As software eats the world and takes control over the jobs and services we care about, it also generates an awful lot of notifications: 3 new emails. 4 new interactions on Twitter. 30 minutes until your next meeting. Your Uber is arriving now. Notifications have essentially become the third runtime of the phone, after apps and the web, and they’ve become increasingly rich and actionable – as well as omnipresent, and sometimes exhausting.

So if they’re so important, what is it that they do, exactly? Over the last few weeks, I’ve posed these questions to a variety of people. Many of them jumped quickly to answers such as: ‘They’re how you find out about things, like email.’ ’They’re how you know what apps to check.’ ‘They interrupt you.’ One person went old school, and mentioned this old museum piece: ‘You’ve got mail!’ That made me smile. The funny thing is, ‘You’ve got mail’ actually used to be helpful, back when we received a handful of email messages per week. But somewhere along the road, ‘You have unread messages’ turned into this:

As email grew from ‘occasional notice’ to ‘constant torrent of inbound messages’, a signal that you have unread email – even one specifying how many unread emails – evolved from useful, to stressful, to sarcastically un-useful. ‘You’ve got mail’ at one point was helpful, because it resolved some ambiguity: it answered the question ‘Do I have any unread emails’ with a definitive Yes or No response. But when you always have unread email, the real question becomes ‘Is anything in there critically important’. ‘You have 17 unread messages’ doesn’t answer that: the true state of your inbox remains ambiguous until you actually go inspect them all. This notification is no longer effective like it used to be, because ambiguity persists.

This is a key point to understand: Effective notifications are ones that resolve ambiguity. If ambiguity persists, the notification is not effective. This may seem obvious, but it actually runs somewhat orthogonal to what most people say that they prioritize in notifications: the notion of ‘relevance’. Many of us will argue that the problem with notifications is that they get too many notices from Expedia advertising promotions on travel packages and other crap they don’t care about, and that’s why they’re overwhelming. I understand that this Expedia notification can be a little annoying, sure, but in my opinion it’s pretty harmless – you can easily and confidently dismiss it in an instant, secure in your knowledge that you have not missed anything important. (Of course, you could always ‘tune’ your notifications, i.e. shut them off, to get rid of all of these interruptions. But then you risk missing the important stuff, and most people aren’t willing to make that tradeoff.) You know what breaks your concentration a lot more? “You have 2 unread email messages, and you’re not going to know what they are until you open your mail app.” This is a much bigger interruption: those emails might be important, but they also might not! Ambiguity persists unless you stop what you’re doing and go check your email. That’s a much bigger problem.

The same problem applies when your phone is in your pocket, and you’re notified about some inbound message by a brief noise or vibration. Here’s where that Expedia notification becomes a real distraction: when your phone buzzes and you have to reach into your pocket, take it out, swipe down, inspect the notification, and only then realize that the interruption wasn’t useful. (And now your phone is open, so you’re probably going to spend the next 3 minutes on Twitter or something like that.) Why is this such an unnecessary distraction? Not, I will assert, because the message wasn’t important. Non-relevance wasn’t the problem – the ambiguity of the notification was the problem.

It’s also why I’ve changed my opinion to some extent about Apple Watch. I had previously written that Google, not Apple, was the company best suited to build a useful wrist-notification device – because they know more about you, and could therefore deliver more contextually-relevant and appropriate notifications at the right times. I still stand by that statement, but I recognize now that context and relevance aren’t everything when it comes to notifications. Simply by moving the subject line of a given notification to within a quick glance, Apple Watch is useful because it helps resolve ambiguity right there in front of you, without having to take out your phone. That’s big.

The Apple Watch example speaks well to the larger problem – why we need notifications at all. Apple Watch – just like your notification tray – isn’t really the right place to consume content, write long messages, or do any ‘serious work’. So why do we care about them? Broadly speaking, we need notifications because the amount of inbound information addressed to us usually exceeds the amount of bandwidth we have available for immediate attention. Effective notifications help solve this problem by resolving ambiguity towards whether incoming information requires our immediate attention, or if not, how it should be triaged. Ambiguity resolution is the rate-limiting step in our intake of inbound information.

So if ambiguity resolution is the rate-limiting step in our intake of inbound information, then what factors determine that rate? I’d argue there are basically two that matter: 1. The quality and clarity of information presented, and 2. The amount of bandwidth we have available. And right now, we’re struggling to keep up with that intake. So what can be done? We have two options. We can either ‘improve’ notifications in some way – which is what nearly everyone I talked to offered as a solution. Make them smarter, make them more relevant, make them more actionable – sure. But all of these suggestions feel like incremental improvements that are squeezing last drops of efficiency out of an otherwise saturated pipe. So what’s left? The other option – which I believe is actually a much better solution – is that we can go find more bandwidth.

So let’s ask the key question: where is there room?

To get to that answer and shed light on where I think we’re going, let’s look at where we used to be: the multi-headed beast called Outlook. Outlook was the one place where our email, calendar, tasks, and the rest of our short term memories resided. It was a single channel for all those functions – and it was totally overloaded. It felt a bit like today, actually: a constant stream of incoming notifications and limited ability to resolve ambiguity efficiently.

But then came the iPhone, and everything changed. Outlook got unbundled into dozens of different apps, and we went from checking our Outlook at discrete times to having our phones on and with us always. These two shifts drastically expanded our bandwidth for inbound uptake. Work email could be categorized and handled separately from iMessages, Twitter interactions, and all sorts of new notifications that didn’t exist before. And we handled it okay, because the shift from one-channel to multichannel and from discrete inbox checking to continuous phone alerts jacked up our bandwidth considerably.

