Did you know that server certificate returned in the TLS server hello can be large? I have seen ones up to 10KB especially when the full chain is included. This can slow down the handshake especially when latency is high. Ways to address this: - Compress the certificate

ChatGPT is the modern version of Flaubert's "Dictionary of Received Ideas" (Dictionnaire des idées reçues), that is, a powerful cliché parroting engine. And, as they say in trading: "what most people know isn't worth knowing."

Everything You Always Wanted to Know About Mathematics [pdf] math.cmu.edu/~jmackey/151_1…

good summary about the floaty rock drama so far “if you’re not following LK99, you’re missing out on the most fun thing happening or the internet now. Feels like the old internet.” reddit.com/r/redscarepod/…

These 94 lines of code are everything that is needed to train a neural network. Everything else is just efficiency. This is my earlier project Micrograd. It implements a scalar-valued auto-grad engine. You start with some numbers at the leafs (usually the input data and the

Kenyans see fellow protesters killed in front of them but they choose to keep fighting Read the eyewitness report from Sky's Africa correspondent Yousra Elbagir trib.al/SkdPUtV

I just opensourced something I have been working on for months. I call it “super prompt” because it also allows some LLMs (claude) to come up with really novel ideas, (picture is an example the prompt is larger). Its built in XML agent format btw. Github in comments.

All main memory workloads eventually end up being bottlenecked by throughput of a DRAM. This further causes bad perf metrics which are reflected in CPU memory stalls. A master class in performance engineering! valkey.io/blog/unlock-on…

Every day on YouTube, people upload 4 million videos and watch 5 billion videos. Handling this staggering traffic requires a vast fleet of servers. So when a new request comes in, where does it go? How do you balance load across so many servers for so many jobs? I love this

This is the entire code needed to reproduce R1 lol Hundreds of Billions of Dollars Later

Moon One of the things we learned with Anthropics circuits work, is how the model explains it's thought process is not always what it's thought process actually was. Probably just something to keep in mind.

1. What every engineer should know when working with distributed systems. francofernando.substack.com/p/the-challeng…

Introducing The Most Advanced Memory System for LLM Agents MIRIX is by far the most advanced memory system in the world, designed to make AI truly remember, learn, and help you over time. Website: mirix.io Paper: arxiv.org/abs/2507.07957 Github:

I was one of the 16 devs in this study. I wanted to speak on my opinions about the causes and mitigation strategies for dev slowdown. I'll say as a "why listen to you?" hook that I experienced a -38% AI-speedup on my assigned issues. I think transparency helps the community.

This is an awesome article. The best part is their note to “build around the KV cache”. If your system prompt remains consistent, your tools remain constant, and you always append to conversation json… you will hit the KV cache often. Cutting down cost and latency.

When should you chunk your documents, before embedding, or after querying? Most RAG systems use 𝗽𝗿𝗲-𝗰𝗵𝘂𝗻𝗸𝗶𝗻𝗴 - the standard approach where you break documents into smaller pieces first, then embed and store them in your vector database. This requires upfront decisions

i recommend reading this

last month, Karpathy dropped the ULTIMATE guide to speed-running your way into LLMs in this project, you’ll build all the essentials, all under 8k lines of code > train the tokenizer, new rust implementation > pretrain a transformer LLM on fineweb > evaluate core score