You walk out of a meeting. There were six action items. You remember three of them clearly. Two more come back when you check your calendar — the meeting after the one before this one had a similar topic, so it jogs something loose. The sixth never resurfaces. The person who needed it never raises it. By the time you find out it mattered, the deadline has slipped and you're three days into a small disaster you could have avoided.
Around the fire is where stories begin. Bonfiyah is where they live — including the parts of them you would have lost.
The forgetting is bigger than you think.
A meta-analysis of meeting-effectiveness research from MIT Sloan in 2018 found that the average professional retains roughly 20% of what was discussed in a meeting one week later. Recall of specific decisions and commitments — the part you actually need — is even lower. Most knowledge work runs on the assumption that meetings produce shared understanding. What they actually produce is a set of approximate, partially-overlapping versions of what each person thinks happened.
This is not a discipline problem. People take notes. Notes get filed. The notes are still partial, because note-taking is a sampling process — you write down the things that catch your attention, in the order they catch it, and the rest dissolves. The action item that was casually mentioned at minute 47 doesn't get written down. The hedge that someone added to a "yes" doesn't get captured. The price that was settled in passing turns into "I think we agreed on forty thousand" three weeks later when the contract is being drafted.
What you've been doing about it.
The historical answer was meeting minutes. The modern answer is some combination of the calendar invite, your notes, a few messages summarizing what was decided, and a follow-up email. None of these are wrong. All of them are partial. Together they produce a record that's roughly the shape of what happened, with the resolution dialed down to whatever someone had time to capture in the moment.
The recent answer is AI summaries. Otter, Notta, and the meeting-bot category produce a bullet-list summary of every meeting they record. This is better than nothing. It is not the right shape for the work, because bullet-list compression is exactly the wrong way to remember a conversation. The summary captures topics; it doesn't capture commitments. It tells you what was discussed; it doesn't tell you who is on the hook for what.
The hook is the thing.
Here is the underlying observation that became Bonfiyah: most of what you actually need from a meeting recording is a structured list of commitments — who promised what, by when, in their own words, with the exact quote anchored to a timestamp you can play back if needed.
Bullet-list summaries don't produce that. Action-item extraction tools that ship with meeting bots produce a flattened version that loses the speaker, loses the deadline-or-no-deadline distinction, and loses the conditional nature of half the commitments — "I'll send the proposal if Mike approves the pricing" collapses into "send the proposal," and the condition that actually governs whether it happens is gone.
Promise Tracker is the Pro AI feature built around this observation. It extracts every commitment from every recording — yours, theirs, the conditional ones, the dated ones, the vague ones — attributes each to the speaker, anchors it to the exact moment in the audio, and tracks it to closure. When the deadline arrives and the deliverable hasn't, it surfaces the original quote, not a paraphrase. Because the speaker attribution holds across recordings, the same person's commitments roll up cleanly whether they showed up in last week's standup or this morning's call.
Why this matters.
Consider the small disaster you avoided. Sarah said in the standup that she'd send the brief by Wednesday. You moved on; she had four more conversations that day; on Wednesday she remembered there was something due but couldn't remember whether it was the brief or the QBR slide. She sent the wrong one. The brief came in Friday. The client review got pushed. The launch date slipped a week.
This costs nothing measurable. It also costs everything that doesn't get measured — the trust erosion, the small re-prioritization that consumed two people's Thursday, the slight tax on the next conversation between you and Sarah where neither of you mentions it but both of you remember.
Pre-Brief is the other side of the same problem. Before your next meeting with Sarah, Bonfiyah generates a one-page recap: open commitments — hers and yours — what was said last time, and what's been resolved since. You walk in already caught up. So does she, when she opens her own Pre-Brief on her phone before the meeting. The whole tax — the forgotten-commitments tax, the partial-recall tax, the unresolved-from-last-time tax — quietly disappears.
What this isn't.
Bonfiyah is not a productivity-philosophy product. It is not going to teach you how to run better meetings, set OKRs, or deliver feedback. It does not require a method or a habit. You record the conversation; the system extracts what mattered; you act on the result. If you're already doing the work of remembering everything yourself, more power to you — Bonfiyah is for the rest of us.
It also is not a meeting bot. It does not join your video calls. It does not email participants to tell them they're being recorded. There's a built-in consent module instead, which surfaces the two-party-consent rule and captures verbal consent in the room — a quieter and more respectful way to handle the legal layer than a bot announcing itself to the call.
On privacy: real-time transcription runs on-device. Audio leaves your iPhone only for the optional cloud-transcription pass you control. And we do not train AI on your transcripts. Those are the boundaries, stated plainly, because vague privacy claims are how this category loses people's trust.
What to do with this.
If "forgetting commitments" is a tax you've been paying without naming, the cheapest experiment is to install Bonfiyah, record three meetings with it, and let the Pro AI suite run on the free trial. It's a universal Apple app — iPhone, iPad, and Mac, synced over iCloud — so it follows you between the room and the desk. By the end of the week you'll either find it changed how you walked into your next meeting, or you won't. If you didn't, the trial costs you nothing. If you did — and the failure mode of the past five years suggests you will — you keep it running and the small-disaster pattern stops being a thing.
Start free, and let the trial do the arguing.
The fire keeps the story warm. Bonfiyah keeps it accurate.
— Richard
