My Notes app had turned into a graveyard of ideas I was sure I would come back to. I never did, and some of them were the best ideas I had all year.
They always showed up at the worst time: walking somewhere, halfway through dinner, deep in client work. I would type three messy lines, close the app, and by the next morning I could not tell what I meant. Having the idea was never the hard part. Keeping it was.
So I built Voice Memo. You open it, talk in plain English for 15 or 30 seconds the way you actually think, and close it. Everything after that is the app's job.
Say "I have a meeting at 7 PM, remind me at 6:30" and your phone or Mac rings at 6:30. Mention a task for Thursday, and on Thursday it sits at the top of your day. Every memo lands in the right project, so ten days of talking about one client end up on one page, with a brief that keeps itself updated.
It also keeps up when you change your mind. Say on Monday that you will redesign the homepage, then on Tuesday "cancel that, fix the checkout first", and Monday's task comes off the list and the old decision is marked replaced. Every change has an Undo.
The part I use most is the Claude connector (MCP). I record "change the button colour on the website" at night, and in the morning Claude reads it and gets it done.
I did not want another subscription, so the rule was $0 from day one. I did not build an AI. I wired free APIs together on a free Cloudflare account.
On privacy, to be precise: it is not on-device. To transcribe and understand a memo, the audio and its text go to the AI providers you add keys for, Groq by default with Cloudflare Workers AI as a fallback. Your transcripts, notes and projects are stored in your own Cloudflare account, the audio copy there is deleted after 30 days unless you pin it, and there is no server of mine in between.
It is single-user, so you deploy your own copy. That takes a Cloudflare account, a free Groq key and a terminal. No iOS app yet, the web app works there.
Android, Mac and web, open source under MIT: https://github.com/hussaintrawad...
Would love to know two things. What would stop you from self-hosting it? And what should it pull out of your voice notes that it does not yet?
PH 用户
to answer the question you asked at the end: people. I'll mention "talked to Sarah about the pricing deck, need to follow up next week" in a memo and that's a relationship thread, not just a task, it has history that matters beyond the one follow-up. right now that probably gets filed under whatever project I was talking about, with the person's name buried in the text. a lightweight per-person view that pulls every memo mentioning someone would turn this into something closer to a personal CRM, which is a much bigger reason to keep using it daily than task extraction alone. also appreciate that you wrote out the privacy tradeoff (audio going to Groq/Workers AI) instead of just saying "self-hosted" and leaving people to assume it's fully local.
My Notes app had turned into a graveyard of ideas I was sure I would come back to. I never did, and some of them were the best ideas I had all year.
They always showed up at the worst time: walking somewhere, halfway through dinner, deep in client work. I would type three messy lines, close the app, and by the next morning I could not tell what I meant. Having the idea was never the hard part. Keeping it was.
So I built Voice Memo. You open it, talk in plain English for 15 or 30 seconds the way you actually think, and close it. Everything after that is the app's job.
Say "I have a meeting at 7 PM, remind me at 6:30" and your phone or Mac rings at 6:30. Mention a task for Thursday, and on Thursday it sits at the top of your day. Every memo lands in the right project, so ten days of talking about one client end up on one page, with a brief that keeps itself updated.
It also keeps up when you change your mind. Say on Monday that you will redesign the homepage, then on Tuesday "cancel that, fix the checkout first", and Monday's task comes off the list and the old decision is marked replaced. Every change has an Undo.
The part I use most is the Claude connector (MCP). I record "change the button colour on the website" at night, and in the morning Claude reads it and gets it done.
I did not want another subscription, so the rule was $0 from day one. I did not build an AI. I wired free APIs together on a free Cloudflare account.
On privacy, to be precise: it is not on-device. To transcribe and understand a memo, the audio and its text go to the AI providers you add keys for, Groq by default with Cloudflare Workers AI as a fallback. Your transcripts, notes and projects are stored in your own Cloudflare account, the audio copy there is deleted after 30 days unless you pin it, and there is no server of mine in between.
It is single-user, so you deploy your own copy. That takes a Cloudflare account, a free Groq key and a terminal. No iOS app yet, the web app works there.
Android, Mac and web, open source under MIT:
https://github.com/hussaintrawad...
Would love to know two things. What would stop you from self-hosting it? And what should it pull out of your voice notes that it does not yet?