Han Wang and the value of a wrong turn

Han Wang's first paying documentation customer could not edit his own documentation. There was no editor. There was no backend. There was a website, assembled over a weekend, and two founders willing to do the work themselves. For a software business, this was a rather personal interpretation of automation.

The customer was Hyperbeam. Wang had called Declan, a friend from his Y Combinator batch, offering to improve the startup's docs. Declan agreed, with the reasonable proviso that he had no time to set anything up. Wang and his co-founder, Hahnbee Lee, migrated the content, revised it and built the site. Hyperbeam paid. After several abandoned products, a modest transaction carried more useful information than another enthusiastic launch.

That scene is a useful place to meet Wang, now Mintlify's co-founder and CEO. His career contains venture capital, software engineering and a Cornell education, but the interesting habit is smaller: he keeps putting himself within reach of somebody else's inconvenience. Sometimes the answer requires a new product. Sometimes it requires fixing somebody's sentences.

A timetable before a company

Wang’s hometown is Surrey, British Columbia. At Cornell, he studied information science in the College of Agriculture and Life Sciences. The connection between building software and helping people was already fairly literal. As a middle school student, he had noticed classmates struggling to manage their timetables and built an app for the school.

University offered more complicated problems and more people to build with. Cornell's Digital Agriculture Hackathon led to FoodFul, an agricultural technology venture. His team won the event's Most Market-Ready category. Wang had entered without much knowledge of the industry; the experience introduced him to entrepreneurship as something that could begin with close observation rather than an elaborate business thesis.

He also joined Cornell's Design & Tech Initiative. There he and Lee worked on a course-planning app, a collaboration whose subject was pleasantly close to home. Students had courses to arrange; the builders were students. The distance between making a decision and discovering whether it helped could be measured in conversations around campus.

Wang's early work also included software engineering at Meta and a stint at Bain Capital Ventures. He and Lee built Pe•ple, a community platform that was acquired by Bettermode. These experiences supplied different views of a software company: writing it, selling it, watching other people finance it. Their next venture would make them revisit the first two with uncomfortable frequency.

Four days to change the pitch

In 2021, Wang and Lee made Figstack, a tool that explained code in ordinary language. They built it in ten days, launched on Product Hunt and attracted thousands of users. The opening reception looked encouraging. The return visits were less convincing. People would try the product without acquiring a compelling reason to keep it.

They decided to change direction. Then YC invited them to interview in four days.

They could present the idea they had applied with, or arrive with something else. They chose something else and assembled a new product, demo and landing page before the interview. The partners questioned the change. That evening, Dalton Caldwell called with an offer to fund them for the Winter 2022 batch. Their willingness to reconsider an idea had counted in their favour.

Admission did not settle the product question. One useful clue came from Figstack's users: some were putting its explanations back into their code as documentation. Wang and Lee developed Doc Writer, an extension that generated docstrings inside a developer's working environment. Usage grew. Revenue remained troublesome. A later attempt to charge $7.99 a month prompted many users to uninstall.

The customer who supplied the missing feature

By September 2022, the pair were trying the documentation platform itself. Their weekend prototype became Hyperbeam's site. They added the ability to edit content by the time Hyperbeam needed it a month later. A product roadmap can sound grand until it has an appointment with a customer who wants to change a paragraph.

The same willingness to intervene shaped the next customers. The founders offered to act as technical writers, spending hours or days cleaning up and migrating documentation. Later, they prepared new versions of potential customers' docs before making contact. The demonstration arrived with much of the persuasion already done.

Lee remembers another revealing request from an early inbound customer. The team wanted to document APIs; Mintlify supported ordinary content pages. Wang spent the week and weekend building the first version of an API playground. Instead of treating the request as a distant feature suggestion, he made it the work immediately in front of him.

These stories explain something that a count of pivots cannot. Wang and Lee were learning which parts of their own labour customers valued. The awkward migrations revealed what onboarding needed. Repeated questions exposed missing capabilities. Selling and building happened close enough together that a complaint could become code before it became an item in a meeting.

Good taste has a revision history

Wang's concern with presentation is unusually specific. In his account of Mintlify's design work, he describes roughly 150 Figma versions of the API playground before the team settled on a design. The speed of that first implementation had not ended the argument about how it should feel.

He rejects the assumption that developers are satisfied with any interface that technically works. They have tasks to finish, preferences and limited patience, like everybody else. Making documentation pleasant to use can determine whether someone understands an API quickly enough to keep working with it. A documentation page is often the first sustained encounter a developer has with another company's product.

“Developers are users too.”

Han Wang, on Mintlify's design principles

His preferred design largely gets out of the reader's way. Navigation, search and layout should make the task feel natural. He has also described asking Vercel founder Guillermo Rauch about the point at which an onboarding experience feels right enough to preserve. The interest is practical: when should a team keep iterating, and when has it earned the right to stop moving the furniture?

That attention has a domestic counterpart. Wang introduced a Thursday gathering called docs and donuts, in which the team sits down together and updates its own documentation. Glazed donuts provide an incentive to inspect the product as a customer would. For a company selling the importance of maintained knowledge, there is something reassuring about putting the maintenance on the calendar.

A different sort of reader arrives

In April 2026, Mintlify announced a $45 million Series B, co-led by Andreessen Horowitz and Salesforce Ventures, at a $500 million valuation. The company put its reach at more than 20,000 companies and over 100 million people annually. Those figures describe the business Wang and Lee built; the change in its audience explains where he wants to take it.

At the time of the announcement, Mintlify said nearly half of documentation traffic came from AI agents and AI-assisted workflows. The instructions that once helped a developer integrate an API were increasingly being consulted by software acting for that developer. Wang's ambition expanded accordingly: keep company knowledge accurate and available wherever those systems need it.

That adds consequences to mundane errors. An old instruction may be repeated by an assistant. An incomplete description can prevent a tool from being used correctly. The job of maintaining documentation follows the information into support systems, internal handbooks and coding workflows. Wang is betting that companies will need to manage those connections as deliberately as they manage their software.

On October 20, Mintlify introduced Autopilot in beta. It reads connected code and workplace context, drafts documentation changes and opens pull requests for people to review. The mechanism is recognisably related to an early experiment the founders tried: notice that software has changed, then help its explanation catch up. The intervening years have supplied a larger product around that problem.

The founder still opening the editor

There is a small tension in Wang's work. He sells tools that reduce the effort required to maintain knowledge, yet his own origin story depends on an extravagant amount of manual effort. That makes sense when the purpose of the effort is to discover what deserves to be automated. A founder correcting somebody else's documentation gets an education along with the invoice.

Even the company's public humour is willing to puncture the tidy biography. Wang lists “Not graduating college” among his LinkedIn honours, with a joke about disappointing his parents. It is a more revealing detail than a polished claim about having always known his destination. He has left room for the untidy parts of getting there.

His recent account of running Mintlify includes time spent testing and fixing the admin dashboard himself. The CEO still wants to encounter the little frictions a user encounters. That preference connects the school timetable, the course planner and the first customer site. Each put him close to the person trying to get something done.

The original Hyperbeam page was missing quite a lot. What it possessed was a pair of founders willing to answer when the customer needed something changed. Wang has spent the years since trying to build that responsiveness into software. The editor eventually arrived. The habit of paying attention was there first.