Nate Sullivan

This is funny, informed, profane, and generally delightful. 

Most organizations cannot ship the most basic applications imaginable with any consistency, and you're out here saying that the best way to remain competitive is to roll out experimental technology that is an order of magnitude more sophisticated than anything else your I.T department runs, which you have no experience hiring for, when the organization has never used a GPU for anything other than junior engineers playing video games with their camera off during standup, and even if you do that all right there is a chance that the problem is simply unsolvable due to the characteristics of your data and business? This isn't a recipe for disaster, it's a cookbook for someone looking to prepare a twelve course fucking catastrophe.

I have no notes. There is going to be some absolutely incredible goalpost-moving over the next five years — unlike crypto, the vague, amorphous “AI” space will be able to declare victory in any number of ways even if the most visible applications never become practical or production-viable. “But Nate, AI is everywhere now!” is a thing you will definitely be able to say without lying in 2026, but also a thing you could have said in 2016 before the market discovered thirsty chatbots and mass-laundering of intellectual property.

As someone who’s been deep (trapped, often) in the Apple ecosystem since 1987, it’s very important to me that the company not… ruin everything, basically. Then again, one of the reasons I’m still here is that Apple 2.0 has been surprisingly good at maintaining the basic customer experience even as they became far, far too large to punk rock their way through things like “should we comply with the request of this lucrative market’s authoritarian regime”.

Being a big public company in 2024 does not usually allow you to prioritize “wisdom". I get it.

As the Generative AI Hype Bubble Spectacular reaches it’s eye-rolling stage — the part where everyone knows how this movie is going to end except the investor class that’s too deep in it now to get out — Apple has apparently assessed the terrain and decided what it’s going to do with all of this stuff. 

1. Throw in commodity transformer text replacement into the operating system level (i.e., shorten this, make it “friendly”, etc., basically everyone’s AI-app from early 2023)
2. Bake some image generation into a few low-stakes areas (chats, emojis, etc.)
3. Give you a HILARIOUSLY walled off integration with ChatGPT where it literally asks you if you want to send this stuff to OpenAI every single time
4. Use an on-device LLM to theoretically make Siri less stupid when it comes to listening 
5. Invest in a whole bunch of non-LLM stuff at the operating system level to allow apps to give Siri the necessary context to be less stupid when it comes to doing stuff

The first three are very real things that I don’t care about at all, and will probably never use except to troll my friends once or twice a year. The fourth one is a win unless it somehow makes things worse, but talking to Siri is already like talking to an old person who doesn’t speak English very well so that seems unlikely. 

The fifth is such a massive, fundamental effort that has so little to do with the kinds of innovations OpenAI, Anthropic, etc. have put into the world, it might as well be an entirely different company/product/industry. Throwing all this under the roof of “AI” is so meaningless and self-deluding I don’t even know where to begin. 

Anyways, the net impact on me as a consumer is… low. The obvious shareholder-appeasement cruft (ChatGPT integration, image generation) is pretty easy to ignore, and Apple’s been pretty good at not shoving stuff like this in my face as long as it’s not an opportunity to sell me one of their services. This isn’t one, and in fact, might even cost them cloud compute sometimes for no revenue, so I feel pretty good they’ll leave me alone. I am most certainly not upgrading any of the (alarmingly numerous) Apple hardware products in my home for any of this, though. I’ll do that at the same rate I always do, because generative nonsense aside, these products continue to marginally improve in generally positive ways. 

Will this make other people upgrade more quickly? I kinda doubt it. Is that the intent? Honestly I kind of doubt that, too — Apple might be just another stupid, next-quarter-minded 2024 corporation like everyone else, but at the same time… they aren’t. They’re the king of the mountain, they print money, and their executives are long-term Apple people shareholders have never been very good at dethroning. So I think they threw those guys a bone (AI! ChatGPT!) in the most non-destructive way possible, and then went on to promote a vision of “intelligence” that is much more inline with what they want to do (the Siri as a real assistant model), but is probably still well outside of their ability to deliver. 

This is a pretty incredible look at (among other things) how their best player being such a massive defensive liability is just killing the Dallas Mavericks, whether casual fans are able to see it or not.

I’ve absolutely played on recreational teams with this problem, where our best scorer and most obviously talented guy was pretty clearly the reason we lost. And I’ve seen in it non-basketball scenarios as well, like incredible sales closers who put up big numbers but suck the life out the rest of the teams, hurt our brand, etc. 

