I
For a long time now, everyone in SF has been talking about agency. This was the case even before AI came along and gave the term an extra layer of importance. Agency in SF means blue blood, grace, election. An agent is a thing which can self-reflect so much that nothing is taken as given. Only an agent can achieve the extraordinary, because only an agent can break the patterns of normality. In the church of agency, there are no rituals. Everything is new all the time. John Ashbery said that ‘God will find the pattern and break it’, and that’s pretty much what agency means: god mode.
Last month I travelled to SF. I went there because I am #building, and SF is where #builders go. My idea in going was to see whether I could take in pill-form a little dose of agency and get my project going hockey-sticks up to the sky. Yet to tell the truth, I was a bit conflicted. It was a funny mix of inferiority/superiority, when you can’t tell if you’re against something because you think you’re above it or because you think it’s too good for you. I didn’t want to play the stereotypical Silicon Valley games I had been warned about: jockeying to be on the most podcasts or in the know re OpenAI. Yet though I didn’t want to play the games, I still wanted to win them. Because that would be nice, wouldn’t it? To know that I could?
The plane took ten hours and the whole time I was fiddling with switches in my head, trying to find the one to make me a superAgent. When I landed I sent emails and texts and LinkedIn DMs to everyone who I could figure out lived there. I met all sorts of people. I drank all sorts of coffees. Each morning I hitched on a Lime scooter and pushed up the insane SF slopes to chinwag with coin-shufflers and VCs and try to follow the manic babble of ‘operators’. To all of them I spoke really fast, making sure that there was no doubt that there were a million other things I could have been doing right at that very moment when I was instead talking to them. I wasn’t there to do any work — that isn’t allowed on an ESTA — but it was just about meeting people and seeing how the system functioned.
(SF is quite amazing. In Europe you hear either great or awful things about San Francisco depending on what your circle looks like. In London I mostly hear the parts about terrible housing, addiction, grim street life, occasionally tempered by zealotry from startup guys who can’t believe that there is just free money everywhere and loads of GPUs and people think it’s cool that you’re a CEO. My experience was not related at all to either of these things: I found that it’s just a very nice place. I couldn’t believe how much I enjoyed walking around it, how pretty some of the areas were with their cartoon purple and yellow facades, how fun it was to crest a hill and see the ocean on one side, mountains on the other, fog rolling in and massive sun burning through. It’s as good to walk as the great European capitals.)
On one of the days of my trip I got invited to a big hacker campus. Hacker spaces are a symbol of the continuity that modern San Francisco has with its history as it has gone from counter culture to cyberculture. In the past, before rents were high all over the city, hacker houses tended to be communal living arrangements plus a shit tonne of wires. Variants of these still persist holding onto their utopian anarchistic vision. Noisebridge, for example, is a workshop run entirely on donations where anyone can just walk in and start playing with their hardware. This sort of thing is the exception to the rule these days, and more often you get these quasi-communal spaces which have all the same surface layer vibes of Noisebridge — messy, hacky, goofy — but which are actually funded by venture capital and function simply as another filter helping investors to select their portfolio. I tried to get into one, and backed out when I realised I would have to take an exam and give away some of my company to live in purple-UV room and eat dinner every night with 11 other dweebs.
The space I got invited to was one which is open and explicit about its nature as an accelerator. It was in a beautiful part of town, in sprawling buildings set back from the street. I scanned my phone to let myself in. What I found there was, in many ways, the opposite of agency. The place was packed with twenty-somethings. Humanoid robots stumbled around the corridors. There was a podcast studio. Hundreds of screens showed Claude Code going at it. “Everything is paid for!” said my friendly host. There was free accommodation for all the residents, as well as free meals. There was a gym, a games room, acres of workshops with soldering irons lying around still warm. Everything you could possibly need to build some weird technology and make a company out of it was there in that building.
You might be wondering why I found this all so notable. After all, London is not exactly the world’s most unfriendly place to innovation and technology. People do stuff here — people do make new things. But the point is that in London you simply do not operate within an environment like the one I saw in San Francisco. In London, when you’re doing something new and unlikely to work you have to recommit to it every day. The odds are you will proven wrong, and deep down you know that your best strategy would be to quit and do the sort of job that everyone else is doing. You have to remind yourself over and over again to be an agent, to do the things that the ideal version of yourself would be doing. This is very often a process you do alone, because your social life is not organised around other people doing the same.
What I saw in San Francisco was the total opposite. The campus I visited, the hacker houses, the evening events, Salesforce Park — it’s all setup such that for a certain class of highly-educated young person, the best option is just to make something new and start a company. That’s basically it. It’s possible to be molly-coddled into it. The system of social status is arranged in a highly idiosyncratic way to make all the founderlings feel great about themselves: being cool is doing a certain kind of post, being invited on a certain kind of podcast. Everyone is told they are special. No one is not being a founder. In SF people love to virtue signal about being ‘short status’, but the same people then chase the coolest VC uber-poster routes that mark you out as top dog. What’s more, the whole thing is set up so that there is literally no downside to founding: failure just means a better job afterwards. Ironically, the concept of agency never felt emptier than when I got to know how early-stage Silicon Valley actually functioned. I was jealous, and not a little awestruck.
