In line to be over

Because the statement that “sf is so back” is given to us in quotation marks, we are being told that it’s not inference.ai who is saying that “sf is so back.”

But it clearly is inference.ai who is saying that “sf is so back.” After all, they bought the billboard. They put the words on it for you to see, as you line up above the San Francisco skyline on the freeway that takes you into the city. They are a tech company, and they are announcing that while San Francisco has, recently, been “so over,” the city is now—in some way that we can inferentially associate with inference.ai themselves—“so back.” They said these words, and they said the quotation marks too. 

They also “said”—though this is a little trickier, but stay with me—everything that isn’t on the billboard. For example, the absent speaker: they could have quoted someone saying those words. They could have, for example, quoted Alexander Stratmoen (“founder of a stealth startup visiting from Canada”) who uttered these words on October 29^th, in line for a TechCrunch Disrupt after-party, and was immortalized for doing so in the pages of the SF Standard. 

On the billboard, however, no speaker is given. These are words that float in the collective consciousness, the social ferment, spoken by everyone because no individual could ever say them sincerely. Try it! Try to say it and you will feel the meaning drain out of it the words as you speak them, like when you try to google “inference.ai” and you find yourself redirected, instead, directly to their strangely ominous red website. They can’t mean anything specific, these words, yet they are a reference everyone recognizes—or that, if you recognize it, you become a part of that specific “everyone” it implies—precisely because they don’t have anything so crude as an origin, or original. Putting actual quotation marks on them is superfluous.

More than superfluous, in fact, putting quotes on this phrase is like adding another negative number to a negative number and making it positive. Because this sentence cannot be spoken as if it were a thought in your brain that you, a discrete individual, had created, and that you are now uttering and communicating to others—with the understanding that until they hear you say these words, they will never have considered the possibility—the act of putting quotation marks on it actually demonstrates a claim to ownership by inference.ai. Before, these words were like a proverb, rhetorical structure lacking substance, a kind of memetic clay ready to be adapted to any context, like the words “inference” and “ai.” But after inference.ai put these words on their billboard, the so being back of sf was, suddenly, being spoken, by someone, by anyone, in an inferential reference—but specifically—to inference.ai.

Let us go deeper. I do not know what inference.ai is, the way a bird doesn’t know what a cat is. But birds know how to behave around a cat, and the fact that they have an I-80 billboard, that they talk this way, that they present themselves this way, tells me how to behave. Not knowing things is sometimes, often, much less important than being able to infer from a context how to blend into it. An observer, sometimes, won’t be able to tell the difference.

At the same time, might not knowing what cats are hinder a bird’s ability to not be eaten? And so I asked a friend to ask ChatGPT what inference.ai is, and whether it’s the reason that sf is so back. Initially, ChatGPT didn’t know what sf was, but after showing it the billboard, and being told that the billboard was in San Francisco, it eventually reported—and this was dutifully relayed to him and then to me—that “The phrase “SF is so back” likely refers to the resurgence or revitalization of San Francisco, often abbreviated as “SF.” The mention of inference.ai suggests that the company is promoting itself in conjunction with this sentiment, possibly tying its technology to the progress or innovation associated with the city’s tech culture. Inference.ai is a platform focusing on AI-related solutions like GPU virtualization. By placing this message on a billboard, the company might be aligning its brand with the renewed energy or optimism surrounding San Francisco’s recovery in areas like technology, culture, or economy.”

That’s not a very good explanation. Here’s a better one.

Inference.ai is in an interesting position in the industry. OpenAI and, collectively, “Big AI,” took on massive legal exposure (and moral sin) for first-mover advantage in what they think and agree is a gold rush. What they think gives them a durable advantage in the market is that they are spending vast amounts of computing effort on training; it’s therefore in their interest to scare everyone else off by bragging about how hard it is, to signal to potential competitors that they shouldn’t even try. But Inference.ai doesn’t do training. It only does inference, applying things that are already trained. 

To understand the distinction, recall that what the thing we are calling “AI” is, actually, in a material sense, is a statistical model. Training is when you show that model a lot of real-world data, typically stolen, with an appropriate framework to make it pick up on patterns, like fitting a regression line to a set of points. Training tends to be very expensive up front and involves a lot of interesting and consequential choices. 

Inference is when you apply the model to data it hasn’t seen before, to see what it does with it. It’s the actual use of the model in the world. And so, Inference.ai is in this interesting position where they’re tee-ed up to make a lot of money, theoretically, off of things that Big AI has already spent a lot of money on. They’re setting up to eat what others have planted, with less legal exposure (and moral sin). As a result, they’re jockeying for name recognition and an image as friendly underdogs in the industry, because that helps them with hiring and if the Big AI hammer comes. If you see Inference mentioned in the tech news, they’re going to be lumped in as “an AI company” because the news tends to presume the thing-ness of AI. But if OpenAI is doing primitive accumulation, expropriating the commons and making it their own—if they are dripping from head to foot, from every pore, with blood and dirt—then a company like inference.ai is the nice legal friendly guys who come and do the nice legal capitalism afterwards, once all the blood and dirt is cleaned off. Of course, Big AI want to be doing the second stuff too. They want to be GE in 1970, even as Inference.ai tries to displace them, and make them merely US Steel in 1970. 

