How Notion Nearly Died Before Finding Its First 1,000 Users
Notion's real story isn't a 2016 explosion. It's two founders in Kyoto, rebuilding from scratch. Here's what actually happened and what it means for your build.
Most people tell the Notion story like this: two founders flew to Kyoto, rebuilt the product, launched in 2016, and the rest is history. That version is clean and easy to share. It's also wrong. Ivan Zhao has said directly that real traction didn't arrive until the Notion 2.0 Product Hunt launch in 2018 — two years after the story supposedly ends. What happened in between is messier, more honest, and far more useful if you're currently staring down a product that isn't growing the way you hoped. Here's what actually went down, what decisions made the difference, and what founders building today can take from it.
Why did Notion almost fail in the first place?
The failure wasn't a lack of money or ambition — it was a product pitch that was too abstract for the market that existed at the time. Notion launched in 2013 with roughly $2M raised from angels including Naval Ravikant and a check from First Round Capital. Ivan Zhao's original vision, borrowed from Douglas Engelbart's 1962 paper "Augmenting Human Intellect," was software as LEGO blocks — composable primitives that let anyone build their own tools. That vision turned out to be correct. It was just about four years ahead of when people were ready to care.
The first technical build made things worse. The team built on Web Components with a CouchDB backend, went offline-first, and ended up with a product that crashed constantly and lost user content. Ivan told Sequoia directly: "Bugs came from everywhere. It was unstable. People were losing their content." By mid-2015 the runway was nearly gone, the team had started quitting, and Ivan made the call to let everyone go. His words from the Figma blog: "If you looked at the burn rate, we all would've died together. It wasn't much of a choice."
The first lesson here is plain: having a correct vision and having the right form factor are two different problems. Notion's eventual product was essentially what Ivan described in 2013. But the 2014 version asked users to imagine a future instead of solving a problem they already had. Users don't buy philosophy. They buy relief.
What actually happened in Kyoto?
Ivan and co-founder Simon Last subleased their San Francisco apartments and office, picked Kyoto off Airbnb because the houses were bigger than Tokyo's, and moved there without speaking a word of Japanese. They rented a two-story house, slept on the bottom floor with bedrooms separated by a paper Shoji screen, and coded. Ivan's most-cited quote from that period: "Neither of us spoke Japanese and nobody there spoke English, so all we did was code in our underwear all day."
That's not just a good story — it describes a specific working condition that matters. No distractions, no investors to update, no team to manage. Just two people with enough runway to rebuild the thing properly.
And they did rebuild it properly. They moved off Web Components onto React. They dropped the offline-first architecture. Most importantly, they landed on the block model that defines Notion today: every element — a paragraph, an image, a database — is the same kind of object, and any block can contain any other block. That single architectural decision is what makes Notion feel like Notion. It didn't come from a sprint or a hackathon. It came from months of quiet, focused rebuilding with nothing left to lose.
Notion 1.0 launched on Product Hunt on August 9th, 2016. It received 5,537 upvotes, 170 comments, and won the Golden Kitty for the year. A real milestone. But Ivan said in Notion's own 2024 blog post celebrating 100 million users: "We started in 2013, but it wasn't until 2018 with the Notion 2.0 Product Hunt launch that we saw signs of traction." 2016 was oxygen, not ignition.
What finally got Notion to 1,000 users?
The block model gave Notion structural flexibility no competitor had. But the thing that actually moved users was the same thing that moves users for most B2B tools: specific, shareable templates solving specific, immediate problems. By 2018 Notion had a richer template library and a community that was building and distributing their own setups. The product gave people something concrete to show each other. That shareability — one person's Notion workspace forwarded to a colleague — is what turned slow growth into compounding growth.
There's also a timing element. By 2018 the no-code conversation had matured. People were ready to think about software as something they could configure, not just consume. Ivan's 2013 vision wasn't wrong — the market just needed four more years to catch up. Product-market fit is partly about product, partly about moment.
