The vibes are weird

The fraying physics of Silicon Valley. Plus, a false fake from Drake?

This is not the ocean we’re used to.

In 2012, it was a party. When I moved to Silicon Valley that year, it was no longer a place for nerdy rebels or idealistic misfits, but a place to get rich without feeling guilty about it. After the financial crisis, a generation of country club brats had started questioning if working at McKinsey really was the most prestigious thing that they could do after graduating from Duke, and tech offered a promising alternative. The Nasdaq was on a steady march upwards from its lows in 2009; huge tech IPOs were back; huge tech acquisitions were back; Google paid as well as Morgan Stanley, gave you unlimited time off, and didn’t make you work until 8 p.m. if you wanted to eat dinner for free. Back then, you could have it all: Money and status and time off, without being evil, without tucking your shirt in, and without having to leave the office for a couple (dozen) beers. 

Of course, the whole thing had an ugly underbelly—but the perpetrators of that era’s corporate culture weren't worried about that yet. The prevailing sentiment in most tech offices was optimism.¹ And not for some promising future that we were building, but for the present. It didn’t matter that the old cliché about tech making the world a better place was already a joke, because for much of Silicon Valley, the world was a pretty good place as it was. 

In 2014, the frat house got raided. The problems in tech’s basement became headlines in the New York Times. Three-character ASCII art and “locker room talk” stopped getting laughs at hack days, and started getting people fired.² The working environment inside Github—which, back then, was the pantheon of tech culture; Mecca; it was every startup’s favorite startup—got its CEO fired. Susan Fowler had a very, very strange year at Uber. It was time to grow up.

Moreover, there weren’t just issues with how tech built what it built; what we were building was starting to look problematic too. Taylor Swift went to war against Apple. Amazon no longer felt like the ethical alternative to Walmart. Trump won, and people began to wonder if Facebook had something to do with it. 

Still, the core promise held: It was prestigious work that paid well. In some ways, the minor reckoning was a good thing for the brand—it was sign of progress, a needed policing of a few bad apples. People began to talk more openly about tech's problems, but few people seemed to entertain working anywhere else. 

Then, the party was back. A paper by some anonymous libertarian got popular. We were promised digital gold, a self-enforcing legal system, freedom from the tyranny of Janet Yellen, a better way to buy drugs, a way to get rich quick. Nobody quite knew what any of it was for, but a16z liked the idea. A few people got rich; most people got scammed; the two were related. 

In 2020, a pandemic, and a flash crash—the Nasdaq fell by 31 percent in 33 days, a handful of startups evaporated, and the market was back to setting all-time highs less than 80 days later. Everyone started working out of their bedrooms, got cabin fever, and lost their minds. Venture capitalists unloaded cannons of money as if there was no tomorrow. You could raise a seed round with a pitch deck and a pulse; every series A was "20 on 100;"³ every series B was done by John Curtius.⁴ “It’s one ETL tool, Michael, what could it cost? 1.5 billion dollars?” We were up to our elbows in punch. 

The bowl got yanked away again; a bank blew up; venture capitalists lectured everyone about spending as if there was no tomorrow. It was time it sharpen our pencils, and tighten our belts, and trim our fats. 

But through it all—and even then, in 2022, in the midst of another barrage of layoffs and fire sales—nothing felt fatal. We’d all seen this before; there were no doubts that the sun would rise again tomorrow. We’d get ‘em next year; the boom times would be back; disappointment did not give way to despair. The arena was still a good place to be, and in time, we were destined to be its champions.⁵

—

In Silicon Valley, there are two sets of physics. The first is the cyclical patterns of the ocean. There are waves: Some technology will be in vogue; others will be out of fashion. Set sail at the right moment and you can find a crest. Plan your voyage wrong, and you get stuck in a storm.

Work here for a while, though, and you get a feel for the seasons. You can see the waves coming; you figure out how to build a boat that can endure rough waters; you can learn how to read the sky and forecast the weather. Careers in tech aren’t linear, like the slow climb up the corporate ladder at a law firm might be, but they’ve still predictable—and often, predictably good. Keep learning how to sail on this ocean, and you’ll eventually find your wave.

Throughout at least the last 12 years, the cycles in Silicon Valley have had this familiar rhythm. There were upbeats and downbeats, but everything stayed in 4/4 time. The debauchery of 2012 had echoes of Liar’s Poker, of my cousins’ stories of Silicon Valley in 1999, and of my classmates’ stories about investment banking in 2006. The first crypto boom was weird, but an eccentric appendage of “core tech,” more financial engineering than software engineering. The mainstream stuff, like SaaS and consumer tech, was still an appealing bet for ambitious entrepreneurs. Throughout the pandemic of 2020 and the profligacy of 2021, the waves got bigger, but they retained their essential physics. And during the bad times of 2022, Silicon Valley’s annoited sages played the usual hits: Hunker down, buckle up, build on, and trust the process.