The real question for Luka is… is he willing to be different? Can he make himself valuable without the ball in his hands, or because he’s less talented in those areas, and excellence is harder to achieve as a result, is that just off the table and everyone else needs to somehow mitigate it?

I’ve worked almost entirely in B2B (or B2B2C) software, so I’ve never had the now borderline cliche experience of working on a free application that starts off being exactly what people want, but is unprofitable, and then needs to be made incrementally worse — forever — first just to break even, and then eventually to grow profits ad infinitum.

At least when you just increase your prices, you can try to provide some additional value (as long as it costs less to provide than the size of the price increase) and wave your hands a lot and hope people accept the tradeoff. But with this, Instagram is just… worse. It’s not even debatable — the business makes tons and tons of money doing what it does, but if the product was worse it could make more. So they made it worse.

Writing user communications for this kind of thing seems like it would be absolutely soul-crushing. 

I mentioned this when I first saw it, but I just cannot get over this ridiculous interview with the CEO of Zoom. Somehow, it manages to simultaneously highlight pretty much all of the major things that are bothering me as I careen into my grumpiest, middle-ag-iest years. It features an incredibly successful (and presumably rich) entrepreneur who sounds completely insane, and utterly high on his own supply. It’s about a company that is extraordinarily pervasive in our professional life — Zoom, dammit! — and yet seems obviously, visibly disinterested in the actual thing that causes us to use them. And finally, it’s got my personal favorite: an ambitious, potentially impossible value proposition for the future that makes absolutely no sense and logically collapses with even the most cursory investigation.

I’ve been working at startups for a long time, and many, many times me and/or my colleagues have made the “giant bong rip” hand motion as we’ve listened to someone visionary explain what they’d like to be doing instead of what we actually do. This is not new to me. But at the same time, it’s also why I’m so hard on my career — I often convince myself that the companies I’ve worked for are less sophisticated or impressive than you might think because they’ve said completely insane things like this, and presumably real companies with real interest in what they’re doing and a firm grasp of what it is would not say those things.

But… here we are. Zoom Communications has a market cap of eighteen and a half billion dollars, and the guy running it unveiled one of the craziest, impractical, and most nonsensical product visions I’ve ever heard to a leading publication without even really being asked about it. He just… said it out loud for no real reason. And it’s fine! The stock is fine. Presumably Zoom still works, maintained and operated by competent subordinates of this man with totally crazy, horrible, unfeasible ideas about what meetings should be like and what would happen if a bunch of large language models added themselves to them and babbled uncontrollably at each other.

Still, that’s not what bothers me the most, or at least what bothers me in a fundamental way that reminds me of so many other things that bother me. For that, you have to go to this section.

NILAY PATEL, The Verge: When you say the hallucination problem will be solved, I am thinking about this literally in terms of Moore’s law because it’s the closest parallel that I can think of to how other CEOs talk to me about AI. I don’t think you spend a lot of time thinking about transistor density on chips. I’m guessing you don’t just assume that Intel and Nvidia and TSMC and all the rest will figure out how to increase transistor density on chips, and Moore’s law will come along, and the chips will be more powerful and you can build more applications. Just correct me if I’m wrong. I’m guessing you—

ERIC YUAN, Zoom: No, you’re so right. Absolutely. This is a technology stack. You have to count on so many others.

NP: So is the AI model hallucination problem down there in the stack, or are you investing in making sure that the rate of hallucinations goes down?

EY: I think solving the AI hallucination problem — I think that’ll be fixed.

NP: But I guess my question is by who? Is it by you, or is it somewhere down the stack?

EY: It’s someone down the stack. 

NP: Okay.

EY: I think either from the chip level or from the LLM itself. However, we are more at the application level: how to make sure to level the AI to improve the application experience, create some innovative feature set, and also, at the same time, how to make sure to support customized personalized LLM as well. That is our work. I mean… pardon my French, but… holy shit. This guy is talking about re-centering a successful, eighteen billion dollar video conferencing company entirely around a technology that (a) doesn’t work reliably, (b) may never work reliably, and (c) is completely outside of his control. It’s like Ford announcing that all their cars are going to run on peanut oil, and just saying “eh, I’m pretty sure Exxon can figure it out. Or maybe Penzoil, I dunno."

