hoangtrungnghia (@nghiaht7) 's Twitter Profile
hoangtrungnghia

@nghiaht7

Data-focused software engineering

ID: 1170630919595544577

calendar_today08-09-2019 09:32:33

870 Tweet

136 Followers

2,2K Following

Zaid Humayun (@redixhumayun) 's Twitter Profile Photo

Are there resources around how b-trees or lsm trees are optimized when designing for cloud storage? Assuming that there are indeed optimizations specifically for cloud storage

Alex Prompter (@alex_prompter) 's Twitter Profile Photo

what are large language models actually doing? i read the 2025 textbook "Foundations of Large Language Models" by tong xiao and jingbo zhu and for the first time, i truly understood how they work. here’s everything you need to know about llms in 3 minutes↓

what are large language models actually doing?

i read the 2025 textbook "Foundations of Large Language Models" by tong xiao and jingbo zhu and for the first time, i truly understood how they work.

here’s everything you need to know about llms in 3 minutes↓
Colin Breck (@breckcs) 's Twitter Profile Photo

Predicting the Future of Distributed Systems: • Databases are being decomposed • Object storage is becoming the database • Edge and cloud are becoming a continuum • Programming toolkits and operational models are ripe for disruption • This is not happening because

Kelly Sommers (@kellabyte) 's Twitter Profile Photo

You asked for longer rants, so here are longer rants! After almost 10 years I'm back to blogging. Thanks for the encouragement. Link below in the thread 🔻

You asked for longer rants, so here are longer rants! After almost 10 years I'm back to blogging. Thanks for the encouragement.

Link below in the thread 🔻
v (@iavins) 's Twitter Profile Photo

The B Tree data structure fascinates me. Databases use B Trees to store data on disk, organizing everything into pages that typically range from 4kb to 8kb. All I/O operations happen in units of these pages. The page looks like this... (1/9)

The B Tree data structure fascinates me. Databases use B Trees to store data on disk, organizing everything into pages that typically range from 4kb to 8kb. All I/O operations happen in units of these pages.

The page looks like this... (1/9)
David Gu (@davidruigu) 's Twitter Profile Photo

we run postgres with 20k concurrent writers. at this scale, it's mandatory to deeply understand the internals of every piece of infra we ran into a performance bottleneck, where our query performance was limited by the rate that Linux could fork child processes! 🧵 👇️

we run postgres with 20k concurrent writers. at this scale, it's mandatory to deeply understand the internals of every piece of infra 

we ran into a performance bottleneck, where our query performance was limited by the rate that Linux could fork child processes!

🧵 👇️
Abhishek Singh (@natoshi_sakmoto) 's Twitter Profile Photo

How to get better at system design (for juniors) A pro tip: try explaining systems like a geek, forget your GF, become a nerd, you will become good soon, trust me. Not kidding but here are my 10cents. 1. Learn the basics first: HTTP/HTTPS, DNS, TCP, TLS, latency, plus OS

Thorsten Ball (@thorstenball) 's Twitter Profile Photo

me, 10 years ago: plz let me work on distributed systems! now: oh god please keep it one process one machine for as long as you can

Abhishek Singh (@natoshi_sakmoto) 's Twitter Profile Photo

Why you should be more active on X? No, it's not about the payouts! I tweet a lot about system design. Many people reply and join the discussions. I read all the comments and try to respond as much as possible. While reading, I think of better points to add - helping others and

Praetor (@fourvork) 's Twitter Profile Photo

how to get Harvard level education for free - pick the field: cs, philosophy, biology, econ, etc - get free courses from Harvard pll, MIT ocw, open Yale, Stanford - full courses with lectures, assignments, and exams included - ask gpt or gemini to build a full harvard-style

blue (@bluewmist) 's Twitter Profile Photo

start with the basics. 8 hours of sleep, 3l of water, meditation, 15 minutes of reading and some movement daily. make this your default routine, do not go back to your old habits. this compounds to affect your entire life.

Paul Graham (@paulg) 's Twitter Profile Photo

I just reread "How to Do Great Work." It's so long! But it also has less fat than most things I've written, which is a weird combination, because usually writing that's long on the macro scale is long on the micro scale too. paulgraham.com/greatwork.html

Abhishek Singh (@natoshi_sakmoto) 's Twitter Profile Photo

You're implementing a distributed counter (total page views) across 100 servers. Each server increments locally and syncs to database every 10 seconds, but the count is always incorrect. What's the bug? [A simple Distributed Systems problem]

Justin Skycak (@justinskycak) 's Twitter Profile Photo

Most people who "fail" at their dreams just run out of time to figure them out before life locks the doors behind them. It's usually not a lack of potential that gets you. It's the lack of urgency.

Lisan al Gaib (@scaling01) 's Twitter Profile Photo

Tri Dao is one of the few people on the planet who would actually deserve a 1 billion dollar compensation package surely he already saved the industry BILLIONS

Andrej Karpathy (@karpathy) 's Twitter Profile Photo

I packaged up the "autoresearch" project into a new self-contained minimal repo if people would like to play over the weekend. It's basically nanochat LLM training core stripped down to a single-GPU, one file version of ~630 lines of code, then: - the human iterates on the

I packaged up the "autoresearch" project into a new self-contained minimal repo if people would like to play over the weekend. It's basically nanochat LLM training core stripped down to a single-GPU, one file version of ~630 lines of code, then:

- the human iterates on the
Andrej Karpathy (@karpathy) 's Twitter Profile Photo

Three days ago I left autoresearch tuning nanochat for ~2 days on depth=12 model. It found ~20 changes that improved the validation loss. I tested these changes yesterday and all of them were additive and transferred to larger (depth=24) models. Stacking up all of these changes,

Three days ago I left autoresearch tuning nanochat for ~2 days on depth=12 model. It found ~20 changes that improved the validation loss. I tested these changes yesterday and all of them were additive and transferred to larger (depth=24) models. Stacking up all of these changes,
Daniel Lemire (@lemire) 's Twitter Profile Photo

Which programming language should you choose for teaching? Though many schools use Java, C#, C or C++, many others have adopted Python. The upside of Python is that it is somewhat easier to get going (helps motivation). The downside is that Python makes it harder to think about

Which programming language should you choose for teaching?

Though many schools use Java, C#, C or C++, many others have adopted Python. The upside of Python is that it is somewhat easier to get going (helps motivation). The downside is that Python makes it harder to think about