It was trite, but seemed basically correct? Even if everything collapsed, the great shroud of the sea would roll on as it rolled five thousand startups ago. We would build another boat and sail on, a little bit wiser about how to navigate whatever storms and swells might be coming next. 

But now, man, I don't know.

—

Because in the last year, the second set of physics in Silicon Valley—the foundational laws of the ocean, of the atmosphere, of what floats and what sinks—seem to be breaking down. 

We have become obsessed with AI. Fifty percent of the startups in Y Combinator are building around it. Eighty-five percent of tech companies in the S&P 500 are talking about it. Companies are rebranding themselves around it. Snowflake is no longer a database, or The Data Cloud, but the AI Data Cloud™. Asana is the “#1 AI work management platform.”⁶ Intercom is AI-first. No, Zendesk is AI-first.⁷ We are impaneling conferences about AI; raising venture funds;⁸ starting newsletters; hosting emergency podcasts. This is not an iPhone moment or the next internet, because those things had naysayers. Now, the only debate seems to be about which side of the AI revolution human beings will be on. 

It’s the crypto gold rush all over again, except the winners are paid in legal tender. It’s the crypto gold rush, except the losers won’t just miss out on some money; they’ll get steamrolled into obsolescence. And so we’re all frantically building new features—60 percent of one VC’s portfolio is already using generative AI in their product. We’re launching new companies. We’re investing in new startups. We’re turning tech on a dime, all trying to skate where the puck is going. 

On the surface, everyone is confident. What they’re selling, you’ll be buying. In an AI-powered future, data quality will be even more critical. Your CRM will be even more critical. Your e-signature management platform will be even more critical. Your useless blog subscriptions will be even more critical. The future is oh so exciting, we say in public. 

But in private, the vibe is different. Beneath the showy confidence, there’s a shaky unease. Nobody knows where the puck is going. The puck is teleporting. The puck is from another dimension—one in which computers speak English and can’t do math. The puck might get obliterated by GPT-5, and our current standing in the social order might get obliterated with it. We launched the AI features, but nobody uses them; are we sure we’re made the right bets? Are we sailing on the same ocean, the ocean we’ve spent our careers exploring? Will the the new ocean be better for someone else? Even if we don’t get outright replaced—which I don’t think is behind the creeping anxiety—are we sure the skills we’ve learned and the boats we’ve built for the old tides will still be seaworthy? Are we sure our new startup is future-proof? Apple events were fun, a chance to see some cool new gadgets, and to think about ways we could use them. OpenAI events are the reaping. We act excited, but deep down, we’re just hoping to survive. 

Is Silicon Valley back, or is Silicon Valley coming unglued? The old foundations are buckling. It’s the end of software—though notes like these don’t feel like just predictions, but self-affirmations, that if we act like we know what’s coming, we can manifest our own continued relevance. Do we build a team to build a business, or just build a team? The best acquisitions are now acquihires—650 million dollars for the leadership team of Inflection AI; 1.3 billion dollars for a bunch of GPUs and a team of AI researchers; 150 million dollars for a new CEO; 1 to 2 billion dollars for the chair of an open-source project. 

In a founder email list I’m on, more people than ever seem to be eschewing venture capital because they want to build a “real” business, or buy a bunch of cleaning companies and become a micro PE firm. Is happiness in Silicon Valley is no longer a traditional startup, but a bootstrapped business and taking fistfuls on longevity pills? Venture capitalists are also preaching fiscal responsibility. It’s the end of easy money; the age of austerity, the year of efficiency, a time to play to make plan—but also, some French AI startup you’ve never heard of just raised 220 million dollars in a seed round. There are no reliable physics anymore. 

Get rid of the classic product management, says one founder worth 10 billion dollars. Get rid of management altogether, says another founder worth ten times that. Elon Musk is seceding to Austin. Founders Fund is seceding to Miami. Politics is banned; politics has Sand Hill Road’s usually chummy cocktail class at each other’s throats. The 2024 election hasn’t even started yet. 

