A few days ago, I wrote about the problem.
Waking up.
Opening your phone.
Not knowing what actually matters.
That part hasn’t changed.
What changed is this:
Now I have something running on my phone.
And it’s… quieter.
Not smart.
Just quieter.
The same morning, slightly different
Phone screen.
Still lights up.
Still a list.
But smaller.
Some things are just… gone.
System noise.
Random background notifications.
Stuff I already know I don’t care about.
And for a second, it feels better.
Lighter.
But then the same question comes back:
Is anything here actually important?
And I still don’t know.
That’s the gap.
That’s the part I haven’t solved yet.
What exists right now
I didn’t start with AI.
I started with control.
So what’s built is mostly foundation:
- Basic UI (kept intentionally quiet)
- Local database (everything stays on device)
- A simple app flow:
- onboarding
- morning page
- all notifications
- settings
Nothing fancy.
Just enough to run the idea end-to-end.
What the app actually does
Right now:
- It listens to incoming notifications
- Removes obvious noise (system, low-signal stuff)
- Lets me define what to ignore
- Stops completely if DND is active
No intelligence.
No ranking.
No decisions.
Just reducing the mess.
And honestly, even that helps.
But it doesn’t solve the real problem.
The real problem (still)
It’s not volume.
It’s uncertainty.
Even with fewer notifications, I still hesitate.
Still scan.
Still second-guess.
Because “less noise” is not the same as “clear signal”.
Where this gets difficult
The next step sounds simple:
Figure out what matters.
It’s not.
Because “important” is not a fixed rule.
It changes:
- based on who it’s from
- when it arrives
- what’s happening in your life
- what you’re expecting that day
You can’t hardcode that.
What I’m building next
Two parts.
Both uncomfortable.
1. On-device anonymizer
Before anything leaves the phone:
- Strip identity
- Reduce messages into patterns
- Turn them into abstract signals
Not:
"John: call me now"
More like:
someone important + unusual time + direct request
2. Decision layer
Somewhere (likely server-side):
- Learn what I actually respond to
- Detect urgency vs routine
- Return a simple answer
Not a feed.
Not a summary.
Just:
this mattered
The part I don’t trust yet
This is where I’m careful.
If the system gets this wrong, even occasionally:
- I stop trusting it
- I go back to checking everything
- The whole thing collapses
So accuracy matters more than features.
And predictability matters more than intelligence.
Still chasing the same thing
Nothing changed from the first post.
I don’t want better notification management.
I want this:
You wake up.
You look once.
You know.
That’s it.
Top comments (0)