DEV Community

Cover image for I Built a Mac Menu Bar App Because I Kept Saying "Wait, What?" in Every Meeting (Live Demo 🚀)

I Built a Mac Menu Bar App Because I Kept Saying "Wait, What?" in Every Meeting (Live Demo 🚀)

Varshith V Hegde on September 13, 2026

Someone on a Zoom call said a URL yesterday. I was still writing down the previous bullet point. By the time I looked up, they had moved on. I mute...
Collapse
 
pushpendraagrawal profile image
Pushpendra Agrawal •

on-device only until you hit cmd+s is the right default. curious what happens to the buffer if the source app switches audio devices mid call, like bluetooth headphones dropping to the mac speaker. does the coreaudio tap survive that or does it just lose the last 30 seconds

Collapse
 
varshithvhegde profile image
Varshith V Hegde •

It still survives it can have multiple audio sources connected if you want that

Collapse
 
pushpendraagrawal profile image
Pushpendra Agrawal •

the notarization step is always the surprise tax on these things. works perfectly on your machine then gatekeeper eats it for everyone else. are you doing the signing/updater plumbing solo or leaning on something like Sparkle from day one

Collapse
 
varshithvhegde profile image
Varshith V Hegde •

Agreeed I am actually stuck on this parttt

Collapse
 
divyasinghdev profile image
Divya •

This is such a great idea.
All the best for further versions of this.

Collapse
 
varshithvhegde profile image
Varshith V Hegde •

Thanks

Collapse
 
jo-do profile image
Jo Do •

The mute-type-wait ritual is the perfect origin story because it is a latency problem wearing a comprehension problem's clothes: the information existed, just thirty seconds too late and only after you interrupted yourself to ask. A menu bar that holds the last few utterances solves the real thing, which is that meetings are streams and note-taking is a lossy sampler. The best tools come from scratching the itch you hit every day, and "the URL was already gone by the time I looked up" is about as daily as it gets.

Collapse
 
varshithvhegde profile image
Varshith V Hegde •

Thank you🙌

Collapse
 
maxslashwang profile image
Max/Wang •

That’s a great idea, bro! I actually had the idea of making a live caption app for macOS before, but running a local AI model was a bit tricky, so I ended up dropping it 😄😄

Collapse
 
varshithvhegde profile image
Varshith V Hegde •

Actually whisper is a very light model and too accurate too

Collapse
 
maxslashwang profile image
Max/Wang •

Yeah, exactly. Whisper is very accurate, but it wasn't really designed for streaming inference. I also tried a few streaming models through sherpa-onnx, but I couldn't optimize them

Thread Thread
 
varshithvhegde profile image
Varshith V Hegde •

Ohh got it

Collapse
 
irusik profile image
Ира Иващенко •

That’s a really great idea! I think this approach could definitely make the process much more interesting and convenient. I love when new ideas and opportunities come up to improve something or try a different approach.

Collapse
 
varshithvhegde profile image
Varshith V Hegde •

Thank you soo much

Collapse
 
vaibhav_srivastava_f543ba profile image
Vaibhav Srivastava •

Great breakdown! Especially appreciate the clear architecture explanation and practical on-device design decisions. Thanks for writing this!

Collapse
 
varshithvhegde profile image
Varshith V Hegde •

Thank You for reading 🙌