Boring AI

A plea to stop the hype

I’ve been thinking about what it takes to build something boring. The kind of software that does what it’s supposed to do and then gets out of your way, that becomes so reliable you stop noticing it entirely.

That kind of boring is actually very hard to achieve. It requires a certain maturity, not just in the technology, but in the people building it. You have to care more about solving problems than about being seen solving them. You have to be willing to do unglamorous work: fixing edge cases, handling errors gracefully, making the same thing work the same way a thousand times in a row. You have to resist the urge to add features just because you can, to make the product louder just because you’re loud. Boring software is the result of a thousand decisions made in favor of the user’s life over the builder’s ego.

I don’t see much of that maturity in AI right now.

The Parasite Strategy

What I see instead is AI being crammed into everything that already works. Browsers, operating systems, productivity suites, search engines — tools that spent decades earning trust through reliability and are now being asked to host something that hasn’t earned anything yet. The strategy is transparent: AI can’t build trust on its own, so it borrows trust from software that already has it. Find a stable host, inject yourself into it, hope the credibility transfers.

It doesn’t, though. Trust is a relationship built over time through consistent behavior. When you stuff an unreliable system into a reliable one, you don’t make the unreliable system more trustworthy, you make the reliable system less so. And people notice. They don’t articulate it in these terms, but they feel it: the tool that used to just work now requires vigilance. The software that used to disappear now demands attention. Something that was theirs has been colonized by something that wants to be seen.

This is happening everywhere, and the backlash is mounting. People are asking why their browser and their operating system need to be agentic, why every piece of software they rely on suddenly has opinions and intentions and memory and feelings. The answer, naturally, is that AI needs distribution, and the easiest distribution is through products people already use.

The promise of AI that would actually matter to people is, for some confounding reason, being ignored. Wasn’t the dream that AI would handle the invisible, tedious layer of your life so that you can focus on what matters? I feel like instead, the past three years have been a super-experiment in extreme data collection and AI psychosis that has nothing to do with helping you out.

I do still think the dream is achievable! The technology is capable of it, at least in constrained contexts, and it’s getting better every day. And people would be thrilled if all they got is a tool that knows its place and has no stake in their real life.

But that’s not what’s being built. Or rather, it’s being built, but it’s not what’s being sold. Because “practical tool” doesn’t sustain the narrative that AI is the most important technology in human history. It doesn’t keep the hype cycle spinning. So instead we get AI as protagonist, AI as companion, AI as emergent being on the verge of consciousness, or superintelligence, or whatever the next chapter of the mythology requires.

The Values Beneath the Product

There’s something deeper going on here, something about what the people building this technology actually believe matters in life. The tech oligarchs talk constantly about transformation, disruption, and being on the right side of history. They build bunkers and develop long-termist philosophies and position themselves at the center of civilizational narratives. Disruption is the whole point. They seem almost disappointed when the apocalypse doesn’t arrive on schedule.

This is a particular set of values, and it’s baked into the products. When you believe that transformation is inherently good, you build things that transform. When you believe that being at the center of history is the highest achievement, you build things that demand to be noticed. When you think stability is boring and ordinary life is something to be transcended, you don’t build tools that help people live their ordinary lives better.

But most people just want their crap to work. They want to get through their days with enough time and energy left over for the things that actually matter: relationships, rest, the small pleasures that make life feel worthwhile. Deep down, they’re hoping to be left alone.

The mismatch between what’s being built and what people actually need is, of course, a values problem. The people with the most power to shape these tools have a fundamentally different vision of what technology should do (and what life should be) than the people who have to use them.

Maturity

What would it look like for AI to grow up? What if it was boring? It would mean building for reliability instead of novelty, prioritizing consistency over capability, and saying no to features that compromise stability, even when they’re supercool. It would mean treating users as people with limited attention and real lives.

That kind of maturity requires something the AI industry doesn’t seem to have right now: humility. You can’t eat a slice of humble pie as you’re running through the metaverse breaking stuff. You won’t have the patience to solve boring problems boringly. The understanding that the most transformative technologies are often the ones that learn to disappear would mean that someone shuts off the spotlight, and that is clearly unacceptable.

The spreadsheet didn’t need to be conscious. The word processor didn’t need to be your friend (let’s save Clippy for a different post). They just needed to work, and they did, and they reshaped the entire world anyway. AI could be that. It could be the next Excel, the next quiet revolution that changes how people do whatever without ever requiring them to care about it. That would be more impressive, in many ways, than the mythology. But it would require the people building it to want something different than what they seem to want.

I’m not holding my breath. But I’m also not giving up on the possibility. The hype will end eventually, and when it does, maybe we’ll finally get the boring AI we were promised.

That would be enough. That would be more than enough.