This episode features Han Wang and Han Bi Li, co-founders of Mintlify, a documentation platform serving major companies like Anthropic, Microsoft, and Coinbase. They're joined by A16Z General Partner discussing how AI agents are transforming documentation from reference material into operational infrastructure.

The conversation explores Mintlify's journey through eight pivots before finding product-market fit, their approach to serving demanding enterprise customers, and how documentation requirements have evolved as AI agents increasingly consume docs directly. The founders share insights on building for developers, scaling customer success, and the future of knowledge management in an AI-driven world.

Eight Pivots to Find the Obvious Solution

Han and Hanbi spent "a solid year and a half wandering in the desert while chewing glass" through multiple failed attempts at developer tools before landing on documentation

The breakthrough came from a weekend prototype: "We forced ourselves to build this very basic version in two days - that's all we got"

Their roommate's company Hyper Beam became the first customer after seeing improved docs: "How do I get it set up right now? Give me access to your cloud for DNS configurations"

"The beauty of having failed so many times was when something did work, there was no mistaking it" - Han on recognizing true product-market fit

AI Agents Transform Documentation from Reference to Infrastructure

Documentation has shifted from explanatory to operational as "coding agents, support bots, and internal AI tools now read documentation directly"

"When they're wrong or outdated, systems break" - incorrect documentation now has real-time consequences across thousands of AI agents

The market expanded dramatically: "It went from being an application to really feeling like an infrastructure product - content that powers AI"

Usage patterns are shifting rapidly: "Maybe 50% for humans, 50% for AI now, but it'll be 10% humans, 90% AI by end of year"

Self-Healing Docs Solve the 25-Year Problem

"No one ever feels good and has gone to bed thinking 'oh, I have great docs'" - the universal pain point of outdated documentation

The core problem: engineers have the most context but "they're not the ones who want to, are incentivized to, and they're not the ones who are paid to" maintain docs

Opus 4.5 was "a big unlock for us because the model became reliable and consistent enough to actually go do it" for self-updating documentation

Enterprise trust has shifted: "A lot more companies are comfortable giving away their context to LLM agents - that was not the case two years ago"

Serving Lightning-Fast Enterprise Customers

Working with AI labs revealed extreme responsiveness: "You would send a Slack message at any time and someone responds within 10 seconds"

Microsoft impressed with their pace: "For a company as large as they are, they are moving lightning fast" - Han on enterprise agility

Anthropic's reaction to Mintlify's speed: "Do you guys have to get on call whenever we send a message?" after receiving instant responses

Early customers drove innovation: Anthropic translated all docs to 12 languages overnight, inspiring Mintlify to build similar capabilities

20 Million Monthly Users and Expanding Beyond Developers

"20 million people came across a Mintlify site last month, whether they know it or not" - Han on their reach and impact

The platform expanded beyond developer docs to help centers, internal documentation, and knowledge management across HR and engineering teams

"A lot more of the world is getting more technical - more people understand Markdown now because it's the language of LLMs"

The mission remains constant: empowering builders like "the Han when he was 11 years old learning to code for the first time"

Resources Mentioned

cuts that's making it so time consuming and cumbersome

umentation. It's an important part, but it's sort of one time. Like there's just a lot of thousand paper cuts that's making it so time consuming and cumbersome. Where now we can probably have a reall

Self-Healing Docs Solve the 25-Year Problem

fting rapidly: "Maybe 50% for humans, 50% for AI now, but it'll be 10% humans, 90% AI by end of year" Self-Healing Docs Solve the 25-Year Problem "No one ever feels good and has gone to bed thinking