Slipping back into the quiet again, where echoes rest and fade, thoughts find stillness, and time forgets. If the winds remember, they'll carry me back next year. 🍂🌘

When the world looks down on you, it's okay. Because that means you're the one looking up, seeing things others can't. You've got the clearest view of all. Even the faintest stars only show themselves to the ones who keep looking up and who never stop, no matter how long it takes

All before your computer even realizes you hit a key ⚡️.

about 200,000 orders in that window. A premium-tier HFT firm operating in the tens of nanoseconds T2T latency could easily top 20 million, and the elite players the ones living in single-digit or 1 nanoseconds T2T latency could push past 200 million (theoretical maxima).

By the time we human does anything i.e. a mouse click, a keypress, whatever and the system actually registers it (roughly around 200-250 ms later), that’s basically an eternity for an algorithm. Even a low-tier HFT outfit running at a modest 1 μs T2T latency could fire off

10x engineer pt.3

Aston Martin 𓆃, Pure 🖤 Magic.

Next generation of developers.

us-east-1 is down Here is why:

https://t.co/P0xiCPNfAo

Jane Street's money printer FPGAs, optimized for high frequency trading, run at <600MHz, while your intel CPU from 20 years ago ran at some GHz. Speed and being low latency are not only about clock speed.

Adopted David Lattimore’s buffer reuse technique (for types sharing the same alignment and size) from the Wild Linker repo. By reusing preallocated buffers instead of frequent heap allocations, achieved measurable performance gains on hot paths with heavy allocation churn.

Interesting experience from @karpathy on this new LLM training repo.

Har Urooj Ka Zawal Aana Hai. - Applies universally to all things.

Fairness died the moment existence smiled and said it never owed you meaning.

While Zstd served us well, it belongs to the past. Enter OpenZL, the next-generation compression engine designed for speed and efficiency. [ https://t.co/ArSCHPTJzt]

Memory access is O(N^[1/3]) https://t.co/4VRZLFhUwm

Short-term smart is nothing but long-term suicide. Don't sugarcoat it, you're just screwing yourself slowly.

Matt’s talks never disappoint. The low level depth he brings is unmatched. 👌🔥

Python@HRT recently completed a firm-wide rollout of lazy imports, cutting startup times, lightening distributed workloads, and boosting developer productivity across the board. Read about how we optimized import times in a massive monorepo: