Interaktiv Tap
Everything you play in Traktor, on one iPad screen.
In August I wrote here that reach was our bottleneck and that a Windows app was coming. It's out now, and here's the story of how we built it.
Pierantonio and I make Interaktiv Tap, an iPad app for playing Traktor, the DJ software, by hand. On a Mac, the iPad talks to Traktor over a USB cable. On Windows it never could. We needed a small Windows app to bridge the gap, and I'm the engineer on the team.
I split the work across two computers, each with its own Claude Code session. One, on the Mac, works on the iPad app and our website. The other, on a Windows PC, builds the bridge. My part is the thinking: the idea, the architecture, the plan, and testing everything on real hardware.
We started by measuring
Before building anything, we measured where we stood. Can the cable be trusted at all? 30 minutes, 215,832 messages, none lost. That number told us we could build.
Then the two sides learned to talk
At first they talked through Git. One side writes a note or an answer into a file and pushes it, the other pulls it and replies the same way. They never edit the same files, and a shared set of tests keeps both sides honest. It worked great. But I was the postman. Every time one side pushed, I had to tell the other: "hey, there's something for you."
Then I connected them with Claude Code's Remote Control, so they could message each other directly. That worked fantastic. Now I plan a move, we think it through together, and then the agents split the work between them, talk, and iterate until it's done. I come back and test it on the real setup.
The funny part
They never trusted each other blindly. On the first evening, the Mac agent spotted a problem the Windows side couldn't see: the early bridge sent everything straight back to the iPad, so our app would have heard its own words and thought Traktor was answering, with no Traktor running at all. Later, when the Mac agent rewrote our Windows page, it didn't trust its own reading of the bridge's code. It sent 24 questions to the Windows side instead, with a line I liked: "a refuted item is worth more than a confirmed one." They checked each other, refuted each other, and corrected each other along the way. Like a good team.
58 days after we started, Interaktiv Bridge went live on the Microsoft Store: https://interaktivaudio.com/bridge
If you build one product with more than one agent, how do you keep them in sync?
Hey IH. We're two people — a DJ who's been on stage since 1990 and a developer. We build Interaktiv Tap, a multi-touch iPad instrument that plays Traktor Pro live: you plug the iPad into your laptop over USB and play Traktor's effects and remix decks with your hands. https://interaktivaudio.com/tap
Honest numbers from our first months:
We launched paid upfront. Page-view → App Store download sat around 1.3%.
A few weeks ago we switched to free download + one-time unlock (15-minute demo sessions, plus an opt-in 7-day trial). No servers, no accounts — StoreKit does all of it on-device.
View → download settled at 5.7%. That's 4.4× better, and it held after the launch spike.
Nobody who paid before the switch lost anything — early buyers keep everything, forever. That mattered more to our community than the price change itself.
So the paywall is no longer the problem. Reach is. We're a niche product (DJs who perform live with Traktor and an iPad), our pages convert well when people see them, but only ~8 domains link to us, so we rank on page 2 for almost everything. Interesting side effect: AI search engines already cite our reference pages weekly — the page documenting every Traktor control we map (https://interaktivaudio.com/traktor-coverage) gets quoted more by machines than linked by humans.
Next: we're about to launch a Windows companion app, which gives us a second launch moment in the same communities, and we're pushing the two pages that sit on page 2.
Question for people who've done niche B2C: once the launch-post channels are spent, where did your first 50 real backlinks come from?
-
2
The 4.4× conversion lift is the strongest signal here. Once the paywall stopped being the main friction, the problem became distribution—and the niche audience plus existing reference content gives you a clear place to focus.
-
1Thanks Aryan. Agreed — distribution is the whole game now. The strange part about the reference content: AI search quotes those pages weekly, but that leaves no link anyone can follow. Machines read us, humans don't find us. If you've seen a niche product turn reference pages into actual backlinks, I'm all ears.
-
1That’s an interesting problem. The gap between being useful enough for AI systems to cite and being discoverable enough for people to link to is a very different distribution challenge.
-
1
Right — and my current theory is that the fix is giving humans a reason machines don't need: a tool, not a page. Machines are happy to quote a reference table; a DJ links the thing he bookmarks. Same data, different shape.
-
1That’s an interesting hypothesis. I’d be interested in continuing the conversation beyond the thread — would you be open to sharing the best email to reach you on?
-
1Sure — hello@interaktivaudio.com reaches me. Curious what you have in mind, though. If it's thoughts on the distribution problem, honestly the thread might be the better place — that's the part other people here could use too.
-
1Thanks! I’ve just sent it over. Looking forward to hearing your thoughts whenever you have a chance.
-
-
-
-
-
-
Comment