← All speakers

Hahnbee Lee

Conference affiliation: Co-Founder & CTO · Mintlify · 2026

Hahnbee Lee is co-founder and chief technology officer of Mintlify, which builds documentation infrastructure for developers and AI agents. Her work spans the systems that organize documentation, give writers control over publication and keep instructions aligned with changing software. As agents increasingly rely on those instructions, she has examined how stale documentation can turn into incorrect answers and actions.

On this page

Finding demand for documentation infrastructure

Hahnbee met co-founder Han Wang through Cornell’s Design & Tech Initiative, where they built a course-planning application together. They later created and sold Pe•ple, a community-management platform. She also worked at Duolingo and had an offer to return full time, but chose to build Mintlify after the founders were accepted into Y Combinator’s Winter 2022 batch. In her account of that decision, she describes learning to interview customers, change direction and become less attached to a particular product as she moved from building technology to building a business.

Their route into documentation began with Figstack, a tool that used OpenAI’s Codex to explain code in English. Launch attention did not translate into sustained use. One behavior proved more revealing: users were copying explanations back into their code as docstrings. Hahnbee and Han followed that behavior into an editor extension that generated docstrings directly inside the developer’s workflow. The extension attracted users, but finding a business remained a separate problem.

The founders also tried detecting when code changes made documentation stale. The stronger commercial response came when they offered documentation hosting: teams wanted a carefully built developer portal without maintaining the infrastructure themselves. They initially centered the product on Markdown and GitHub version control, fitting engineers’ existing tools. Mintlify’s development through these experiments explains the move from generating individual pieces of documentation to operating the system in which teams maintain and publish it.

Giving documentation teams better tools

Hahnbee’s technical writing explains the design decisions behind that infrastructure, including how teams describe navigation and control which content reaches readers.

  • A navigation model that matches the documentation. In her account of Mintlify’s move from mint.json to docs.json, Hahnbee explains why coordinating separate fields for tabs, anchors and versions had become difficult to reason about. The team replaced those interdependent fields with one recursive structure. Teams could nest languages, versions and tabs without making the documentation hierarchy mirror the repository’s folders.
  • Control over content and publication. Hahnbee argues that faster drafting alone cannot remove documentation bottlenecks. Writers also need to keep material internal, reuse content and reorganize pages safely. Her article on the writer’s development experience illustrates those needs with .mintignore exclusions and relative imports: teams can exclude files from publication and reuse material while organizing repositories around their work.

Keeping an agent’s instructions current

Hahnbee’s case for automated documentation updates begins with a concrete failure: a chatbot gives obsolete billing instructions because its source documentation was never updated. Improving the chatbot’s answer requires maintaining the material it consults.

She describes a team-built Mintlify workflow that combines event or schedule triggers, relevant code repositories and an agent working in a sandbox. The repositories give the agent context for proposed documentation changes. Depending on configuration, those changes can become a pull request for review or be committed directly. The workflow gives teams a choice about how much review to require before an automated edit reaches publication.

Hahnbee also joined Coinbase’s Lincoln Murr for a conversation about AI agents as documentation readers. That changing readership extends the problem her work addresses: documentation must remain navigable and maintainable as both people and software use it to understand a product.

1 conference talk

Loading more talks and highlights…

References