How Entellya actually builds this
The Notion story is a lesson in how long it takes to get architecture right when you're doing it mostly by hand. Two founders, eight-plus months, a complete rewrite. That's what it cost to land on the block model.
At Entellya, we use AI to BUILD the products — not just embed AI features inside them. Tools like Cursor, Claude, Bolt, and v0 sit at the center of how we design, generate, and iterate on code. Flutter and Next.js handle the mobile and web stacks. Supabase handles the data layer. The result is that architectural decisions that once took months of manual rewriting can be explored, tested, and validated in days.
That matters most at the stage Notion was stuck in — post-funding, pre-traction, needing to pivot the technical foundation without burning the remaining runway. Trendio raised 1M€ and reached 100k users. Celering runs at 20k+ daily users. Apprecio crossed 100k downloads on Google Play. None of those outcomes happened because the underlying code was written slowly and carefully by hand. They happened because the build process itself was fast enough to let the product chase the market signal in real time.
When you can rebuild something in weeks instead of months, you make different decisions. You test more. You throw out more. You don't get stuck defending a technical choice just because it took too long to make.
What this means for your timeline and budget
Notion's Kyoto rewrite took somewhere between eight months and a year for two experienced engineers working full time. Today, a comparable architectural pivot — rethinking your data model, migrating your stack, redesigning core interactions — is a fundamentally different scope with the right tooling and team.
This isn't about cutting corners. The Notion block model is elegant precisely because it was thought through carefully. Speed doesn't mean shallow. It means you spend your time on the decisions that matter — what the product should do — rather than the mechanical work of making it do that thing.
If you're at the stage Notion was in mid-2015 — the product technically works but nobody's using it, and you're trying to figure out whether to pivot or push — that's exactly the moment where build speed becomes a survival variable. Burning six months on a rewrite you're not sure will work is a different kind of bet than burning six weeks.
If you want to talk through where your product is and what a fast, AI-assisted rebuild would actually look like, book a call here.
Build something like this.
Tell us your idea. We'll come back in 24h with a plan, a timeline, and a price that'll surprise you.
Start your projectFAQ
How long does a full product rebuild actually take with an AI-assisted team?
It depends on the scope, but for a mobile or web product with a clear direction, we're typically talking weeks rather than months. A comparable Notion-style architectural overhaul — new data model, new stack, redesigned core interactions — would be a very different timeline today than it was in 2015. We use Cursor, Claude, Bolt, v0, Flutter, and Next.js together, which compresses the mechanical parts of engineering significantly.
Is it better to iterate on a broken product or rebuild from scratch?
Notion's answer was rebuild — and it worked. But they also had a clear architectural problem (the Web Components and CouchDB stack) and a specific solution (blocks on React). If your problem is technical debt with a known fix, a targeted rebuild often makes sense. If your problem is that you don't know what to build yet, rebuilding won't help. Get the direction right first, then build fast.
What's the difference between Notion's block model and a regular database?
In a standard database, different content types live in different tables with different schemas. Notion's block model treats every content primitive — text, image, checkbox, table row — as the same kind of object. Any block can contain any other block. That structural uniformity is what lets users combine pages, databases, kanban boards, and calendars inside each other without the product needing a custom integration for each combination. It's a clean architectural idea that took a Kyoto rewrite to get right.
Does early traction on Product Hunt actually mean product-market fit?
Notion's story is a clear answer: not necessarily. Their 2016 launch won Product of the Month and a Golden Kitty award. Two years later, Ivan described 2018 as when they first saw real traction signs. A strong launch gives you an initial user base and some press. Product-market fit is when users come back, tell other people, and your growth compounds without you pushing it every week.
What kind of products does Entellya build?
Landing pages, mobile apps, SaaS platforms, dashboards, and AI chatbots. The common thread is that we use AI throughout the build process itself — in the design phase, in code generation, in iteration — not just as a feature inside the finished product. Clients include Wio Bank, Dorydoc, Kanara, and others across fintech, health, and consumer apps.