The Funeral That Never Happened: How Ivan Zhao Fled to Kyoto, Rebuilt Notion From Zero, and Built the App That Ate Microsoft Office

The Office That Got Sold

In late 2015, a moving truck pulled up outside a small office in San Francisco and hauled away the desks, chairs, and monitors of a startup called Notion Labs. The company wasn't being acquired. It wasn't scaling up. It was being gutted by its own founders.

Ivan Zhao, Notion's co-founder, had run out of road. The product — then called Notion V1 — was a strange, ambitious tool that let users snap together documents, kanban boards, and calendars like Lego blocks. Critics were polite. Users were confused. The app crashed constantly. Money was nearly gone, the team had shrunk to a skeleton crew, and the obvious move — the move every advisor in Silicon Valley would have told him to make — was to shut it down, return what capital was left, and get a job somewhere stable.

Instead, Zhao and his co-founder Simon Last did something that looked like surrender but was actually the opposite. They laid off almost everyone. They sold the furniture. And they left the country.

Exile in Kyoto

Zhao and Last moved to Kyoto, Japan — a city of temples, quiet mornings, and almost none of the ambient noise of the Bay Area. No Sand Hill Road. No Slack pings from investors asking about growth metrics. No pressure to perform a startup for an audience.

They rented a small apartment. They lived cheaply. And for the better part of a year, it was just the two of them, a whiteboard, and the humbling realization that the entire foundation of their product was wrong.

Notion V1 had tried to be everything by building everything separately. Documents were one system. Kanban boards were another. Calendars, databases, wikis — each was its own hard-coded primitive, bolted onto the others with glue code. Every new feature meant building an entirely new subsystem, and every subsystem interacted badly with the ones before it. The app wasn't broken because the team lacked talent. It was broken because the architecture itself couldn't scale with the ambition.

In Kyoto, away from the noise, Zhao kept returning to an idea he'd been circling for years, heavily influenced by Bret Victor's writing on "tools for thought" — the belief that software should be a malleable medium for human cognition, not a rigid product with fixed features. If Notion was ever going to be a true thinking tool, it couldn't be a collection of separate apps stitched together. It had to be built from one single, infinitely composable primitive.

The Turning Point: Everything Is a Block

The rewrite that emerged from that year became Notion V2, and its core idea was almost absurdly simple: everything is a block.

A paragraph is a block. A heading is a block. A to-do item, an image, an embedded tweet, a table, a database, a page itself — all blocks. And critically, any block can contain other blocks. A toggle can hold a kanban board. A kanban board can hold pages. A page can hold a database that itself behaves like a spreadsheet, a calendar, or a gallery depending on how you choose to view it.

Technically, this meant collapsing dozens of bespoke data models into one recursive tree structure. Under the hood, every block is stored as a row in Postgres, carrying a parent pointer, a type field, and a flexible content payload. Rendering the page is just a tree walk: start at the root block, recursively render its children, and let the type field decide whether a given node renders as text, an image, or a live database view. Moving a block is as simple as reassigning a parent ID. Nesting is infinite by default, not by exception.

That architecture solved the collaboration problem too. Because every block is an addressable node with a stable ID, Notion could layer CRDTs (conflict-free replicated data types) on top to handle concurrent multi-user editing — letting two people type in the same document, even the same block, without stepping on each other's keystrokes or needing a central lock. And because the block model was uniform, Notion could eventually expose it directly through a public API: developers didn't get a bespoke "documents" endpoint and a separate "databases" endpoint — they got blocks, the same atomic unit the whole product was built on.

The genius of the design wasn't just technical elegance. It was that it made Notion infinitely flexible without Notion's own team having to predict every use case. Users didn't wait for Notion to ship a CRM feature — they built CRMs out of nested databases and linked blocks. They didn't wait for a habit tracker — they built one out of toggles and checkboxes. The product became a platform disguised as a notes app, and the roadmap was effectively crowdsourced by its most obsessive users.

The Cult That Built Itself

When Notion relaunched in 2016 with this new architecture, there was no sales team, no ad budget, no enterprise outreach plan. Growth came from somewhere almost nobody tracks in a pitch deck: template culture.

Power users — designers, productivity obsessives, indie hackers — started building elaborate personal dashboards, content calendars, and wikis inside Notion, then sharing screenshots on Twitter. Each screenshot was a tiny advertisement for the product's flexibility, and each shared template was a tutorial, a sales pitch, and a community artifact all at once. Notion didn't need a marketing department because its users were voluntarily running one, driven by genuine enthusiasm for a tool that felt less like software and more like a canvas.

Zhao's design philosophy reinforced this. He was notoriously resistant to feature creep, rejecting the typical SaaS playbook of bolting on new modules to chase every competitor's checklist. The goal wasn't to have the most features. It was to have the fewest primitives that could compose into the most outcomes — a philosophy lifted almost directly from Victor's "tools for thought" essays, which argued that software should amplify human thinking rather than constrain it into pre-built templates.

The Empire Strikes Back

By 2021, Notion was valued at $2 billion. By late 2021, that number had jumped to $10 billion, fueled entirely by organic growth and word of mouth. In 2023, Notion shipped Notion AI, weaving generative writing and summarization directly into the block editor — not as a separate chatbot, but as another tool living inside the same recursive canvas. More recently, the company has pushed further into "databases as apps," letting users build lightweight internal tools and automations without ever leaving a page.

The company that once sold its furniture and skipped town now claims more than 100 million users, including a staggering share of Fortune 500 teams who've quietly replaced their wikis, their project trackers, and chunks of Microsoft Office itself with Notion workspaces.

Microsoft noticed. In 2021, Microsoft launched Loop, a block-based collaborative workspace that borrowed Notion's core conceit almost component-for-component — portable, nestable blocks that live across documents, chat, and apps. It was, in the eyes of nearly every tech journalist who covered the launch, an admission that the giant of productivity software was chasing a two-person startup that had nearly folded six years earlier in a Kyoto apartment.

The Legacy

The lesson of Notion isn't just "pivot when things go wrong." Plenty of startups pivot and still die. The lesson is that Ivan Zhao and Simon Last didn't pivot the business model — they pivoted the abstraction. They went back to first principles, found the single atomic unit their entire vision depended on, and had the discipline to burn down a mediocre product chasing a great one.

Sometimes the only way to find what you're actually building is to almost lose the chance to build it at all.