Monitoring the situation

Last December, touring Susquehanna’s office, the sheer number of screens at each trader’s desk confirmed a narrative I had long known but had never believed: that traders must monitor so many situations.

On every firm’s website, overlaid with company descriptions chock full of corporate buzz-words (“we commit our capital, global reach, technology, and expertise to help institutions meet their trading goals”) and vague characterisations of the work done (“curious people work together on deep problems”) are promotional photos of the workplace where everyone—and I mean everyone—has absurd multi-monitor set-ups—that one image of the new JPMorgan office building comes to mind. Ubiquitous is the Bloomberg terminal. But never before had I bought this story—surely the work of a trader was not so passive.

I doubted, though, that SIG traders deploying more capital than I could comprehend were putting on a charade for a ragtag group of undergrads. Their success as a firm implied their workflows were optimised. And thus, although some traders balanced on half yoga balls while others lazed in expensive looking chairs, though some stood and others sat and others alternated, the invariant of so many monitors (I’d wager they had six on average) had to be optimal. My question thus morphed from ‘is this the case?’ to ‘why is this the case?’

I think I understand now. These days, I typically have four windows open in my foreground:

1. Command Prompt

In the bottom left of my screen is typically a terminal with five tabs: three run instances of Claude Code, one SSHs into a Mac Mini running Openclaw, and one holds the data feed of a dashboard.py, in which rushes a ferocious river of websocket retrievals and API calls.

It’s a terribly abstract affair. I saw some tweet about how a 10× dev in 2026 “sits motionless, like a spider in the centre of its web, but that web has a thousand radiations, and he knows well every quiver of each of them. He does little himself. He only plans. But his agents are numerous and splendidly organised." I do not know if my agents are splendidly organised (perhaps all my tools merely elevate me to a 1× dev), but I certainly relate to the rest. I do very little besides orchestrate—I approve Claude’s permission requests when it asks, and that’s about it. And these days more and more I activate dangerously-skip-permissions…

Back in the pandemic, when ChatGPT had just released, the user experience felt like that of a Yukon prospector or a snow plow. You’d have Chat spit out a mountain of text, and trudge through it with a metal detector, panning for gold—precisely why teachers (rightfully, in my opinion), recommended it for brainstorming or outlining—while shovelling the slop aside. But these days, the roles are reversed. I trod through Claude’s work and might catch edge cases or lazy implementations, but per Karpathy "the bits contributed by the programmer are increasingly sparse and between."

I don’t feel behind, raw capability-wise. I maintain two tenets: first , "whatever level of abstraction you are handing off to your agents you should probably be doing one level above that” and second , "if we're working a manual labor job it's fine to have AI lift heavy things for us because the actual goal is to move the thing, not to lift it. This is the exact opposite of going to the gym, where the goal is to lift the weight, not to move it.” But that’s not what I’m scared of losing out on.

My friend lamented to me two days ago that Cursor trivialises so many of the aspects of coding which he enjoys. I can only sympathise; I am lucky to have specced during high school mostly into math and not so much into CS. But I fear the future. I really love writing, and it’s only a matter of time before some lab releases a model which spits out Pynchonesque prose with none of the cringe tricolors and meaningless metaphors currently indicative of LLM slop. This will be hard, obviously, since prose is difficult to verify and thus difficult to reinforce. But I am not naive. Banal but always applicable : "I want AI to do my laundry and dishes so that I can do art and writing, not for AI to do my art and writing so that I can do my laundry and dishes.” It’s terrifying. Maybe Atlas is right; perhaps the only things that will matter in a year or two will be taste and agency.

All I do is sit back and watch while Claude waddles or puzzles or schleps or combombulates; there’s an odd element of cuckoldry. What occupies me in the meantime?

2. Ring Game (NLHE)

Directly above my terminal, in the top left corner of my screen, I have a ClubWPT Gold 0.01/0.02 table open, something which demands periodic attention but which mostly serves to entertain me.

Poker is particularly thematic, I believe, since its gameplay consists—like everything else here—of nothing more than decision-making. I decide whether to fold, call, or 3-bet and then on a meta-level whether the table is still soft, whether the player to my left is still loose and passive, and whether I must top off or cash out.

