Lost Years, Lego Bricks &
Sugar-Coated Broccoli

Ivan Zhao

Co-founder & CEO, Notion

MAR 6 2025

Core Concept

Sugar-Coated
Broccoli

"People don't want to eat the broccoli but people like sugar, give them the sugar then hide the broccoli inside of it."

Notion's real vision: everyone builds their own software (no-code power for all)
The sugar: familiar productivity tool people already understand and need daily
Took 2+ years to realise the world needed sugar first, broccoli second
The deeper you use Notion, the more you discover the hidden broccoli inside

Framework

Notion's Lego Model: Bricks → Boxes → Solutions

lost years before PMF

2×

full code rewrites

$10B+

valuation reached

Bricks: flexible primitives, database blocks, page nesting, relations, the underlying no-code power most users never see
Boxes: pre-assembled Lego sets for a specific use case, wiki, project management, CRM, what most users actually buy
Solutions: named enterprise offerings that sit on a P&L, what enterprise buyers require before they sign
B2C2B flywheel: personal users adopt bricks → discover power → bring Notion to work → enterprise buys boxes
AI unlocks the original dream: AI agents can now assemble Lego bricks into any custom software for any user, fulfilling Ivan's 2013 vision without requiring users to code

Ivan's hard-won lesson Hardcore Lego fans love bricks. Most people want the Lego box. Ivan stayed in brick-first mode too long. The strategic shift: keep building bricks, but sell boxes, and now, let AI assemble the bricks for everyone.

Playbook

Surviving the Lost Years & Staying Lean

How to stick with a hard idea

Reset freely: Notion scrapped its entire code base twice. Better abstractions compound faster than grinding on a bad foundation, you can catch up all the progress you lost, quickly
Beware false momentum: momentum will carry you in the wrong direction the same way it carries an LLM that can't back-track from a wrong token, chain-of-thought models are better at resetting; so should you
Two energy sources: building for business (score-card, competition) vs. building for value (what you want to exist). Both are real. The durable one is value, it kept Ivan going through four lost years
Change the scenery: Ivan and Simon laid off everyone, subleased the SF apartment and office, and coded 18 hrs/day in Japan. The physical reset unlocked the product reset
Daily personal reset: "I just go to sleep, next day I'm reset." High is never too high, low never too low. Equanimity is a founder super-power

Small bus, talent density

The small bus metaphor

A smaller bus turns corners faster, accelerates harder, and is easier to steer. You choose who sits next to you, that determines speed more than headcount. Notion tracks revenue-per-employee, not total headcount.

No salespeople until $10M+ ARR

Ivan, Simon, and Akshay could each design, code, market, and close deals. When founders cover multiple functions, the headcount stays low by default, and the internal communication overhead stays manageable.

Craft is the company philosophy

Conference rooms named after timeless tools, rice cookers, Lamy pens, original Macs. The goal: build software that lasts decades, not 18-month product cycles. "Craft = applying your values to your technical know-how to make clever trade-offs."

Tactics

Building Horizontal & the $10B Near-Death

Anchor to a billion-user wedge: Notes and docs have 1B+ daily users, that gave Notion its top-of-funnel. Calendar is next (another 1B+ category); email is in progress
Build to your values or face organ rejection: When Notion hard-coded "sprints" instead of composing them from Lego blocks, engineers, customers, and the code base all pushed back. Took 9–12 months to unwind
Don't go too late on infra: Notion ran on one Postgres instance until COVID. Usage exploded, weeks away from total outage. Sharding saved the company. "Don't do premature optimisation, but plan ahead a little"
AI rewards horizontal data: The more information lives in one place, the better AI can reason across it. Bundled tools + AI = compounding advantage over fragmented SaaS stacks
Steal from outside tech: Romance of Three Kingdoms, complex-systems biology, 1970s Engelbart papers, patterns from other domains surface trade-offs competitors miss entirely

The COVID doomsday clock One Postgres instance. Weeks of runway before total data loss. All engineers halted features, shipped sharding under pressure. The lesson: avoid premature optimisation, but don't let urgency become crisis.

Contrarian

Myths About Building Enduring Products

✗ Build exactly what users ask for INSTEAD → ✓ Balance personal taste with market signal. Too much of yourself = research project; too much market = commodity. The sweet spot is authentic craft that also happens to be useful for others.

✗ Product-market fit hits you like a thunderbolt INSTEAD → ✓ For Notion it was a gradual brightening, the PMF signal was VCs sending dog treats to a private office address. "Oh, good, people care now." No binary pop, no milestone celebration. Just a slow, steady ramp.

✗ More engineers = faster product INSTEAD → ✓ Internal communication overhead grows super-linearly with headcount. A small, talent-dense bus outpaces a large, misaligned one. Notion was profitable and growing with fewer engineers than most competitors had in PM roles alone.

✗ Horizontal platforms always lose to point solutions INSTEAD → ✓ The bundling/unbundling cycle is eternal, Romance of Three Kingdoms: "long united, must divide; long divided, must unite." We're at the tail end of SaaS fragmentation. AI rewards horizontal data unification. Notion is #2 in project management and #3 in CRM without intentionally building either.

Sugar-CoatedBroccoli

Notion's Lego Model: Bricks → Boxes → Solutions

Surviving the Lost Years & Staying Lean

Building Horizontal & the $10B Near-Death

Myths About Building Enduring Products

Sugar-Coated
Broccoli