their core concept was right and his wasn't. Taste, in his framing, is not innate; it's the ability to run a mental "virtual machine" that predicts how a specific in-group will react to an idea, and it's built through the same iterate-with-feedback loop that trains a model, which is why he's skeptical of the "taste is the last human advantage" narrative.

The episode also covers token economics and where AI's impact concentrates. Max says Notion currently treats token spend as effectively unlimited by design, not tracking it closely because doing so during an exploration phase would be optimizing for the wrong thing, while predicting that most companies will start asking hard ROI questions about AI spend within six to twelve months. He pushes back on the idea that AI is transforming domains like marketing or writing directly; instead he thinks non-coding progress is overstated and that the real mechanism is software eating the world faster, as the falling cost of writing code lets every other domain (HR, ops, sales) absorb more of its own software-writing capability.

The conversation closes with rapid-fire segments covering contrarian and failure "corners," and a lightning round. Notable threads include Max's argument that deliberate exclusivity (building for a small, high-leverage group rather than trying to please everyone) can be a winning strategy, his half-joking claim that knowledge work already functions as a form of universal basic income, and book recommendations spanning "Code" by Charles Petzold, "Tools for Conviviality" by Ivan Illich, and "Seeing Like a State," all centered on the same throughline: understanding the real material and real system you're working with, rather than the legible abstraction of it.

Notable Quotes

"I don't think agency is very evenly distributed in the world. And I think people who have true agency and they understand that the world around them is malleable will do great." - Max Schoening

"All the great products have something tiny that is a superpower. One tiny core that is so exceptionally good... One of the biggest pitfalls is if you get into the loop of, if I just add one more thing to the product, it'll be finally great. That never works." - Max Schoening

"Malleable software is the idea that software works closer to the interest of the people that use it than the interest of the corporation that makes it." - Max Schoening

"I think there's a real danger in sort of making the token spend the metric to boast about, which is the same as when people boast about how many lines of code they've written in a day." - Max Schoening

"You just have to do reps. It's almost like training a model... it seems very much how we train models too." - Max Schoening, on how taste is built