But, honestly, there is very little decision-making even there. Most of my decisions are automatic: preflop charts I have memorised and routine exploitations (tighten range, squeeze higher, c-bet less, value bet more, underbluff, overfold) I know by heart—amateurs love yapping about playing the man as if that adds some romantic element to the game, but any good player knows that exploitation, like everything else, is systematised easily. Even when I do make a decision, e.g. after postflop deciding whether to check it back or bet, the decision tree is discrete, not continuous—for the layman, I should mention that well-played poker has fixed bet sizes (otherwise you would be telegraphing the strength of your hand).

Of course, beauty can still be found in discrete things. Per combinatorics, a lot of options over a lot of game-states yields ‘unique’ and thus beautiful lines—more games of chess than atoms in the universe and all those trite factoids. And I suppose on some level, continuity arises by simply taking the limits of discretion; even art is merely millions of micro-decisions.

That being said, though I occasionally play a gorgeous hand, trapping expertly with the nuts or hero-calling with my bottom pair bluff-catcher on the wettest board you’ve ever seen, the vast majority of gameplay is mechanical. I know which button I’ll click within a microsecond of seeing what I’m dealt. And I don’t necessarily dislike this. I often joke to my friends that in another world I would not mind incredibly repetitive labor. I am mostly serious; Gogol’s Akaky or Melville’s Bartleby lived great lives. I’d derive a great deal of meaning from such mindless tasks; I find Zetamac and Monkeytype and Human Benchmark amazingly entertaining and thoroughly enjoyed the Optiver OA. I think everything which can be said about this has already been said, but, nevertheless, something feels off when the creative process is so divorced from the fruits of labor.

3. Firefox

Directly to the right of my terminal and my poker is a thin and tall window, a browser open most of the time to its pinned tab, localhost:8000, in which runs the aforementioned dashboard.py, and on which I monitor around eighty markets, three hundred limit orders, as well as a variety of statistics (my liquid cash, api usage, connectivity, etcetera). It is here, perhaps, that the phrase ‘monitoring the situation’ is most apt. The frontend reads from some SQL tables, a half dozen python scripts, and hundreds of websockets. It does some calculations in the backend before calling APIs which place orders. I don’t move an inch, and it makes me quite decent money, a few hundred dollars a week.

But though this profit is ostensibly the fruit of some labor, and I eat that fruit via purchasing Poke bowls from Hutch or burritos from Tiffin, I often wonder if it’s the fruit of my labor. There are concerns with capitalism alienating the worker from the fruits of his labor; I experience the converse: I enjoy the fruits while it seems Claude deserves them (though I suppose Claude does receive them; I pay Anthropic for Max). At any rate, I am just monitoring the situation.

And so, what is my workflow? I monitor my agents, letting them run amok, and monitor an automated dashboard, letting it trade itself. I monitor a poker table—perhaps the singular most monitor-able game, folding 85% of hands preflop anyhow. My eyes flicker around, watching markets glow purple when my orders get traded with, watching my opponents battle with each other after I’ve long since folded, watching Claude’s little ASCII orange star, hexagonal, spin around in illusory circles. It’s a game of whack-a-mole, almost, but I rarely hit any of the moles; they take care of themselves. And so my keyboard is rarely used, and my hands are occupied by entirely meaningless distractions; I shuffle fourteen poker chips in my left hand and spin a pen in the right.

4. Aquile Reader

Then finally to the far right of my screen I have a libgen.li derived epub open, and my eyes spend most their time there. Because what else is there to do? I’ll scroll Twitter until I hit my time limit, play some bullet on lichess, and reach over to open some vile X-rated websites on O.’s laptop while he’s in the bathroom. And then I get bored, and so I open a book. I wrote to a friend in November that I wanted a life wherein I do technical work and have sufficient work-life balance such that I can read and write and think. Is this not precisely what I wanted?

I’m not sure. It feels like on some level the technical work here is not work. I suppose I alluded to this earlier, but I want to flesh the idea out a bit more. Walter Benjamin famously argues in The Work of Art in the Age of Mechanical Reproduction that mechanical reproduction devalues the work of art. I think we probably have the same case here. The inputs I give are simply not wide enough in degrees of freedom for the output to be anything I can call my own. It’s a bit interesting, since the work is probably nonfungible and thus what he would call authentic. But the locale/context renders it without what he calls aura. Which I agree with, in all senses of the word.

This is not to say I am unfulfilled or despondent. I think it’s a pretty nice workflow. I get to read and write and study and so long as I halfheartedly monitor the situation Claude chugs away on new projects while prior projects trade for me. But I am not sure I like where we are going.