10,000 microwave enthusiasts to attend annual microwave conference in Las Vegas
Pushing on the asymptote.
Ok, this would be a weird event. But, like, why? Microwaves are everywhere; they are unambiguously valuable; some people1 use them nearly every day; they’re powered by an almost magical technology that, it turns out, actually can support lots of conferences.
Why not have a convention for the appliance enthusiasts too? Imagine the possibilities! Industry leaders gather to talk about the latest trends in countertop microwave technology. They share bold predictions on how the microwave landscape is changing, and explain why it’s critical for every family to invest in modernizing their microwaving capabilities. Breakout sessions teach people creative ways to use microwaves—you can’t keep up with the Joneses if the only thing you’re heating in your microwave is food. Microwave champions tell stories about how they’ve fully migrated their households away from legacy ovens and broilers. Microwave vendors launch daring new products—a microwave with wifi, a microwave with AI, a microwave durable enough for the enterprise kitchenette, a microwave that opens from the left. Every family today needs a microwave, we’re told. Having a microwave is table stakes; you’re falling behind the modern family if you don’t have a microwave. All the happiest families have a microwave.
Despite all this potential, my naive guess is that this conference doesn’t exist because nobody cares about what big microwave builds next. As useful as the appliance is, there isn’t much else for it to do. Sure, microwaves could be marginally better—Apple could make one that is remarkably simple to use; Toshiba could make one that cooks a Hot Pocket evenly. But today’s microwaves do the job they were intended to do, they all do it well enough, and we don’t need them to do anything else. Even Wirecutter—a professional quibbler if there ever was one—agrees: “The newest models don’t work any better than old ones did.” No matter how many 4K displays Apple puts on a microwave, it’s still a basic appliance, a utilitarian commodity, a fifty-year-old revolution with no second act.
In other words, microwaves are a product with a ceiling. The relationship between how advanced a microwave is and how useful it is has hit its asymptote. It peaked as a “glorified microwave popper,” and not the transformational cooking tool that the food industry thought it would be:
The microwave oven has become one of the country's most beloved appliances. About 87 percent of American households have at least one.
But despite its popularity, the microwave has failed to live up to its early promise. Once touted as a revolutionary cooking tool, it has become primarily a reheating device -- most often for coffee or tea…
"When the microwave first came out, people thought they had stumbled on nirvana," Mr. Vierhile added. "It's not the appliance the food industry thought it would be. It's a major disappointment."
What’s notable about this story is that it’s not declaring the microwave dead; to the contrary, the microwave is everywhere. It’s obituary for the microwave’s potential.
The modern data appliance
For those us who work in data, it’s a worthwhile warning: Data, just like the microwave, has no God-given right to be important.
It’s surprisingly easy to forget that, if you work in this industry long enough. We’ve been reciting the same talking points for so long—every company today needs to be data-driven; being data-driven is stakes; you’re falling behind the modern company if you aren’t data-driven; all the most successful companies are data-driven—that we’ve built an entire ecosystem on the implicit assumption that they’re true. It’s an unquestioned truth that corporate data contains vast amount of value; that tomorrow’s businesses will infuse more decisions and processes with data that yesterday’s; that data-driven companies will outperform their Philistine competitors; and that all of this inevitable. Despite the various problems we’ve had in driving adoption or finding value—so much so that the Harvard Business Review has a beat writer for failed data projects—our collective confidence endures: The future will be built on data.2
But there’s a different future that’s also out there: One in which data is everywhere, but the revolution is nowhere. It’s one where data infrastructure mostly exists to feed BI tools, BI tools mostly exist to crank out dashboards, and the whole thing is little more than a utilitarian appliance. Good decisions may not be buried in our warehouses, waiting to be data sciened out. Massive efficiencies may not be an a few streaming pipelines and ML models away. Reporting could be our plateau; our databases could be haystacks, full of nothing but hay. And our fitful struggles to reinvent how businesses are run may not be because of our failures as data teams, or because we’ve chosen to focus on tools over bUsInEsS vAlUe; they may be because data, as a means for delivering that value, is tapped out.3
The harsh truth is that if you look over the long arc of the last ten years—from when Redshift kicked off this whole cycle off until now—you’d have to conclude that we’re more appliance than avant-garde; more operational grease than organizational revolution. Take BI, for instance. In our 72 attempts to redefine BI, we still haven’t broken the dashboarding barrier. No effort to do it differently—BI for business teams! For startups! For the enterprise! Open source! With a code-free semantic layer! With no semantic layer! With an external semantic layer!—has escaped the gravity of basic reporting.4 If at first you don’t succeed, try, try again. If after 72 times you don’t succeed, maybe something’s wrong?
