Most important take away

In an era where AI puts skills at everyone’s fingertips, the differentiator becomes agency — the willingness to act as if the world is malleable and you can change it. Max’s mantra: “drive Notion like it’s stolen.” Skills are commoditized; the people who realize “the world is made up by people no smarter than you” and just start making things will pull ahead. The first 10% of every project is now free, so the leverage shifts to taste, craft, and the relentless reps that build them.

Summary

Key themes:

1. Agency > skills. AI removes “I don’t know how” as an excuse. The new differentiator is whether you act on the world or wait to be told what to do. Founders had this; now everyone needs it. “Drive Notion like it’s stolen” — contribute as if you owned the place.

2. Make things to develop agency. Tinkering is the path. The Steve Jobs insight (“the world is made up by people no smarter than you”) only lands when you’ve built something. Home cooking counts. Tool-making is innately human.

3. Designers and PMs should code — to master the medium, not to ship to production. The Brett Victor “stop drawing dead fish” rule: a static Figma chat mock can’t capture an AI experience. Notion built a small LLM-friendly playground for designers to prototype. Max would rather hire a designer who deeply understands agent loops than one who can tweak production CSS but doesn’t get how agents work.

4. The first 10% of every project is now free. No more PRDs — just ship the janky demo. “Demos not memos.” This means it’s cheaper to fan out 10 agents on 10 paths and pick what works. Iteration is built in earlier; waterfall is dead.

5. The last 10% is still 90% of the work. Vibe-coding is producing more software, not better software. The “engineering” of making something work for 100M users hasn’t gotten easier — quality is the missing virtue. Max wants Apple’s “machined unibody aluminum” engineering ethic to come back.

6. Software 1.0 is eating new domains. Models are getting exponentially better at coding but not significantly better at writing or other generalist tasks. The implication: “software eating the world” accelerates — every domain (HR, marketing, ops) gets self-coding superpowers, not “AI-native non-coding tools.”

7. Malleable software is the future. Software should serve the user’s interests, not the corporation’s. Today apps are monolithic — UI, data, and behavior are glued together. Max wants ownership over your computing life via communal, real-time-collaborative malleable platforms. Notion is positioning itself as that operating system.

8. The SaaS apocalypse is exaggerated. People don’t actually want to maintain software — that’s the value of “as a service.” Tools become more general (back to ’90s primitives like word processors, file maker pro) but stay maintained by specialists. Anthropic uses Slack, not a custom rebuild. “Why waste time?”

9. Token spend is the new lines-of-code. Don’t optimize for it as a leaderboard metric. But don’t ignore it either — in 6-12 months companies will have uncomfortable ROI conversations. Notion currently has unlimited token spend by policy.

10. Smarter ≠ always better; speed and modality matter. Once you can’t see the pixels, you don’t need a sharper retina display. Once intelligence is “good enough” for a task, latency, locality, and cost become the real edge. Max is more excited about the exoskeleton than “a god in a box.”

11. Taste is back-propagation. “Run a virtual machine in your head where, given an idea, you can predict whether your in-group will like it.” Build it through reps with feedback. Designers with taste have side projects and are constantly trying new tools. Surround yourself with tasteful objects (Notion’s conference rooms named after the first typewriter, Macintosh, Porsche 911).

12. Great products have a tiny core that’s exceptionally good. Heroku: git push heroku master. Dropbox: the menu-bar sync icon (so good it became your “do I have internet” indicator). GitHub: the pull request. Notion: blocks + slash commands + collab. Most product death-spirals come from “if I just add one more feature it’ll finally be great.” Almost never works.

13. Being first is overrated; durability matters. AirPods weren’t first. Anthropic was way behind OpenAI. Build for generations, not for what’s trending today.

14. Notion AI agent worked because Notion is an OS, not an app. Agents need context. A connected workspace gives them substrate. This is “made for such a time as this” — Notion’s bet on being a malleable OS pays off in the agent era.

15. Obviously good + incremental correctness. Notion uses literal “obviously good” stickers. iPhone, ChatGPT — nobody argued they weren’t obviously good. But don’t hide in a cave waiting for perfection: increase shots on goal, then consolidate evolutionary branches back into a “naked robotic core.”

Career advice:

• Cultivate agency by making things. Side projects, home cooking, tinkering. The skill of agency is built, not given.
• Drive your job “like it’s stolen.” If you’re a PM/designer, learn to prototype in code — to master the medium, not necessarily to ship. The real test: do you understand how agent loops work in their native material?
• Work hard between 18 and 25. Don’t fall for the “permanent underclass” frenzy. Don’t believe the “this is the last train” narrative. Work intensely early, ease up later. The amplitude of worry should be low; the work output should be high.
• Don’t optimize for short-term frenzy. Read computer science history. History repeats more than it innovates. People who cling to certainty in an uncertain era will be miserable.
• Specialize on the edges as roles merge. As designers code and engineers design, the risk is losing specialists. Stay deep in your edge — delight, craft, manufacturing-grade engineering — even as the middle homogenizes.
• For young Silicon Valley specifically: A lot of the current cohort doesn’t actually love computers — they’re chasing money. You’ll be hollow if that’s your only motivation. Get curious about how things actually work.
• Pursue exclusivity, not universality, when it fits. “The world is run by group chats of eight people.” Notion would upset its top 500M users by trying to serve 8B. Build for the top of the class in your in-group.
• The fail mode in hiring: Loosening the “designer who can code” requirement created a slippery slope. Hire fewer, more polymath people.