But then things got unruly. The permissionless innovation that makes the mobile internet so great also gave us lots and lots of parallel channels to attend to, and it became a serious pain to deal with them all separately. So what happened? All of those alerts got bundled into the notification tray, giving us one place where we could find them all. See a pattern here? It’s the bundling and unbundling cycle at work – and as Jim Barksdale of Netscape said twenty years ago, “There’s only two ways I know of to make money: bundling and unbundling.” It’s a pretty safe bet that, at any given point, we’re in the middle of a bundling or unbundling cycle – it’s just often very hard to recognize until later on, in hindsight.

So now where are we? Our capacity to handle inbound information, which had been unbundled into Apps and then bundled into the notification tray, is getting saturated. It overwhelms us because as that capacity becomes saturated, our ability to resolve ambiguity – the rate limiting step of inbound uptake – becomes increasingly taxed. And the suggestions I hear from other people- ‘make notifications smarter, less intrusive, more time sensitive, more contextually relevant’ – only reinforce the notion that we’re squeezing the last drops of efficiency out of a pipe that’s otherwise full. That pipe is the time domain. And if time is tapped out, where should we look? Where is there room?

No, not space space. I mean right here, on our phones, already with us and in plain sight. I’m talking about our brains’ ability to scan fine spacial details, seamlessly take in spacial information, categorize, process, and remember visuospatial input is incredibly powerful, and barely leveraged with today’s notification systems. A sizeable chunk of our brains – roughly speaking, the upper-rear part – is dedicated to processing information about space, movement, and location. And it’s very good at it: we visually experience a coherent, stable, uninterrupted view of where things are in the world, yet that representation is continuously being formed on a millisecond-basis out of thousands of abstract and ambiguous visual components. Before this becomes too much of a science talk, just remember this – the amount of neural horsepower that our brains have evolved to process incoming ambiguous spacial information is a lot more powerful than the parts we’re currently using to handle notifications – specifically, our language processing and short term memory areas. We’re ready for our next major unbundling: an unbundling of notifications from one discrete physical place (the notification tray) out onto broader spatial real estate.

How would this look in real life? I think we’ve been offered a glimpse of the future from – wait for it – Facebook. Specifically, their new brilliant implementation of Chat Heads as an interaction layer. For anyone who hasn’t had the benefit of using Chat Heads yet, they’re completely different than anything else you’ve seen from your phone, notifications and interactions-wise. When you receive a new message, instead of that message going into the notifications bar with everything else, it pops up as a little floating bubble that looks like the face of its sender. You can then manipulate these heads like they’re physical bubbles – you can open them up, push them around, shove them in the corner to deal with later, or flick them away and out of sight. You can stack them into layers, deal with them as groups, or manipulate them individually whenever necessary. And best of all, instead of relying on your short term declarative memory to handle these inbound packets and resolve ambiguity- like you have to use when dealing with a notification tray – you can lean on your visuospatial memory instead, and the substantial neuronal horsepower it has to spare. It may seem like a small difference, but it’s actually a big deal: you’ve shifted from using a part of your brain that has a lot more bandwidth available. This isn’t squeezing a drop of efficiency out of an existing pipe – it’s widening the pipe. A lot.

Chat Heads right now are still very basic – there isn’t a ton you can do with them that you couldn’t do before in some way. But I think there’s huge potential for them and for similar initiatives. It feels a little bit like using Slack right now. Sure, Slack right now is mostly just chat – for the time being. But just like how with Slack you can feel the enormous potential for helper bots and channels to automate so much of our workday, with Chat Heads and our visuospatial working memory I can imagine an ocean of potential that’s waiting to be properly used. The old adage “You only use 10% of your brain” is usually a bit of a nonsense statement, but here’s it’s actually somewhat applicable.

It also would make for a very nice continuation of the bundling-and-unbundling arc that has shaped notifications over the last ten years, and could keep it going for another ten or more. Starting from Outlook (a single channel we checked at discrete times), we unbundled into multiple channels (Apps) that were with us all the time, and widened our ambiguity-resolution bandwidth considerably. But then that got messy, so we bundled all of those channels into one physical place (the notification tray), which was very efficient but ultimately became saturated – we ran out of resolution in the time domain to handle everything at once. I’m willing to bet that the next big unbundling will be an unbundling of the notification tray out into physical space – across our screens, maybe in our watches, who knows wherever else- but in a way that dramatically improves our ambiguity resolution bandwidth.

So what should we expect in the next few years? I imagine we’ll see more apps and services try to escape the confines of the notification tray, just as Facebook has done with Chat Heads, and that a lot of people will protest it as a ‘huge invasion of blah blah blah’ or whatever . Then, something subtle but important will happen: we’ll stop complaining about it, because it won’t feel like a big deal anymore. It’ll just feel normal. We’ll learn how to take in, manipulate, and distribute our notifications across various physical spaces and environments- maybe the watch or something else will play a strong role here, or maybe not. Our capacity for information and notification uptake will keep on rising, because we’ll have learned new strategies for handling and resolving ambiguity. And nothing will feel that different: no huge changes, no sharp adjustments. But the tick, tock, tick of the bundling-unbundling cycle will keep on driving us forward, and preserve for another day the only constant thing about the internet: always more, always more.