LLMs having no understanding of what they’re spitting out, but using pattern recognition to make it kinda-sorta hard to figure out exactly when they are wrong — that’s the whole game right now. That is everything when it comes to making this technology, and all the things people are extrapolating from it, worth literally anything at all. Yuan is excited by the “AI” summaries of meetings. Great! Neat! That is an extremely niche feature with, potentially, some real value, which is why he’s not focusing on it but instead using it as an example of how close he is to allowing people to building virtual, digital doubles who speak, argue, and make decisions on your behalf. The second thing is a much, much bigger thing — a world changing thing, honestly — but it should also go without saying that it is infinitely more complicated and difficult to execute since a computer connected to actual, operational actions simply cannot be wrong the way a computer spitting out text for some bored office drone to skim can be. It’s questionable whether present, best-in-class levels of LLM accuracy are acceptable for even the corporate-virtue-signaling email writing tasks they’re starting to be pointed at now. It is not questionable at all that those levels of accuracy, or even significantly improved ones, are totally unacceptable for authorizing a purchase, or even scheduling a follow-up meeting.

I’m sure it’s a lot of fun to envision interesting product scenarios that sit on top of a magical, accurate, Jarvis-from-Iron-Man-like artificial intelligence apparatus with a reasonably priced API sticking out the back. But it’s not “product management”, or business strategy, or really anything more sophisticated than a couple of stoned college guys try to process the ending of “Her” in 2013. The hard part is building Jarvis, not fantasizing about all the cool things Jarvis could do if he existed. We had all kinds of fun concept fantasies and Dick Tracy watch scenarios floating around in our heads going back to the 1970s (or earlier!), but none of them really meant anything until Apple built us an iPhone to run those applications. And honestly, this feels even more egregious, like if somebody saw a rotary phone and started talking about his idea for Pokemon Go.

The Hologram Isn’t The Hard Part, or the Important Part

The largely forgettable movie “I Robot” was randomly on cable the other night, and I watched the beginning for a while, occasionally looking over at my vacuum and giggling. There’s a scene where the dead guy Will Smith is investigating has left behind this sort of interactive hologram, and Will Smith asks it all these questions. He gets some answers, but every once in a while it says something to the effect of “I’m sorry, I can only answer questions that I’ve been programmed to answer”. Which… makes total sense! This thing is important, and was left behind to help a police officer solve a murder, for God’s sake. I love imagining an LLM-powered version of this, answering correctly 92% of the time, but never admitting it didn’t know an answer and just making plausible sounding stuff up the other 8% of the time. “Sure, I was murdered. It was a band of criminals, seeking my money. Robbery gone wrong, you know? How else can I help you?"

This is all a long-winded, oft-made argument that whatever we decide to call AI — giant if-then statements attached to holograms, probability models, whatever — is not going to fulfill the dreams of the Silicon Valley elite unless either (a) it’s basically right all the time, or (b) it has an excellent grasp of whether it’s right about something and it’s at least right about that basically all the time. Again, this is the ball game right now. If you want to launder stock photos and generate spam sales outreach, maybe do some code suggestion, you’re good. Grab your API key and see what the market will bear for your probability driven idea. But if you think we’re going to automate like… actual people? Who make decisions? Who are held accountable for those decisions? First of all, you gotta figure out if LLMs can even do this, and it’s not looking great right now. And then, if not, you’ve got to invent something else.

Nope! That’s “someone down the stack”. Maybe it’s the chip! Maybe the chip will make large language models turn into… something else? Something better, and also fundamentally different. Chips do that all the time, right? Or maybe we’ll get a transparent, visible decision making engine from an LLM provider, even though no language model works like that because they literally add randomness to make interactions feel more interesting and human-like at the expense of predictability and exactitude.

Look, I’ve clearly had it with this LLM nonsense, and that’s in no small part because it stunk to high heaven from day one, if I’m being honest. And it might feel like I’m an ideologue on this, but I’m really not —  I’ve done the reasonable thing and admitted where the utility may lie. I’ve even built test applications with this stuff. I have an OpenAI developer account, for crying out loud.

But this is some of the worst, most red-flag waving magical thinking I’ve seen from tech in my entire life, and we just sat through crypto! Apple is frantically trying to get away from having to buy freaking cellular modems from Qualcomm because that apparently limits their ability to implement their medium-term vision for iPhones, a product that exists and is real and has literally changed the world. Meanwhile, Zoom and the rest of these guys are just standing around, taking their huge, imaginary bong rips (or real ones, I don’t know, I’m not there) and waiting — just waiting! — for some band of machine learning elves to make it all possible. And when they do, Zoom will be there, leveraging this fantasy technology for the dumbest, most impractical purposes you can think of.

