Your most requested feature? Might be the worst thing you could build.

That's one of the lessons from my conversation with Mitchell Tan, co-founder of Kondo.

Kondo is the Superhuman for LinkedIn DMs — a tool that turns the messy LinkedIn inbox into something you actually want to use. Keyboard shortcuts, labels, split inboxes, reminders. Built by a 3-person team. No public roadmap. And they say no to almost everything users ask for.

I've been using Kondo for months and it's genuinely changed how I handle LinkedIn. If you use LinkedIn professionally, you have to try it.

So I had to get Mitchell on the podcast to understand how they think about product.

🧠 𝗪𝗵𝗮𝘁 𝘆𝗼𝘂'𝗹𝗹 𝗹𝗲𝗮𝗿𝗻 𝗶𝗻 𝘁𝗵𝗶𝘀 𝗲𝗽𝗶𝘀𝗼𝗱𝗲:

0:00 - What is Kondo and how it started
1:40 - From recruiting firm to building a Superhuman-inspired SaaS
7:45 - Why "copying" great products is harder than it sounds
9:11 - The invisible UX details that make Superhuman feel faster (even when it's not)
13:52 - How LinkedIn DMs are fundamentally different from email
19:43 - Why Kondo has no public roadmap and ignores most feature requests
23:52 - The 3-question framework to decide what to build next
38:52 - How to define your aha moment (and why it's different for everyone)
49:33 - Why one-on-one onboarding calls still matter (and when to stop doing them)
55:51 - Forcing users through onboarding vs letting them explore freely
1:01:14 - Linear's hidden UX gems that inspire Kondo

💡 𝗔𝗰𝘁𝗶𝗼𝗻𝗮𝗯𝗹𝗲 𝘁𝗮𝗸𝗲𝗮𝘄𝗮𝘆𝘀 𝗳𝗿𝗼𝗺 𝗠𝗶𝘁𝗰𝗵𝗲𝗹𝗹:

→ 𝗕𝗲𝗳𝗼𝗿𝗲 𝗯𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗮𝗻𝘆 𝗳𝗲𝗮𝘁𝘂𝗿𝗲, 𝗮𝘀𝗸 𝟯 𝗾𝘂𝗲𝘀𝘁𝗶𝗼𝗻𝘀
Does it let you charge more? Does it convert non-payers? Does it reduce churn? If all three are no — don't build it.

→ 𝗖𝗼𝗽𝘆 𝗴𝗿𝗲𝗮𝘁 𝗽𝗿𝗼𝗱𝘂𝗰𝘁𝘀, 𝗯𝘂𝘁 𝘂𝗻𝗱𝗲𝗿𝘀𝘁𝗮𝗻𝗱 𝘁𝗵𝗲 𝗶𝗻𝘁𝗲𝗻𝘁 𝗳𝗶𝗿𝘀𝘁
Superhuman tracks your mouse direction to adjust UI response times. You can't just copy pixels — you need to understand why things work before deciding what to borrow.

→ 𝗞𝗻𝗼𝘄 𝘄𝗵𝗮𝘁 𝗡𝗢𝗧 𝘁𝗼 𝗰𝗼𝗽𝘆
Superhuman's split inbox works for email. But LinkedIn DMs are different — all conversations with one person live in one thread. So Kondo built labels that create split tabs instead. Same goal, different execution.

→ 𝗙𝗼𝗿𝗰𝗲 𝘂𝘀𝗲𝗿𝘀 𝘁𝗵𝗿𝗼𝘂𝗴𝗵 𝗼𝗻𝗯𝗼𝗮𝗿𝗱𝗶𝗻𝗴 — 𝗯𝘂𝘁 𝗸𝗻𝗼𝘄 𝘆𝗼𝘂𝗿 𝗮𝘂𝗱𝗶𝗲𝗻𝗰𝗲
Superhuman blocks clicking during their tutorial — keyboard only. Kondo allows clicking because not everyone buys for shortcuts. Be opinionated, but match it to why people actually pay you.

→ 𝗣𝗲𝗿𝗰𝗲𝗶𝘃𝗲𝗱 𝘀𝗽𝗲𝗲𝗱 > 𝗮𝗰𝘁𝘂𝗮𝗹 𝘀𝗽𝗲𝗲𝗱
Kondo had the same response times as Superhuman but felt slower. The difference? Micro-animations. A subtle slide when you archive. Small details that make software feel alive.