II
For some reason the internet is illegal in San Francisco. At least, it was for me. My UK provider gave me terrible 5G, so I bought an eSim and ran T-mobile, and it was even slower. You fight for a table in a café and — once you get over the cringe — ask for the wifi code, except the wifi is like 0.5 Mb/s and you barely open your email before giving up. I’m guessing that if you’re a Big Tech (SLOGAN) engineer that you’re on some secret network managed via Discord. I was out and about and I wanted to chat with some people back home, and since I had given up on regular cafés decided to head into the Capital One co-working space which you can use for free as some kind of GTM strategy by the bank. In here there were all sorts. Founder types but also loads of people studying, many of them apparently high-school.
I found a corner and logged in. The guy sitting next to me had AirPods in and was on a call. He was tall and solid-looking, had a nice natural authority. His call was about some business, presumably his startup, and he was having a long rambling conversation. Every few minutes he would say, ‘don’t worry bro, we’ll win. We’ll win though.’
We’ll win. I probably smiled to myself or something and knuckled down with whatever I was doing and didn’t think much about it until a few days later I saw an interview with Sam Altman where he sat back on a big chair and in his fried little voice said that there are whole categories of labour that will be ‘totally gone’ in a few years. Dario Amodei, Altman’s CEO-nemesis, chief admiral of Anthropic, said that half of all white collar jobs could be gone in the next five years.
Altman and Amodei are the princes of the Valley, Henry Fords whose orders will dominate the future of American industry. When I heard them speaking, I realised that there was a pattern to the way they related to the world. These claims about AI futures, the way they prophesy doom as well their own dominance, are a special kind of Silicon Valley high-agency talk. It connects Altman and Amodei to each other, and it connects both of them kind to my guy in the Capital One Café predicting that he will win it all. A particular skill in Silicon Valley is to imagine a future in which there will be nothing but you and your ideas, and to make other people believe it. This skill is the speck of dust that betrays the goldmine of agency.
These kinds of claims about the future are intimately related to the capital structure of Silicon Valley. Before a startup has any money, any customers, any IP, any employees, what is it? It is simply a claim about the future. For many companies, it will never be much more than that, taking on millions of funding and achieving little in the process.
III
Let’s say you’re at some event.
‘What are you building?’ people ask.
No one says we have a limited window to sell chat interfaces to an industry with slow adoption rates before the frontier labs build more integrations and do it much better than we ever could.
This is a theory of the future, but it’s not going to win you any support. No one is going to be impressed or ask your opinion on things.
A more likely kind of response would be: The world works thanks to goods transported on ships. And shipping only works with maritime insurance. We are building the first agentic maritime insurance underwriter to remove the bureaucratic bottleneck from logistics entirely. By 2030, all ships will be directed and commanded by robots.
Press him on that last claim, why don’t you? Yes, you standing to the side, looking at your phone as if we don’t know you’re listening. Ask him what he thinks 2030 will be like.
Well, of course this is just on current trends. I’d give maybe 50% chance.
The Bay Area Rationalist culture can be heard in this grasping after probabilities. The paradox: expressing your uncertainty about the future by putting an exact number on it. Reality is unknowable, but not so unknowable that you can’t say how unknowable it is.
IV
Lawrence Ferlinghetti’s not-bad poem ‘Dog’ is about a dog walking through the streets of San Francisco.
The dog trots freely in the street
and sees reality
and the things he sees
are bigger than himself
He pads past puddles, babies, and policemen. The dog is open-minded, yet close enough to the ground to smell the urine and dirt that make up the city. He is ‘a real live / barking / democratic dog / engaged in real / free enterprise’. The poem ends with describing the dog as if he is like the HMV dog ‘with his head cocked sideways’,
looking
like a living questionmark
into the
great gramaphone
of puzzling existence
with its wondrous hollow horn
which always seems
just about to spout forth
some Victorious answer
to everything
The streets of San Francisco in real life are not as bustling or as hay-wire as they are in this poem, but when you go around them you still sometimes feel like this dog. You look at everything, slightly puzzled, democratically puzzled. In a way that is different to the other American cities I have visited, there is a democratic urgency about SF which you can see in the density, the fire-escapes, the blocks which tumble down the hills. The notorious zoning and the sarcastically-bad rents counterintuitively contribute to the sense that everything is a bit disrupted but in a useful way. Even the places from Vertigo which should feel like museum pieces don’t because you when you see them you just remember how that movie chops your brain up. No wonder Dog looks at it like a ‘living questionmark’.
I read this poem in the famous City Lights Bookstore in North Beach. It is an extremely American store, with lots of awful history books and a philosophy section full of Chomsky. I didn’t know this at the time, but it turns out that Lawrence Ferlinghetti was in fact the guy who started the store in 1953. Ferlinghetti was the first person to publish Howl in 1955, which led to his store getting raided by the FBI for obscenity. He was an anarchist trouble-maker, perhaps the arch figure of SF’s radical past, still looking down and forward at everything that happens in the city today.