Disclosure: I have a pretty broad 401k managed by someone else so, God help me, I probably own some Zoom stock. 

When assessing the “dirty” aspect of a basketball play, a good but non-all-encompassing question to ask is “if I did this in a pickup game, would someone fight me”. The play in question one hundred percent qualifies, not because it’s so physically egregious (it’s not), but because it’s such an obvious cheap shot (the ball isn’t even in bounds!) that it’s clear the aggressor wanted everyone to know it was a cheap shot. 

This is not “I am a tougher basketball player than you”, this is “I can do whatever I want to you”. And look, I get that a lot of people are horrified by the Matt Barnes-es of the world imploring for some extra-legal, on-court revenge justice. I’m usually not into that stuff either, to be honest. But that’s only when the powers that be are making a real effort to regulate what’s happening to people on the court. If somebody goes after one of my best players, after the play, the league needs to throw him or her out of the game, period. Otherwise, the 12th man is coming out and putting the aggressor into the basket stanchion on the next play, and the league can sort all that out instead.

I sure hope so. I am really sick of logging into bloated web applications. While we’re at it, let’s see if this can handle photo uploads. 

Godspeed, bits.

I don't know a polite way to say it, but the CEO of Zoom sounds like he's completely lost his mind. You should read this whole Verge interview to experience his bizarre, dystopian, AI-double fantasy for yourself, but also to witness the perfect example of zero-accountability AI-fueled investment hype, pictured here:

"It's someone down the stack." That's the most revealing quote I've heard in a while about today's AI hype.

Quitting social media over the last five years has been a net positive for me, but I do miss having a place for quick observations and small rants. Turns out it's not the audience that does it for me, it's just the therapeutic exercise of documentation. Crazy, huh?

Anyways, I'm adding a feed here for short things like that, and we'll see how it goes. This also gives me an excuse to try hooking something up to ActivityPub.

When I'm not pretending to be a tech visionary, I work in product marketing. This means that everyone in my personal life (most of whom do not in work in marketing for software startups) thinks my job is basically to come up with cool advertisements, and most of those people have suggestions. Meanwhile, every job listing I'm qualified for is essentially an elaborate demand to prove with 100% certainty that hiring me will make everyone want to buy their product.

This divide has dominated my career, mostly because I came of professional age in the "let the data decide" era. If you squint really hard, every once in a while it feels we just might be able to track everything important, put it all in a big pivot table, and then just... do what it says. But most of the time, things haven't worked that way at all. Instead, non-experts with massive confirmation bias now just have access to an infinite number of allegedly unbiased justifications for whatever it is they wanted to do.

In the words of my AI-generated co-worker Brian:

Anyways, this is already veering into more of an anti-data rant than I want it to, mostly because I've gone on my own versions of those before and the full posts have the necessary disclaimers and squishy "not-all-data!" qualifiers, so just read those if you're looking for that.

Instead, today I'm here to talk about limitations.

The Limitations of Art

I don't know if any amount of squinting is sufficient to view me as an "artist", but go ahead and try. It'll probably go better than trying to view me as anything else. Still, I think a lot of people assume my data skepticism is rooted in my own love of, for lack of a less soulless term, "qualitative creativity". I like to write, I like narratives that make people excited and my kids laugh and my wife remind me that there are other people in the restaurant. I'm guilty of all of that, and it's not exactly a major leap to assume I'm always going to pick a good story over whatever the data says.

But you'd be wrong. I've heard a lot of big, exciting narratives in my time about what was going on with the business and what we needed to do. Eventually, though, I also had Salesforce. Was our Sales data bad and wildly deceptive? Yes, it was. But that just made it take a little more work to figure out what was really going on, and once I did, my interest in exciting stories that didn't align with what the system was logging dropped to approximately zero. Because those stories were bullshit.

That's the downside of relying entirely on the art side of things. Everybody can have an idea, and no matter how stupid or delusional it is, it's probably going to seem great to someone. Even worse, if presented properly and cynically enough, that idea can prey on people's emotions and insecurities and drive you in the absolute wrong direction until you fly off a cliff, hand-in-hand like two fugitives in a convertible fleeing local law enforcement.

My Dad's an electrical engineer who was drafted into the executive ranks when I was in high school, so he rarely had to settle people down with business data. He was able to lean on physics, and explain to his fellow leaders that something was an incorrect assumption because it wasn't how, you know, thermodynamics worked or whatever.

