Inside Our Product Hunt Launch

Introduction

This is How We Grow, where we write up what we're doing to grow Kai. Two days after our Product Hunt launch, we recorded this one in a single room for the first time, and David, Kai's co-founder and CEO, sat in with Lambert and me. I wanted three things from him: what a launch is for when you've been live for weeks, how to read a fifth place, and what we're changing now.

Prefer to watch or listen? Here's the full episode

What the launch was for

A deadline for the whole team

Kai was already live. We had paying customers and subscribers well before Tuesday 13 October, so I asked David what launching even means at that point.

His answer: with AI it's always easy to add one more improvement, so without a date the product is never ready. A launch is an artificial deadline that forces the whole team to push as hard as it can, and ours made us go through every step of the funnel:

1. The paywalls we had to work through.
2. The first seconds in the app, once a new user is in.
3. The positioning: what we're telling the world Kai is, and who we're talking to.

What we got out of it, in David's words, is a V1: "this is our baseline now." More traffic and a big creator picking us up would always be welcome, but as a baseline to iterate on, the launch did its job.

Every channel we could reach

For Lambert, launching as a growth marketer meant making sure the world knows, so 2 months before the date he sat down and mapped every channel we could use:

1. Morgen's users, starting with the ones interested in AI.
2. Kai's own users.
3. Both websites, morgen.so and hirekai.ai.
4. Social: reels on Instagram, LinkedIn and X.
5. A big YouTube creator we work with, talking about Kai.
6. Friends and family, brought in early to follow what we're building and make some noise on the day.

Lambert's launch board, as it stood 11 days out.

Fifth, and why it helped

Top one or nothing

Before the launch I asked what we wanted out of it. Lambert didn't hesitate: top one or nothing. On Product Hunt you're either first or you're the rest. Nobody brags about second or third.

We finished fifth. The night before, Lambert had asked Claude for a forecast of the day, and its dashboard said about 321 votes and around #4.

Claude's forecast, the night before.

195 votes, and twice the usual traffic

The morning went to plan. Product Hunt hides the vote counts for the first four hours, and when they appeared at noon, Lisbon time, we were fourth with 110 votes, ahead of the 87 Claude had forecast by then. The afternoon didn't. The forecast counted on Product Hunt's own visitors to keep voting, and the next 20 hours added 85 votes.

Muse Wonder passed us at 14:30. We then traded fourth place with Fina Business until 04:15, when its overnight run of close to 100 votes took it past us for good, and we closed the day fifth with 195. Articos MCP won with 325.

The real day, vote by vote.

The website did better than the ranking. hirekai.ai had 677 visitors on launch day and 738 the day after, about twice the 365 a day of the two weeks before, and its best two days since early September. Product Hunt sent 479 of the 1,412 visitors over those two days, Instagram 145 and Google 128.

Visitors on hirekai.ai, around the launch.

Lambert's first reaction was anger. Two days later his read is different. Product Hunt on its own didn't bring what we expected, and it's mostly people asking other people to vote, which he doesn't find healthy. But everything we did around it brought traffic and new users, and the morning we recorded, the first new paid users had come in. The launch also forced the rebuild of the onboarding, the website and the pricing page, with a reel every 2 days on top.

Product Hunt has changed

David has launched there before. When Morgen did, about 1,200 users came in that day, with 20 to 30 other products launching. Now it's hundreds a day. On our Tuesday, 567 products were posted, and 29 of them made the homepage. Building is cheap, people relaunch every few days, and the platform has to fight bots. The messages offering to sell us votes started the minute we went live.

His harder point is about the badge itself. Product of the Day is a vanity metric. It doesn't tell you that you have product-market fit; only a week or two of data does, once you can see whether people stay and whether they pay. When Morgen made Product of the Day, we read it as the moment to scale, and that call shaped the next 3 months of strategy in the wrong direction, when we should have kept working on the product.

The question we'd have skipped

My own read is that the flop gave us something. If we'd finished first, I'm not sure we'd be asking what we're really building, and for whom. A win in disguise.

David's version: Kai is in an insanely overcrowded market, where everyone is building a personal or AI assistant, and 3 to 10 launches a day sit close to what we launched. Being top of that group for a day wasn't enough. So the work now is to get extremely sharp, and it starts with the aha moment. Kai gives value in several places, and we haven't decided which one is the moment a user is hooked:

1. The first email they send with Kai.
2. The first time they see what Kai has done.
3. The first time Kai does the work for them.

Picking one decides what we optimize for, from the message to every step of the funnel. David's bet is one clear niche, served all the way.

The Instagram growth hack

The best thing to come out of launch week wasn't on Product Hunt. Lambert started spending more and more of his time on social, and he found what I think is a real growth hack.

It started with Jonathan, who joined us to make content: watching his process gave Lambert the push. Around the same time Opus 5.5 came out and turned out to be good at editing, and we had the raw material most people lack, close to 1,000 videos recorded since February. Lambert had a list of content ideas he'd never had time to edit, so he started running them through Claude. The first results weren't great, and he iterated until they were.

Then Instagram gave the account its trial reels: a new reel goes first to people who don't follow you. Lambert calls it the biggest hack he's seen to grow an account, because once the editing is automated you can post many versions of one idea and keep what works.

Runaway, the reel. One reel, set to Kanye West's Runaway, covered the whole journey from February to launch day. Lambert posted 20 versions of it, 3 or 4 a day. The account went from 300 followers to more than 700, and people started messaging to ask what we're building, because they'd watched the journey.

The reel, from 300 days to launch day.

The first sign-ups and activated users from social are already coming in. Lambert's rough math is that 10,000 followers in 2 months could mean around 100 new users every 2 or 3 days, and he'd be the first to say it's a guess. David wants in too: part of why he joined the episode was to give Lambert's pipeline clips of himself.

We had partners pushing too. Tool Finder, whose YouTube channel has half a million followers, made a big video on Kai ahead of the launch.

The messier parts

• We missed the goal. Fifth, not first, with 195 votes against a forecast of 321, and no Product of the Day badge.
• Product Hunt is a voting game now. Hundreds of launches a day, bots, and people selling votes. On its own, it didn't bring the users we expected.
• We moved too fast in places. Lambert spent the 2 days after the launch cleaning up the week's work so it can be repeated.
• Followers aren't customers yet. 400 new followers and the first sign-ups from social are a start, not a channel we can count on.
• We still haven't named our aha moment. Until we do, we can't say what we're optimizing for.

What's next

The work now is a post-mortem, as David puts it: keep what worked, scale it, and narrow down. That means the positioning (one niche, served sharply), the aha moment, and the content machine, with David's clips feeding Lambert's pipeline.

If you want to follow along, the whole series lives at How We Grow, and the changelog shows what we ship week to week.