drew

Aspiring Quant // Ex SWE

[drew @DrewMeade1]

when will a jepa model be able to handle dividends in arrears, lord knows llms can’t

[drew @DrewMeade1]

@EMostaque noble prize larping now

[Mathematica @mathemetica]

The relationships among common probability distributions through limits, special cases, and transformations.

Examples include the binomial approaching the normal as n → ∞ (μ = np, σ² = np(1 − p)), Poisson as the limit of binomial (λ = np, n → ∞), exponential as gamma(r = 1), uniform as beta(α = 1, β = 1), and Cauchy as t(ν = 1).

These insights guide statisticians in choosing appropriate models and approximations for data analysis in finance, epidemiology, and engineering.

[drew @DrewMeade1]

@dedene probably still worse than prime opus 4.6

[drew @DrewMeade1]

@HowToAI_ @zerohedge JEPA

[drew @DrewMeade1]

i need more time to study

[drew @DrewMeade1]

new prospa album saved dance music in real time

[drew @DrewMeade1]

@basecamp Can't see the tasks on the schedule for a specific day in full screen. Please fix this.

[Ivan Velichko @iximiuz]

LeetCode but for Linux, Docker, Kubernetes, and Networking problems? 🤔

245 hands-on challenges (over half are free) based on real-world scenarios that come with:

- Preconfigured environments
- Automated verification
- Hints and editorial solutions

https://t.co/ZTwhoT3YZx

[drew @DrewMeade1]

training myself to be a delusional optimist

[Dan Rosenblum @SharkAlertsBio]

Here is a snippet of what I wrote this morning $ABVX

[drew @DrewMeade1]

It starts

[alphaXiv @askalphaxiv]

"Learn from your own latents, not tokens: A Sample Complexity Theory"

This paper explains why data2vec and JEPA can learn with much less data.

They showed that when data has hidden hierarchy, token prediction becomes harder as the hierarchy gets deeper. But latent prediction keeps the learning problem simple at every level.

Which suggests that models may learn faster when they stop predicting raw tokens and start predicting their own abstractions.

[drew @DrewMeade1]

@InvestingVisual As of today 😭

[will depue @willdepue]

@a1zhang this shit is so annoying man. 'recursive language models' you mean putting your agent in a for loop. wow. incredible work man. truly groundbreaking. i'm sure you were the first person to think of this

[drew @DrewMeade1]

I’d say @AnthropicAI has some bugs to work out

[die goldene horde @david_streever]

@longironcondor the children yearn to tighten the bid-ask spread

[drew @DrewMeade1]

@RepLuna 🐘

[drew @DrewMeade1]

@silver__tsuki They don't want to expose the losses to the masses

[drew @DrewMeade1]

@Teknium Burned 100k tokens in 10 mins for me. Not lazy but definitely inefficient.

[Tokyo Deep Value @TokyoDeepValue]

Fuckers in the lecture hall telling me, always on the goddamn CFA forums

"Japanese net-nets ain't 'bout this, Japanese net-nets ain't 'bout that"

My boys trading at half of fucking liquidation value

He, he, they say ol' Ben just makin' it up

Shut the fuck up

Y'all CFAs ain't know shit

All y'all motherfuckers talkin' about

"Net-nets a value trap

Net-nets ain't growth

Net-nets a relic"

Shut the fuck up

Y'all don't sit with these tickers

Y'all don't know they got more cash than market cap

Trading below working capital night after night and shit

Bhai been compounding quietly since fuckin' Showa era

Motherfucker, stop fuckin' playin' 'em like that

Them DCF savages out there with they terminal growth rates

If I catch another motherfucker talking

sweet about "quality compounders at 40x earnings"

I'm fucking beatin' they ass with Security Analysis

I'm not fucking playin' no more