Or, try a 73rd time—but now, with large language models.
If you wanted to create an addictive app for short-form videos, you could do in one of two ways. You could either pay a bunch of famous actors and directors to make polished content and hope that they do a good job; or, you could make it easy for anyone to make anything, figure out how to identify the best stuff, and show that to people. Quality creation or curated quantity; choose your fighter.
Quibi got steamrolled partly because it underestimated how good amateur content creators could be—but mostly because it underestimated the the immense power of their scale. An infinite number of monkeys typing on an infinite number of keyboards won’t just reproduce Shakespeare; they’ll create something far better.5 Quibi failed because it fought that; TikTok has been transformative because it captured it.
For tomorrow’s BI to be more than today’s BI—and data tooling to be more than an appliance—we need to capture the same thing.
If we could accelerate how quickly people can get answers to novel question, everything changes. Companies would go from having a handful of people pushing on analytical boundaries to having hundreds or thousands. People would look up simple questions as freely as we google for things, without wondering if the answer was worth the time it took to find it. Not only would this make organizations much more informed, but it’d also help them uncover far more of the truly meaningful insights that we so often chase but can’t find.
Call me a cultist,6 but I think LLMs could actually get us there. My rough theory of BI is something like this: Most people are actually reasonably good analysts, and plenty capable of asking good questions without a professional chaperone. But we don’t see this because their methods are imprecise; there’s a lot of trial and error. If extracting an answer from a BI tool takes too long—i.e., minutes instead of seconds—people aren’t able to cycle through enough trials to cover meaningful ground. Rather of recognizing this incomplete analysis, we score it as bad analysis, and put guardrails around it, to protect people from making mistakes. Do the mundanes stuff, we say, but data teams will do the meaningful work. Shoot your own home videos, and leave the important content to the professionals.
The only interface that will let people get the answers fast enough and painlessly enough to break this cycle is natural language. The corporate masses won’t learn to code; drag-and-drop OLAP data cubes are too confusing; and I refuse to endorse a world where we all walk around with a View-Master strapped to our face. If we can figure out how to make a natural language interface work—truly work, where people can answer complex questions, and be as creative with it as they are in their spreadsheet munging—we can turn Quibi’s curated and fatally constrained production studio into a TikToks’ crowdsourced content engine.
If there’s value in our data, it’s the people that will find it. And if the data ecosystem has a second act—one that’s worthy of a 10,000 person conference in Las Vegas—it’ll be the people that will power it.
Speaking of conferences and Las Vegas, Mode—my employer, who tolerates this blog in exchange for the periodic plug—is doing lots of things there for Snowflake Summit. There with be a booth (2123-C); there will be a happy hour (fun fact: it’s normal bowling but my grandfather holds world record in duckpin bowling, which is a real game and not a DuckDB event); there will an executive suite with food and drinks (and probably a microwave though idk for sure); and there will be a fancy dinner (it’s invite only, hmu if you want to come, I have some sway but not that much tbh; I’ll also be there, which if you’re like, I’d go if you weren’t there, register on the website and list me as a dietary restriction, and you’ll probably win that fight).
It’s also striking that other corporate disciples are considerably more understated about their value. Products and businesses can win markets with great design; in fact, its advantages are probably less ambiguous than those that come from being data-driven. And yet, I rarely hear breathless claims that companies can’t win without being design-driven, or that we all face an urgent need to modernize our design tooling. It’s perhaps telling that they’re content showing people the value they currently provide, whereas we feel compelled to tell people about the value we could provide.
Though plenty of other people have raised similar doubts, most skeptics put the blame on us: We think too much about tools and not enough about stakeholders, business objectives, and The Bottom Line. That’s fine (though as I’ve said before, I think tooling is the only durable way to make things better); still, that solution rests on implicit assumption that data is very valuable, if only we use it right. I’m not so sure.
This isn’t to say there aren’t businesses to be built here. People typically buy products for relatively mundane reasons: They’re easier to use; they’re cheaper; they’re more closely aligned with their personal preferences. There’s money to be made selling a more comfortable chair without revolutionizing how we sit down.
This is roughly analogous to how AI models learn. Volume, it turns out, is a far more effective teacher than skill.