Pajeet Davidson

They don’t understand that having corporate leadership comprised of a vellalar, a jatt, and two goltis is the Indian equivalent of having two black guys, a white guy, and a latino. This is as diverse as it gets

@anarch97 Even in America it’s been wholly debunked. IQ means way less than ppl think

and the 'great sort' DEFINITELY isn't true outside the first world. A 130 IQ middle-class Nigerian is not guaranteed great success, A 140 IQ Algerian is not boarding a flight to sf

Yea we all told him to do that. All men get together and conspire to rape. It’s our favorite past time

@nfntcyberking They’re gonna have a streamer who’s entire schtick is personalitymaxxing

Im like 3 years they will start teaching courses on how to have a personality

Unreal Tommy Shelby tat at the broad street run today

@jonah_b @milesgr_ The real edge her is the top shot route. Sell the tools to find good strats, tax the winners

@milesgr_ the other think is that, in theory, selling it should commoditize the strategy and make it not work

One thing I still don’t get is companies selling AI trading bots. If your AI trading bot is actually good at trading, why would you sell it and not just make money with it? To me, it feels similar to selling “how to get rich” courses. This is a strong view, weakly held. Would love for someone to steelman why a company with an agent that can profitably trade would rationally choose to sell it instead of use it.

@jonah_b Anything being sold is intentionally shit and can be traded against

If you’re serious about robot learning you (unfortunately) need to know about video compression. Camera streams dominate data volumes for most datasets at 90+% even when compressed. Video is more complicated to deal with but the size wins are too big to give up. The unit of compression is a Group Of Pictures (GOP). In the simplest case (what you should use in robotics), GOPs start with a keyframe (I-frame) that is followed by several delta frames (P-frames). Delta frames only need to encode the difference to the previous frame which is where the compression win comes from. That means to decode frame 15 of a 30-frame GOP you need to feed all the preceding frames in the GOP to the decoder to get out that one frame. The GOP controls the tradeoff between random access and compression. Why does this matter for robot learning? Because while training, dataloader performance is dominated by fetching and decoding video. To build a streaming dataloader (you need this for large datasets) it needs to take GOPs into consideration when fetching data for a time step. It’s hard enough to build a dataloader that doesn’t starve your GPUs that most teams forgo flexibility. That means researchers at most of the best funded robotics efforts currently wait around for large export jobs before training can start after each change to the dataset mix or the wrong hyperparameter. This situation obviously won’t last since they all know that experiment cycle times is a key lever to fast progress and the competitive pressure is enormous. If you want to compete in this space you need both flexibility and performance.

post pics of himself at Nobu and looks like a faggot

how is bro still alive

the lcs is an impeccable parody of a professional sports broadcast

Sucking titties be so awkward she just lays there dead silent, probably thinking yeah nigga you my son now

@danrobinson Have spent a lot of time on this in the past two months

I think there is low-hanging research harness innovation coming, because the best prompters usually say they are adding very little expertise—just nudging Codex/Pro/Claude generically—but ralph loops don’t work well for open-ended research yet Looking for people to work on this

Aight bro

@yenwod_ @DrDavidSimic how is full replayability even possible

@DrDavidSimic If your system is replayable you don't need extra observability at runtime, you can instead download the logs, replay, and observe the system state at any point in time

When trying to fix negative markouts in your market making system, top of the list should be digging into fills that happened during inflight cancels. These are trades your system knew were going to be toxic, but you got filled anyway. This may happen because you're still too slow on the tails and quoting too tight for your latency, or it could be an inconsistency in your business logic. Tracking these events involves accepting a small amount of overhead in the hot path but is generally worth it, especially in the testing phase with smaller amounts of capital. When such events cross a certain threshold, I like to attach a cancel and fill context for detailed analysis of the system state before and after. This can include details about the BBO, alpha states, latency deltas, skew, etc. Then, once such events have been reduced to an acceptable level, compile them out to shave the extra nanos off the hot path.

I don’t understand why they need 72m

The fact that everyone immediately realized the story was fake after the guy’s face reveal more or less proves that physiognomy is real

This team refuses to do anything but play to the level of their opponents. They only show up when they’re underdogs

I dare Shams Charania to break some news that turns out to be wrong