What I can tell you is what we've learned from watching people actually use the thing, which is the part I find more interesting than any chart anyway. Some of it confirmed what we believed. Some of it surprised us. A bit of it we got wrong and have been fixing in the open.
Around the fire, you don't know what a gathering will become until people sit down and start talking. Building in the open is the same — you light the fire and then you watch who pulls up a seat and what they do once they're there. Here's what we've seen in the first sixty nights.
What people actually reached for.
The thing I most wanted to learn was which features people would return to — not which demoed well, but which became habits. The honest pattern: the legible, load-bearing ones held up. Promise Tracker is the feature people describe back to me unprompted, in their own words, which is the surest sign something landed. When a user explains your feature to you better than your marketing does, you built the right thing. The shape of it — who said they'd do what, in their words, tracked to closure — turns into a habit because it solves a problem people feel weekly.
People Memory was the quieter surprise. People don't always talk about it, but they rely on it — the continuity of the same person recognised across recordings is the thing they'd miss most if it vanished, even when it's not the thing they'd name first. It's infrastructure that earns its keep by being invisible, which is exactly how we hoped it would feel. The voice layer underneath it — how we recognise one person across separate sessions — is on the Voice ID page if you want the mechanism.
What surprised us.
A few things genuinely caught us off guard, and surprises are the most valuable thing a launch produces.
The rooms were even more varied than we expected. We built for the in-person conversation the meeting bots can't reach, and we were right that it's a real and underserved space — but the specific rooms people brought Bonfiyah into were wider than our own list. Conversations we hadn't centred in our heads turned out to be exactly the ones some people most wanted captured. The bet on the room, not the call, held; our imagination about which rooms was too narrow, and our users corrected it for us.
People care about the consent tooling more than we predicted. We built consent into every tier because we thought it was the right thing to do, half-expecting it to be a quiet background virtue. Instead it turned out to be something people actively value and mention — the surfaced rule, the captured yes, the exportable log. It seems we underestimated how many people record real conversations and have quietly wanted to do it properly without a good way to. That was a good thing to be wrong about. To be clear about what that tooling is: it surfaces the rule for where you are and helps you capture a verbal yes. It is not legal advice, and it never tells you a recording is fine to make — that judgement stays with you.
What we got wrong (and are fixing in the open).
Building in the open means showing the misses too, so: we shipped some rough edges. The most useful feedback has been about the unglamorous parts — the moments where the product asked the user to do work it should have done for them, or surfaced something a beat too late to be useful, or got a detail wrong in a way that made someone double-check it. None of it was the big stuff; the foundations held. It was the thousand small frictions you only discover when real people use software in conditions you didn't picture. We've been working those down steadily, and the rhythm of ship, watch, fix has been the actual shape of these sixty days.
The one principle that's been reinforced rather than challenged: when in doubt, be precise rather than impressive. Every time we've leaned on the honest, specific version — of a privacy claim, of what a feature does, of what the consent tooling is and isn't — it's aged well. Every place we were tempted toward a rounder, shinier claim, the precise version was the one users actually trusted. On privacy that precision matters most: real-time transcription runs on-device, and audio leaves your iPhone only for the optional cloud-transcription pass you control. The careful version of that sentence has earned more trust than any sweeping one would have. Candor compounds.
What's next.
I'll resist the roadmap-as-marketing move and keep this honest. The near-term work is less about new flagship features and more about making the ones that landed better — tightening the rough edges, making the load-bearing features more reliable in more conditions, closing the gap between what the product does and what it should do in the messy real-world moments. The cross-recording reasoning that makes Bonfiyah different — the AI layer that reasons across your whole library rather than one recording at a time — has a lot of headroom, and the most interesting work is making it sharper rather than adding surfaces alongside it.
We're still building in the open, still shipping fast, and still more interested in what the product gets wrong for you than in telling you it's finished. It isn't. That's the fun part.
What this isn't.
This isn't a victory lap, and it isn't a hockey-stick post dressed up as a reflection. Sixty days is early; I don't know yet which of these patterns will hold and which were the noise of a launch window, and I'd be lying if I claimed otherwise. It also isn't a complaint — the first two months have been genuinely encouraging, just in the quiet, qualitative way that matters more to me than a number I can't yet trust. What it is, is an honest check-in from a founder who said he'd build in the open and meant it: here's what we've seen, here's what surprised us, here's what we're fixing. The numbers can come later, when they actually mean something.
The fire's lit and people are gathered around it. Now the work is keeping it warm.
Pull up a seat.
If you've been watching from the edge of the firelight, this is the open invitation. Bonfiyah is free to start, and runs as one app across iPhone, iPad and Mac with iCloud sync, so wherever you keep your conversations it's there. The most useful thing you can do — for us and for what the product becomes — is use it on your real conversations and tell us what it gets right and what it gets wrong. Sixty days in, the thing shaping Bonfiyah most isn't our roadmap. It's what people do once they sit down.
You never know what a gathering becomes until people start talking. Sixty nights in, this one's just getting going — come tell your story by the fire.
— Richard