Is it the end of something? Is some era dead? Are we getting getting eaten? Is it time to build? I don’t know. The ocean is changing too quickly; the weather is too unfamiliar. The moment betrays buzzy headlines and showy predictions. Whatever gets made now will likely be upended two or three times before the sea settles down. The best reason to build a boat today may not be to make something that endures, but simply to be out on the ocean, for whenever it finds its rhythm again. Because the water is convulsing in weird ways, and you don’t fight the sea.

---

Drake fakes deep fake?

If you’re a famous rapper and you’re in the middle of a very public rap battle with another famous rapper, you have two choices:

1. You can drop a new diss song. 

2. You can not drop a new diss song. 

There are your only two choices, and neither are ideal. If you don’t put out a song, people—including the person you’re battling—might call you a coward. This is bad in a rap battle. You cannot win a rap battle without battling in it.⁹ 

But putting out a song is risky. If people like the song, they’ll make memes about it and you will score a lot of points. This will be good for you and bad for your opponent. If people don’t like the song, however, people will also make memes about it—but they will be bad for you and good for your opponent. Hence the choice, about what to do, if you are in a rap battle: Don’t drop a song and take a guaranteed small loss, or do drop and wait for the internet to decide if you put out a banger or a bomb. 

Ideally, you wouldn’t have to make this choice. Your best option is option three: Record the song and figure out if people like it before you drop it. Unfortunately, as a famous rapper, you can’t do this. Nothing would lose a rap battle faster than people finding out that you focus grouped your diss song. And you can’t secretly release the song either—everyone knows what your voice sounds like. 

Anyway, last week, Drake, who was last seen getting humiliated by Kendrick Lamar in a rap battle, dropped a new song on Soundcloud. Or maybe he didn’t?

Late on Monday, the internet exploded with a version of the Plain White T’s moody 2005 hit sung in Caribbean patois in what sounds like Drake‘s voice (but these days, who knows?). The cover, titled “Wah Gwan Delilah” and filled with Toronto references and slang, was uploaded by SoundCloud user Snowd4y and claims to feature Drake.

The track has more than 200,000 plays on SoundCloud and immediately went viral on Twitter. But is it real?

Previously, if you heard a song with Drake’s voice, you wouldn’t wonder if it was Drake or not. Now, with AI, we don’t know anymore. It could be Drake; it could be a random woman in a basement with a Speechify subscription and a Soundcloud account. It could be The Weeknd; it could be ghostwriter997. It could be Eminem; it could be David Guetta. We used to be able to know if a song was real by listening to it; now, the only way to know it’s real is if person in it tells us that it is.

But they don’t have to tell the truth. Drake can make a song, release it under some other account, promote it without saying if it’s his, and then decide if he wants to tell people he made it. The flops can be fun fanfic. The hits can be his. But they’re actually all his.¹⁰ 

When people talk about the dangers of AI, they often talk about deep fakes, which are AI-generated images and videos that appear real. As bad as these are, they’re only half the problem. We’ll also have to deal with false fakes. People can now—very plausibly!—claim that something that’s real was made by AI. That dud? Not mine. That embarrassing photo? Adobe Generative Fill and a smear campaign. That hot mic? I never said that. 

Though “my account got hacked” was always a suspect excuse for doing something dumb online, it was useful because it could plant a seed of unresolvable doubt. Unless someone admitted that they were lying, you could never prove the negative. “That’s AI-generated” could become the same generic defense—never probable, but now, always possible. 

1. The optimism was less widespread in women's bathrooms, whisper networks, and crowdsourced Google docs.

2. Three months before this clown show, roughly the same thing happened at a hack day of a company where I worked. Some guy made a bunch of suggestive jokes that eventually culminated in him unzipping his pants and pulling a rolled up piece of paper out of them; he was fired before the next presentation was over. Which, on one hand, yes, obviously. On the other hand, had he done that a year ago, he probably would’ve gotten a slap on the wrist in public and high fives in private. 

3. As in, raise 20 million dollars at a 100 million dollar valuation.

4. I mean, no, obviously not; this is a joke. But it’s not that much of a joke! Curtius did 250 deals! In five years at Tiger! That’s one a week! He was invested in half—half!—of Forbes’ top 100 cloud companies. That is wild!

5. To be clear, the arena wasn’t a good place for everyone, and lots of people have been badly mistreated in it. But the arena was an overwhelmingly desirable place to be. That’s one of the reasons why tech’s toxic elements are so pernicious: They exclude people from what could otherwise be a very rewarding career. 

6. What on earth does that even mean.

7. Which one is first??! Which Joe’s is the original one????

8. These funds invest in AI, not CSS.

9. Or maybe you can?

10. Ditto for Drake’s kids.