Buddy AI Note
Hand Buddy the whole task. Get the files back.
141 followers
Hand Buddy the whole task. Get the files back.
141 followers
Buddy AI Note is an AI task assistant built around your daily note. Hand Buddy a task and get the files back: slides, documents, spreadsheets. It also drafts email replies in your voice. You see the price first and approve what goes out.
Free Options
Launch Team / Built With
Render Sandboxes: a computer for every agentRun untrusted code safely. Idle sandboxes cost nothing
Promoted
The plan-then-confirm boundary is the right design instinct. The question is whether you've drawn the line in the right place.
"Research brief" running on one click feels underspecified. If Buddy is searching the web, summarising sources, and writing a document, that's not obviously safer than sending an email. What's the actual definition of "safe" here and is it user-configurable or hardcoded by you?
The bigger validation question: how many of your current users have actually let Buddy execute a task rather than just organising their notes? The memo-to-plan loop is useful but that's a glorified to-do list. The execution layer is where the real value lives and also where churn will happen if it gets one step wrong on something important.
What does a successful week look like in your usage data right now?
@sergio_jivan This is a very good question we've gotten on here, thanks for actually digging in. On "safe", you're right that "low-risk" was a lazy word. The real line isn't how much work the step does, it's who it touches and whether you can undo it. A research brief just writes a doc into your own workspace, so if it's garbage you delete it and nobody ever saw it. A sent email is gone the second it leaves, sitting in someone else's inbox with your name on it. The clearest example: blocking focus time on your own calendar runs automatically, but the exact same action stops and asks the moment there are other attendees on it, because now it's reaching someone else. So a research brief can totally be wrong, it just can't hurt a relationship, and that's the specific thing the confirm step is there to catch. On whether that's hardcoded by me or yours to change, this is the part I'm happy to point you to. The defaults are mine, but they're only defaults. There's a settings section with a row for every action Buddy can take and three options each: always ask, auto unless it touches someone else, or always run. "Execute all" only fires the ones you've set to auto. So if you think research briefs should ask first, flip that one and they will. The line I picked is just the starting point. On the to-do list thing, you're completely right. The memo to plan loop on its own is a fancy checklist, and the execution layer is where the actual value is and also where people will leave if it gets one important thing wrong. That's the bet and yeah, it's the scary part. On the data, honest answer is we launched on mobile today, so I don't have an execution rate I'd trust enough to quote you. I'd rather tell you that than throw out a number based on 10 people. The week I'm watching for isn't memos written, it's how many people let Buddy actually run a step and then come back and do it again the next week. That repeat is the whole thing. Ask me in three weeks and I'll give you the real number, and I'll come back here to tell you what it turned out to be.
The hard part with note-to-task tools is handling ambiguous input, which is most notes most of the time. Curious what the approach looks like under the hood - structured extraction or something more generative?
Do you have plans for Evernote or OneNote integration? And how are you thinking about sync consistency when the same note gets edited across platforms?
@demi_tan Good question ! It's the core problem. Under the hood it's generative reading, structured output: the model reads the messy note as-is (no required format), then emits typed tasks through a tool call, each with a confidence score and a plan of discrete steps. So it's not regex/structured-extraction that breaks on ambiguity, but it's also not free-form generation, the structure is imposed at the output layer, not demanded at the input.
Two guards keep ambiguity honest: a grounding rule (every step has to trace back to something you actually wrote, no invented tasks, no generic advice), and we surface the confidence rather than hide it, so a vague note produces a low-confidence suggestion you can reject in one tap instead of a wrong action. Multi-topic notes get split into separate threads rather than flattened into one summary.
No Evernote/OneNote integration planned right now, we're focused on the notes→action loop inside Buddy first. Sync consistency across external editors is exactly why; we'd rather not promise it until we can do it without the "edited in two places" race. Appreciate the questions.
Congrats on the launch! I love the memo-first idea.
Asking as a neurodivergent person: how does Buddy handle a voice note that jumps between five completely different things?
@marie_saxon Thank you, and honestly this is one of the cases we most want to get right. A voice note that jumps around is normal, that is just how thinking out loud works. Buddy does not try to flatten it into one tidy summary. It pulls the separate threads apart, so a ramble that touches five different things comes back as those five things, each as its own item you can act on or ignore. The goal is messy in, structured out, you should never have to organize your thoughts before you say them. That is the whole point. I would genuinely love for you to try a real rambly one and tell me where it does and does not hold up, that kind of input is exactly what makes it better.
@maxence_leguery_podtech This is so cool! It sounds like something that could genuinely make life much easier. Thank you!
@marie_saxon 🙏 Thank you Marie, that genuinely means a lot. If you give it a go, I would love to hear how it holds up for you. Either way, thanks for the kind words and for asking the question that made me explain it properly.
Turning notes into an actual daily plan is the part most note apps miss. The interesting question is whether the plan stays grounded in what the user actually captured, or starts becoming generic productivity advice. Curious how you keep it practical day to day.
@vidur_saini Good question ! This is exactly the failure mode we were scared of. The plan only ever gets built from what you actually wrote or said, not from some generic "here's how to be productive" template. Buddy pulls the concrete things out of your memo and turns those into steps, so if you didn't capture it, it doesn't invent it. It also leans on your own past notes and docs rather than a generic model, which keeps it sounding like your day instead of a self help blog. The one rule we hold is that every step should trace back to something you said. And the longer you use it, the more it pulls from your own history, so it gets more like you over time, not more generic.
When Buddy turns a memo into a plan, can you see what it extracted and edit it before anything gets scheduled?
@karimbenkeroum Yes. That's exactly how it works. When Buddy turns a memo into a plan, it lays out each item it extracted as a separate step, and you can open any step to edit the details before it runs. Anything that actually touches your calendar or email like scheduling, sending or rescheduling defaults to Always ask, so it waits in a review queue until you approve it.
Would love to hear how it feels once you try it 🙌
Works on iPad too?
@syed_shayanur_rahman On iPad the best experience right now is the web app at ainote.tech, runs great in Safari. A native iPad version is on the way as the current mobile app is designed for iPhone.