Anyway, as I was reading the poem I realised that Ferlinghetti’s ending was the best description I had ever read for the specific way of speaking and imagining that I had noticed in my time in San Francisco. The Dog notices that the city always seems ‘just about to spout forth / some Victorious answer / to everything’. The answer, forever just incoming, forever pre-empting, the answer which will declare victory over everything — this the founder’s refrain of we’ll win bro, or Altman’s vision of a world without workers and only GPTs.
The victorious answer is a derivative of SF’s agency culture. It is a great way to get ahead, to dress for the job you want. It is a prediction which says what the world will be like if everything works out exactly according to your ideas. It is a prediction, but it has to be outrageous, and it has to be entirely yours. The victorious answer is when Peter Thiel says that we will soon be living under ‘the moderate rule of the anti-Christ’. It’s what Marc Andreesen means when he talks about how technology will 'drive prices effectively to zero’.
The victorious answer is victorious because it states what will need to be true for the person saying it to be the winner. When Musk claims that the species will be interplanetary, he is also saying ‘if I get it the way I want it’. But the victorious answer is also victorious because the speaker is asserting a victory over you, the listener, by saying something so enormous, so brazen. They are unleashing a bold gambit in the discursive game, and the question is whether you will accept it or fail to meet the challenge.
V
I was in SF because I have a startup. My startup is one of thousands of startups. We are trying to do something about the crisis in construction productivity. There are many challenges, many things which we do not yet know. There is also the constant temptation to forecast the ways in which it might in fact work, all the ways in which the world’s problems might be solved by me and reality itself might sway to my command.
This means that there is the constant temptation to talk in the same way that everyone around you is talking, to start dropping high-agency victorious answers in every conversation you have. If I’m tempted, imagine what it’s like for someone younger than me, someone less cynical and filled with UK pessimism. Imagine you’re a 22 year old, you’ve just moved to SF after doing a high-status degree at a high-status school which it would have been cooler to drop out from but your parents were too strict. You move to SF because it’s serious money, and when you get there you live in some hacker house with 5 other people, one of whom was the Moldovan u21 chess champion, another was a co-author with Karpathy, and the last one has already exited a crypto wallet startup for 8 figures. Imagine you live with such people, and you happen to be at some networking or quasi-social event, and you’re chit-chatting away, maybe to one of the few people in this city that isn’t a founder, and they decide to ask you a question about something which maybe you know a little bit about but not really enough to appreciate quite how little you understand. Of course, if you’re this guy, you don’t defer, you don’t parry the question, you have a go, you answer, you find yourself saying something like, ‘well, yes, I would say that there is a 50% chance of humanity being interplanetary before the end of the century.’
You’d be right to model this behaviour, because it’s not just the Amodeis and the Altmans who are showing you the way — this behaviour is also exhibited by the substrate of the next industrial revolution.
It’s not just the nakedly ambitious who talk like this. Even the people who kind of see themselves as antagonistic to the CEO AI prophets use the victorious answer: I’m talking about the AI Safety Guys. A month ago or so there was a minor kerfuffle in the biz about an epically detailed forecast done by some A-list posters called AI2027 which made the rather king-size claim that it is plausible that by 2030 we could all be murdered by a bioweapon unleashed by a naughty AI. AI Safety guys are not — at least not so openly — going after status and money, but the way they speculate about the future takes the same structure as everyone else. When someone gives their p(doom), for example, they are saying that ‘if I am right, no one is right’. It is a kind of victorious answer in that it too is a prediction about how only they can win.
It is well known that Silicon Valley thought is a mixture of democracy and domination. The projected worldview of the first wave of platform technology was democratic — this is the ethos of ‘here comes everybody’. Yet at the same time the fact that everyone did join platforms meant that the companies that run platforms achieved a level of success rarely seen in the history of capitalism. Google when it started in the late 90s was goofy, irreverent, and unstructured. When you get 3 billion users and a spending power more than most militaries, then those things are impossible to maintain. As Mark Zuckerberg said in a 2010 memo, ‘the cost of an open culture is that we all have to protect the confidential information we share internally’. Openness very much resembles a more normal closedness. In the transition from startup culture to big corporate culture, a lot of these companies have lived the contradictions between democratisation and domination.
The play of democracy and dominance will persist in the AI timeline, too. If intelligence is commoditised, then high factors of it become available to people for whom it otherwise was out of reach. At the same time, AI adoption can prove effective at centralising intelligence. For all the talk of open models, OpenAI still dominates chatbot usage. It is unprecedented to have a plurality of the world’s cognitive load is done by one entity — although the word entity is quite odd here.
As long as there is technology, and there are people trying to make more of it, there will be people proclaiming their victorious answers. They will continue to voice their opinion in true democratic spirit, and those opinions will continue to be about total lightcone domination. While they are doing this, there will still be many little dogs listening to them with their little questionmark faces. They will be staring into the gramophone which is not playing anything and yet there is deafening noise all around. The answer is new; it is victorious. It is the soul of an agent. The agent is you.
banger
i liked this