This is why inference.ai is both saying that SF is so back and disavowing it. There is an AI gold rush happening, and they want to get rich, but they don’t want to be associated with the people who murdered all the indigenous people to get the gold (metaphorically speaking). They want to ride the wave. They want to be carried on the froth, the foam, and be washed clean by the surf.

Now I’ve said these words, but I didn’t really say them. I couldn’t. I don’t know these things, or at least I didn’t at first. But I was able to say them, and now I know them, because I asked a friend—a bird who knows a lot about cats—and he told me these things, and because I know how words and sentences work, I was able to consolidate what he told me into this very plausible-seeming summary of the situation. You probably read it and thought to yourself, huh, this person probably knows what he’s talking about. But I don’t, except to the extent that I trust what I’ve been told. And I don’t know how he knows these things. Which is fine. If I knew all the things that he knows, there wouldn’t be much point in talking to him. 

We could stop there. This might be enough. But when I found a podcast interview with the founder of inference.ai, and started to listen to it, and was almost overwhelmed with an almost lethal disinterest and apathy, it clarified for me what I wasn’t looking for here. I don’t want to go deeper, it turns out. What I really don’t care very much about is what inference.ai is doing, or their strategies, or their prospects. I don’t care to go deeper, to know what inference.ai is, what is behind that name, these words, or what meaning their speakers have, in their souls, when they say them. I don’t want to know what Alexander Stratmoen meant when he said that “sf is so back.”

I just want not to be eaten. And in that sense, what a cat is thinking, as it pounces, might not be the most useful information; maybe what we want to understand is the physics of how it moves through space. Maybe to understand this billboard, we have to go wider.

The phrase on that billboard, of course, is an echo of an already stale discourse, from a moment in time in which the phrases “It’s so over” and “We’re so back” not only emerged into the memetic fabric of our consciousnesses—everywhere and nowhere at once—but did so as a way to describe a distinctly cyclical state of perpetual reversion. It came out of a moment, we might say, when the pandemic was both over and always coming back, a permanent present state of affairs which had impressed upon us all the impermanence of everything (even of the pandemic itself). It was in this sense that being so over only reminded you that soon you’d be so back (just as being so back was a promise of a return to being so over once again). “The two phrases draw their power from each other and expect or imply one another,” as Max Read put it; “Within ever back is an over; within every over is a back.” The phrase articulates

“a particular post-pandemic vibe. We’re emerging from an unprecedented political-economic situation into an unclear future, dependent on monthly updates of case numbers, or inflation indices, or unemployment rates, hoping for signs of a stable forward trajectory. “It’s so over”/”we’re so back” embraces the spirit of volatility, overreaction, over-extrapolation, and over-commitment that characterizes incremental updates around unusual developments.”

What’s most interesting about this billboard, then, is not what is in it, but what is precisely not in it. To say “sf is so over” is not to say several crucial things, which are nevertheless “out there” in the linguistic fabric through which utterances travel to be understood, and like the missing rug that tied the room together, we “hear” them all the more distinctly because they have been removed. 

First, to say “sf is so over” is not to say the word “we,” as well as removing the interesting way that being back is a property of the collective, of our pluralized first person (in contrast to the alignment of over-ness with being an it, a thing). More importantly, as the dyadic nature of the meme has been removed—becoming one sentence instead of two—it’s become a sentence with a center, sf, what those who don’t know San Francisco as The City call it. It’s become the claim that San Francisco is back even while denying the inevitable return to overness (of which even the very word “back” cannot fail to remind us).

What is “sf,” this modestly un-capitalized version of San Francisco that has displaced the utopian collective? In a sense, it is The City: the center of the center, a financial district whose centrality is structural more than geographic and an imperial city in the sense Gary Brechin described it as being. For it to be back is to describe a return, again, to the kind of extractive centrality that had, as of recently, threatened to be over in San Francisco. Whether because of woke, because of the pandemic, because of crime—or simply because tech keeps laying people off—something was rotten, something was off. Now we’re back, the sign wants to say, or wants someone to say so that we can quote them, and follow them, and trust and believe what they’re saying. That sign wants to look at the last ten years and say: we’re back on the path that we thought we were on a decade ago, and weren’t. But now that we’re back, we’ll never be over again.

It is, in this sense, a very Trumpian “back.” This billboard is going up now, not in 2022 when the overness of the pandemic had first threatened to emerge—or in 2023, when it had become a parodic echo, mocking the hope that anything ever ends—but now, in 2025, just days away from the return to America’s greatness, again, a return it can’t help but rhyme with. Yet just as San Francisco “moderates” must pretend they aren’t Trumpian, the billboard can’t quite admit what it’s saying. It must disavow it, like the Trumpian locution through which other people are always saying the thing that he is saying, so that he doesn’t have to say it (“sf is back? Wow. I didn’t know that, you’re telling me now for the first time.”) But of course, he’s in on the joke. He plays the game, he spews the bullshit; he knows he’s more grin than cat, even if he also has teeth. He even tells you, because he also knows that knowing won’t keep from being eaten. Knowing is beside the point, because the billboard is not for you. It’s for the people in that line, in October, trying to get into a tech conference after-party. For people in that line, it tells them that if you have a ticket, then the experience is amazing.