Product strategies:

• Demos not memos.
• Find the tiny core. Resist “one more feature” syndrome. Periodically consolidate evolutionary branches.
• Use jobs-to-be-done as a zoom-out reminder (“be the user for a second; would you buy this?”), not as religion.
• Make tools that turn users into superheroes (Cathy Sierra). Cloud Code/Codex give you bragging rights about your code; old code-review tools roast you publicly. Same product space, opposite positioning.
• Ship to increase shots on goal — but ruthlessly consolidate after.
• Make your product agent-readable (a connected workspace, clear permissions, clean APIs). That’s why Notion’s AI agent works.

Hot take on UBI: “We already have universal basic income. It’s called knowledge work.” Max means we already get paid extraordinarily well to type letters into a box in air-conditioned rooms; we’ll find new reasons to insert ourselves into the loop with agents.

Stocks/companies named: Notion (private), Anthropic, OpenAI, GitHub (Microsoft / MSFT), Heroku (Salesforce / CRM), Cursor, Vercel, Replit, Sierra, Clay, WorkOS, Vanta, Dropbox (DBX), Snapchat (SNAP), Apple (AAPL), Meta (META), Google (GOOG). No stock recommendations.

Chapter Summaries

1. Cold open: agency as the new differentiator — Skills are at your fingertips via AI; the gap is whether you act on the world. “Drive Notion like it’s stolen.”

2. Origin story: designers prototyping in code at Notion — Brett Victor’s “stop drawing dead fish.” Notion built a small LLM-friendly playground for designers to prototype AI features in code rather than Figma.

3. Designers should code — to master the medium — Not for utility of shipping, but to interrogate the material. Max would take a designer who deeply understands agent loops over one who can tweak CSS but doesn’t.

4. Specialists at the edges as roles merge — Hardware metaphor: prototyping is 3D-printed layer lines; engineering is the manufacturing. Don’t lose delight, craft, and manufacturing-grade engineering as the middle homogenizes.

5. Cultivating agency by making things — Tinker, cook, build. The Steve Jobs insight (“world made by people no smarter than you”) only lands after you’ve made something.

6. Malleable software — Software should serve users, not corporations. Apps glue UI/data/behavior together; users can’t tweak. Notion as malleable OS.

7. Pinned video: Dieter Rams criticizing chairs — “Neither orderly nor properly chaotic.” Design should be useful first, beautiful second. Test of usefulness: can it be changed?

8. The SaaS apocalypse is exaggerated — People don’t want to maintain software. Tools become more general (90s-style primitives) but stay as-a-service. Anthropic uses Slack.

9. What’s most changed about the job — The first 10% of every project is now free. Demos not memos. Iteration is built in.

10. What’s the next leap? — Plain text/markdown is durable. Faster inference may bring back direct manipulation (mold the clay rather than queue jobs). Saturation of intelligence ≠ saturation of usefulness.

11. Token spend & ROI — Notion’s policy is currently unlimited. Don’t make token spend a leaderboard metric. But ROI conversations are coming in 6-12 months.

12. Software is eating the world (still) — Models exponentially better at coding, not at other domains. So coding capability spreads into HR, ops, marketing — software 1.0 just gets cheaper.

13. All winning agents are coding agents — They build the skills they need. All harnesses look like coding agents now.

14. What if the model gap widens vs. narrows — Cloud-wars analogy: businesses want choice, not lock-in. Heroku vs. Kubernetes — Kubernetes won by making ops teams superheroes, not replacing them.

15. Cathy Sierra’s “make users superheroes” — Cloud Code/Codex gives you bragging rights for your code; old roast-style review tools fail. Frame products around making the user great, not the product.

16. Why Notion’s AI agent succeeded — Agents need context. Notion’s connected workspace + OS-like substrate is the perfect AI environment. “Made for such a time as this.”

17. Shipping fast at scale — Internal “shots on goal” mantra. Reminding people to be less precious. Feature count is the new lines-of-code (silly metric). Quality is what’s missing.

18. Obviously good + incremental correctness — Notion’s “obviously good” stickers. Don’t hide in a cave; iterate. But periodically consolidate evolutionary branches back to a tiny core.

19. Taste is back-propagation — Build via reps with feedback for a defined in-group. Designers with taste have side projects and constantly try new tools. Surround yourself with tasteful objects.

20. Tiny-core theory of great products — Heroku, GitHub, Dropbox, Snapchat, Notion all had a single exceptional core. “One more feature” rarely saves a product without one.

21. Being first is overrated — Durability over virality. AirPods, Anthropic. Build for the long game.

22. JTBD as zoom-out, not religion — Be the user for a second. Would you buy what you just built?

23. Hot take on UBI — “We already have UBI. It’s called knowledge work.” Half-joke, half-real: we live unusually well; we’ll find new reasons to be in the loop.

24. Contrarian corner — small-groups theory — Inclusivity isn’t always great. The world is run by group chats of eight. Be okay being exclusive (the TBPN model: 8K listeners, $X00M acquisition).

25. Fail corner — Loosened “designer who can code” hiring criterion → slippery slope. Started a 2014 Notion competitor; spent forever polishing the editor while Notion shipped a worse editor with a better core. Lesson: get the tiny core right before iterating.

26. Closing advice for younger people — Don’t fall for the “last train” frenzy. Work hard 18-25. Read CS history. Tune down the worry amplitude.

27. Lightning round — Books: Code (Petzold), Tools for Conviviality (Illich), Seeing Like a State. Products: Ghostty terminal, Moshi (mobile coding), Corne keyboard, CRKD pocket knife. German words: Tüftler (tinkerer, no derogatory edge) and Verbraucher (user, but emphasizing using up). Motto: “the universe is change, life is what you make it.” How to help him: go for a walk and notice that everything around you was made by people no smarter than you.