Salesforce data is not thermodynamics. In fact, if thermodynamics was as reliable as your GA4 data, there'd be no way to determine how long it would take to make toast. But if you take proper precautions and analyze things very critically, they can be powerful tools for tethering yourself to reality.

And obviously, science doesn't have to come in the form of numbers. A customer can say they wanted a feature, and that's obviously a more concrete observation than just a feeling that a customer would want it. Is it better to pull out more than a single random anecdote? Yes, it is. Do people who go too far down the art road with their narrative often fail to even do that? Yes, they do, because often that's how far away they are from reality.

In other words, the worst case scenario for leaning solely on the art of marketing is... bullshit. And while non-marketers and casual observers of marketing might think "hey, that's what you guys do!", those of us trying to grow businesses know that drifting into bullshit is usually the beginning of the end.

The Limitations of Science

Weirdly enough, my criticism of people's data usage is often that they're using data incompetently, in quasi-bad-faith, or even outright unethically to advance... well, bullshit. And that is one huge challenge that comes with mass access to data. It's actually why I'm so worried about mass access to generative AI at work -- it's not the "replacing jobs" part I'm worried about, it's the idea of a bunch of people cranking out spam with the lame lack of discipline or literacy they showed for twenty years when given easy access to things like user and buyer data.

But, to be fair, "you can do it wrong" is not really a unique limitation of data or science -- it's just easier to dismiss bad art than bad science. Most people feel comfortable dismissing art that they think is worthless, but the language of data and science carries with it an authority you have to overcome with your own knowledge of the material. Hipsters and scenesters who didn't think my band was punk rock enough aside, it's easier to bully with science than art.

The bigger, more impactful problem with data as your source of inspiration is that it rarely leads to truly new, outside-the-box ideas. Maybe you don't collect the relevant data, which makes sense because you don't actually do it yet. Maybe there's a benefit or impact that isn't consistently shown in the data, or it's only shown in broad measures like customer satisfaction that don't confirm specific assumptions about your product. Either way, there's a fundamental limitation on how creative you can get if you're 100% tethered to basing your ideas on observable things that already exist.

I've certainly triggered the art part of my brain with some realization provided to me by data, but I've never had a transformative, exciting idea that was simply telegraphed by the data in front of me. And while that might not seem like an existential blocker to innovation, it's a lot harder than you'd think to get a 2020s-era organization to act on a great idea that was only inspired by data. There will always be a less radical, less artistic idea with more support from data, and in my experience, that's the idea most companies will be willing to commit to.

In fact, this process has happened so frequently, and so consistently in my professional journey that I'm honestly kind of shocked when the opposite takes place, as it has with virtual/augmented reality platforms and generative AI. While you can now dig up some data to justify basically anything at this point (and I assume it's been done with these ad nauseam in various meetings at the companies pushing these concepts), it seems pretty clear that this is coming from somewhere else. The only concrete data points we seem to have about generative AI, for instance, is that ChatGPT got a ton of usage right out of the gate, and that businesses are convinced they want to use AI as long as you don't actually define what its capabilities, costs, or limitations are. Cool, let's ship it, I guess. In some ways, "the computer is a person!" is one of those data-inspired, art-led approaches to product innovation I mentioned earlier. If it's right, you end up with what many in the media and investment world seem to expect -- a revolutionary new platform that, when we look back, will seem too obvious to have missed but required a fundamental, visionary leap away from the past.

Or... you get the opposite. A derivative, copy-cat idea that's then poorly applied to nonsensical use cases by a bunch of people with their fingers in their ears who watched too many of the same movies growing up. Either way, the art vs. science debate won't get resolved, because either approach could plausibly take a good chunk of the credit (or blame, depending on how this plays out). If everyone migrates to the metaverse (or we hilariously and self-servingly redefine the term to encompass something different that people actually want), is that because of a qualitative adherence to Mark Zuckerberg's vision against all indicators? Or will some team of analysts tell us the whole thing was obvious based on a rigorous review of such and such indicators? If we end up in my personal hell arguing with our computers through command lines, does that mean the tech bros obsessed with building the computer from Star Trek saw the future, or was this an obvious evolution from engagement with Siri & Alexa over the years?

I have my takes. I don't think they matter, though, because the entire thing will look different depending on who's looking at it. And my friends will still think I make ads for a living.