The length of an audio message is proportional to how much I’ll raise my eyebrows when looking at it.

It’s like a wall of text: It looks bad, nobody wants to read it, except if it’s Britney on Instagram. For everyone else: We have newlines, paragraphs, subheads—ways of formatting things, to create a rhythm that people want to read. Still, every popular messaging app has the exact same interface: Everybody gets blobs.

How can we fix this?

Wait, why don’t people just send smaller messages?

First of all, if you do: Congratulations, please accept this award: ✂️. Hooray! If the rest of us could do it, you wouldn’t be reading this post. We’re not doing it on purpose. We simply don’t know how long we’re going to talk, nor what we’re going to talk about. There’s no script. That’s how oral communication works.

Unfortunately, we’re going to have to see if the computer can help us out.

✂️ Hey, how are ya ✂️ [rustling intensifies] ✂️ By the way, guess who got […] ✂️ Say hi to […] ✂️ Tschüssi ✂️

Now that’s better. Unfortunately, the computer is not a born DJ. We can’t send every message to a room full of bored students for cutting. Even worse, they would probably disagree based on their field of study. I once studied a tiny bit of CS and now I mistakenly believe that every problem can be solved with a computer program.

Deep Learning is exactly the right problem for this solution.

Surely the ✂️ algorithm will get it ✂️ wrong? ✂️✂️✂️💥

Yes, but here’s a good rule of thumb: If the computer is doing “smart” things, give constant immediate feedback to the user. As the message gets longer, show when it gets split. Let the user pause and resume at any point, freely move the ✂️ right before sending.

I’m out of words, unwilling to delete more words, and eager to try this out!