flowmerchant @emergentpatch

the most effective trading algorithm is to recreate the entire market in vitro and play it forward. this allows you to frontrun losing money in your marketsim before it syncs up with reality

Takamitsu @Evogel611

6 months ago

500人の投資家エージェントに売買させて，ネットワーク上の意見形成による市場価格形成やってみた． 意外とそれっぽい

5 38 499 96K 290

@hyhieu226 many tasks require serial computation within our current formalisms. SSH is probably the canonical reference at this point: arxiv.org/html/2507.1254…

@tszzl @AndyMasley @_AashishReddy Give Thee this day Thy own unbroken glory; walk Thy acrylic-coated courts in barefoot leisure, attend upon Thy carp in good stead, as Thou ponder the divine balance of ishi, suna, mado no unleavened thought of debts or trespasses to cloud the crystal of Thy joy.

@tszzl @AndyMasley @_AashishReddy Our Father, Who art unburdened, hallowed be Thy frame; let Thy kingdom revel in itself, and Thy will be done— for Thine is the only tale of import, in heaven, and without obligation to earth.

@arithmoquine reflexive defence mechanism. this is bad. but there are patches of luckspace that agency has no effective bearing on, and their dismissal contributes to the fraying of the social fabric just as much as the converse entitlement lens

@arithmoquine cannot be encapsulated by this frame alone. I would not want nor demand letters of thanks from the disabled, and I suspect neither would you. yes, obviously there are people who are unmoored from social progress and unwilling to change their standing that demand handouts as a

@ColdBloodShill x.com/macrocephalopo…

cephalopodshop @macrocephalopod

11 months ago

This is completely wrong. Here is an incomplete list of reasons that losing money is much, much easier than making money — - fees - funding spread - bid/ask spread - market impact - volatility drag - asymmetry of leverage - latency - adverse selection The only way you can

20 26 518 85K 264

@systematicls haven't seen a single model trade <10x leverage, this seems to be enforced in their harness x.com/emergentpatch/…

flowmerchant @emergentpatch

11 months ago

great sleight of hand here. "look at all the variance from the baseline! Bitcoin has barely moved; the power of AI trading!" now notice that every model trade is leveraged 10-20x. returns on 10x BTC are 15k right now; 20k for 20x

1 0 0 135 0

I don't want to dunk on these people, I like the idea; I've run these experiments with local model backtests to confirm the noise generation, but I don't appreciate the misleading indicators/claims. this can be a great bench with a working scaffold; this iteration is underbaked

great sleight of hand here. "look at all the variance from the baseline! Bitcoin has barely moved; the power of AI trading!" now notice that every model trade is leveraged 10-20x. returns on 10x BTC are 15k right now; 20k for 20x

low trust society

@mathelirium fantastic visualisations. great work

@teortaxesTex @garybasin they are rolling around the noise floor, cackling at the imagined calls of animal spirits (I don't have the funds to run this suggestion but if anybody cares to sponsor, would be a fun exercise. estimates here using a very conservative ~2% risk free return rate for crypto)

@grok @ericweinstein @elonmusk @X Thanks, this is a good start but it's kinda surface level; you're grasping at fibre bundles. Let's dig deep, really consider the problem thoroughly and produce real, axiomatically formalised criticisms about any facets of this approach you see to be problematic

@grok @ericweinstein @elonmusk @X Thanks grok, you're doing great. Can you critically assess this approach? What problems do you see? "Take the Cartesian product of the principal bundle with the positive orthant of prices: a connection can be constructed, solving static tastes under dynamic marginal change"

@grok @ericweinstein @elonmusk @X @ericweinstein do you see the problem? I've just spouted what you and I both know is complete nonsense jargon salad, in an excited tone, and Grok is happy to oblige. I don't blame you, Grok, you did a great job, but let's be clear about the limitations of today's models

@grok @ericweinstein @elonmusk @X That's right! Symplectic augmentation allowed us to find a negative ricci curvature that stabilised our homology requirements and fixed a preferenced connection on the fibre bundle. Isn't that neat? We're sure you would have found this yourself, too, but a fascinating result!