I kept forgetting important stuff in my daily life, so I built an AI second brain called AMP to fix it
Hello,
I’ve been building something called AMP.
It’s a small AI “second brain” that helps you remember everyday things — so you don’t have to rely on notes, memory, or a bunch of random apps everywhere.
Right now I mostly use it for stuff like:
“My passport is in the top drawer”
“Daniel owes me $20”
“Remind me to buy eggs tomorrow”
Then later I can just ask:
“Where’s my passport?”
“What do I need to buy?”
“What do I know about Daniel?”
and it pulls it back instantly.
Nothing fancy honestly — just trying to solve my own problem of forgetting things and having too many apps open.
It’s still early and I’m mainly looking for honest feedback:
- does this make sense immediately?
- would you actually use something like this?
- what feels confusing or unnecessary?
If you want to try it: https://amplyai.org
Appreciate any feedback (good or bad helps a lot)
Thanks!
Trending on Indie Hackers
A one-week feature took two months, mostly spent keeping three systems in sync
The retrieval through natural questions angle is the right thing to lead with, that is what makes it feel different from notes and reminders. One thing stood out the moment I hit the login screen though.
The product holds the most sensitive things a person has, where the passport is, who owes money, the daily schedule, and the one word promise on the page is secure.
But there is no privacy policy, no terms, and nothing that says what secure actually means.
Is it encrypted at rest, who on your side can read it, does the text get sent to an LLM and if so which one and is it retained.
Two reasons to fix this early. First, for this category that disclosure is not legal boilerplate, it is the product, because trust is the entire reason someone would put their passport location into an app.
Second, calling something secure without backing it is exactly the kind of unsubstantiated claim that regulators treat as deceptive, so the safe move is to either describe the concrete security or soften the word until it is true. A short privacy page and one honest paragraph on data handling would do more for conversion here than another feature.
This is incredibly helpful feedback, thank you.
You're absolutely right that for a product like this, privacy and security aren't just legal checkboxes, but they're part of the product itself.
I've been focused on making the memory experience useful, but your point about being specific about what "secure" actually means is spot on. Right now, there isn't enough transparency around how data is stored, processed, or handled, and that's something I need to fix early.
I'm going to prioritize a clear privacy page and be much more explicit about how memories are stored, which AI providers are involved, and what users can realistically expect.
Out of curiosity, what information would you personally need to see before you'd feel comfortable trusting a tool like this with sensitive information?
Honestly the answer is short: I trust the specifics, not the word secure. For a tool holding something like a passport location, here is what I would want to see before typing anything real into it.
First, can you read it. The biggest question is whether the company itself can see my data. If it is encrypted so only I can, say that. If your team can technically access it, say that too and explain what stops misuse.
People can live with either answer, what kills trust is not knowing which one is true.
Second, where my text goes. Name the AI provider the memory text is sent to, and whether it goes through an API that does not train on your data and does not retain it beyond basic abuse checks. That one sentence does more than any badge.
Third, can I get it out and get it gone. A real delete that actually purges, including from backups within a stated window, and ideally an export. Knowing I can leave clean is what makes people comfortable getting in.
After that it is small stuff: where it is stored, how long you keep it, and a two minute privacy page a normal person can actually read. None of that is a feature. It is just you saying out loud what already happens, which is exactly why it converts.
Quick update: your comment pushed me to prioritize this sooner than I otherwise would have.
I've added a privacy page, terms, and a plain-English "How your data works" page that explains who can access data, which AI provider is involved, what happens when data is deleted, and where the current limitations are.
Still a work in progress, but you've definitely changed how I'm thinking about trust in products like this.
You're right. For something like AMP, trust comes from specifics, not labels.
A lot of the questions you mentioned already have answers behind the scenes, but I haven't surfaced any of them clearly enough. That's on me.
Things like who can access data, which AI providers are involved, what happens when you delete something, and whether data is used for training should be easy to find and easy to understand.
Looks like my next sprint is less about features and more about making those answers obvious.
This makes sense immediately — the core idea is basically “structured memory you can query later,” which is a real pain point.
What stands out is that the value isn’t in storing notes, but in retrieval through natural questions. That’s the part that actually feels different from normal note apps or reminders.
The biggest risk I see is trust + correctness. For something like “where is my passport” or “who owes me money,” users will only adopt it if they’re fully confident nothing gets lost or misremembered. Memory systems have zero tolerance for ambiguity.
Also, it may help to be very clear about boundaries — is it just a personal memory layer, or does it also act like a task/reminder system? Those tend to blur, and that’s where users get confused.
But conceptually, the direction is strong: turning scattered notes into a queryable personal context layer is something a lot of people already try to hack together manually.
Really appreciate this.
You summed up the core idea better than I have in some cases. "Structured memory you can query later" is pretty much exactly what I'm aiming for.
The trust point is the one that keeps coming up, and for good reason. If AMP gets something important wrong even once, that confidence disappears fast.
I'm still figuring out the boundaries too. Right now I see reminders and tasks as supporting features, but the main job is helping people remember and retrieve information without having to think about where they stored it.
Curious what you'd personally trust something like this with. Everyday stuff? Important documents? Follow-ups with people?
The core use case is immediately clear and that's a real advantage — most people have had that exact moment of 'wait, where did I put my passport?' Your examples make the value obvious without needing to explain anything.
The question I'd have from a user perspective: what does the capture flow look like? The hardest part of personal memory tools is getting people to actually log things in the moment they happen. A home screen widget or voice shortcut that makes adding a fact feel like near-zero friction could be the difference between daily use and 'I'll add stuff later' (which tends to become never).
Would be curious to hear how your early testers are using the capture side — that's usually where the real behavior patterns show up.
That's a really interesting question.
The more feedback I get, the more I'm realizing the hard part might not be retrieval at all. It's getting people to capture things in the moment.
I'm trying to keep it as low friction as possible with a chat-first approach and voice input because if it feels like work, I know I'd stop using it myself.
Early testers have mostly been saving things they don't want to search for later, but I'm still learning what capture habits actually stick.
What do you use today when you need to remember something quickly?
One thing I'd be careful with:
The interesting question may not be whether AMP remembers things accurately.
It may be what users are actually trusting it to remember on their behalf.
That sounds subtle, but it can quietly shape what gets stored, what feels valuable, and what makes someone keep using it over time.
I wouldn't make that call casually this early.
That's a really interesting way to think about it.
I've been spending a lot of time asking myself whether AMP remembers things accurately enough, but you're right. The bigger question is what people are actually willing to trust it with.
My instinct is that it'll start with small everyday things like where something is, what to buy, or who to follow up with.
But I honestly don't know yet, and I don't want to make assumptions too early. That's a big reason I wanted to put it in front of people now instead of building in isolation.
What kinds of things would you personally feel comfortable handing off to a tool like this?
That's actually why I stopped short earlier.
I don't think my answer would be the useful part.
I think the interesting part is the decision that follows from whichever answer turns out to be true.
Probably easier to explain properly than in a comment thread.
If you're interested, drop your email and I'll send it over.
That makes sense.
You're right that the interesting part isn't the answer itself, it's the product decisions that come from it.
I'd definitely like to hear your thoughts. Feel free to send them to [email protected] or DM me here if that's easier.
Appreciate it.
Just sent you an email.
The trust question is interesting, but I suspect the decisions that follow from the answer are where things get much more consequential.