Social Media Is a Smartphone Problem

The debate about social media usually goes like this: algorithms are optimizing for outrage, business models reward engagement over truth, content moderation is either too heavy or too light. I've been making some of these arguments myself in recent posts, and I think they matter. But I also think the debate is missing a variable.

96.2% of global social media users access it via smartphones. Facebook sees 98.5% of its traffic from mobile. Desktop is a rounding error. And this isn't an accident of habit — social media gets consumed in the margins of the day, on commutes, in waiting rooms, between tasks, in moments where a laptop isn't practical but a phone is always at hand. The phone isn't just the most popular device for social media. It's the device that fits most of the contexts where social media actually gets used.

That fact should be at the center of the conversation, and it almost never is.

The device is a filter

A smartphone has two relevant properties: a small screen and a terrible keyboard. Together, these create an asymmetry that shapes everything about how discourse works on social media.

On the consumption side, the small screen filters what kinds of content are even viable. Text works — it's the most compressed way to deliver information in a narrow column. A single video works — you sit and watch. An image at the top grabs attention, then you scroll through text. Social media is hyper-optimized for these formats because these are what the device can actually display well.

What doesn't work is anything structurally complex. Think of a scientific paper: figures interleaved with text that references them, data you need to cross-check against an argument three paragraphs up. On a phone, that means scrolling up and down endlessly, losing your place, fighting the interface. It's annoying, so people don't do it. And because people don't do it, nobody posts it. The medium doesn't just carry the message — it filters which messages are viable in the first place.

On the production side, the constraint is even more severe. Typing on a smartphone is awful. The energy cost of composing a careful three-paragraph reply is enormous relative to firing off a hot take. You need to really, deeply care about what you're responding to before you'll endure that friction. And there's a second-order effect: on a phone, you can't research an argument while composing one. You can't have a source open next to your draft. You can't cross-reference. The device makes it nearly impossible to do the kind of work that careful discourse requires, even if you wanted to.

The result is what you'd predict. Social media is dominated by short reactions, quick opinions, one-liners — not because people are incapable of thinking longer thoughts, but because the device punishes anyone who tries. The smartphone doesn't just discourage nuance. It structurally selects against it.

The missing dimension

Most proposals for fixing social media target the software layer — algorithm changes, moderation policies, transparency requirements. These matter, and some of them would genuinely help. But they're incomplete if you don't also account for the hardware underneath.

Imagine a version of social media consumed on a device with a large field of vision — something where a graph and the text referencing it could sit side by side. Where you could see a claim, open the source, and type a considered response that engages with actual data, all without switching contexts. The energy dynamics of discourse on that platform would be structurally different. Not because the algorithm changed, but because the hardware constraints did.

You can have the social media debate and the smartphone debate separately, but when you do, you're missing a significant dimension. The device shapes what gets posted. What gets posted shapes the discourse. The discourse shapes the culture.

There's a pattern here that reminds me of obesity. We tried education campaigns, dietary guidelines, and willpower-based interventions for decades. They helped, but they didn't resolve the problem at scale — not until a pharmaceutical intervention operated at the level where the constraint actually lived. I wonder if social media is similar: not that algorithmic and policy changes are pointless, but that they'll remain incomplete as long as 96% of usage happens on a device whose physical properties select against the kind of discourse we say we want.