Related posts

AI research engineer working on large language models. He writes the Ahead of AI newsletter and is the author of Build a Large Language Model (From Scratch).

From one piece Recent Developments in LLM Architectures: KV Sharing, mHC, and Compressed Attention 2 beliefs, in the piece's order there

• Our readingkorrents.com

  Larger language models already have enough capacity that extra per-layer embedding tricks add little benefit to them.

  Their words

  However, larger models already have sufficient capacity where these extra embeddings may not help that much.

  magazine.sebastianraschka.com

• Our readingkorrents.com

  As reasoning models and agents retain more tokens for longer, KV-cache size and attention cost become the main constraint on scaling context.

  LLMs

  Their words

  As reasoning models and agent workflows keep more tokens around (for longer), KV-cache size, memory traffic, and attention cost quickly become the main constraints, and LLM developers are adding a growing number of architecture tricks to reduce those costs.

  magazine.sebastianraschka.com

Co-creator of Django and creator of Datasette; writes daily at simonwillison.net.

• Our readingkorrents.com

  Keeping a codebase clean now pays off directly rather than aesthetically, because a coding agent will copy the patterns already in it almost exactly.

  Their words

  one of the magic tricks about these things is that they they're incredibly consistent. If you've got a code base with a bunch of patterns in, they will follow those patterns almost to a T.

  youtube.com

Software engineer and a long-time technical lead of the Go programming language; author of plan9port and of the research.swtch.com essays.

• Our readingkorrents.com

  Writing a technique down, however basic it seems, is the first step to making it usable by everyone.

  Their words

  But writing a trick down is the first step to making sure everyone can do it: magic tricks can be basic but not obvious.

  research.swtch.com

Deep-learning researcher and Turing Award laureate; founder of the Mila institute.

• Our readingkorrents.com

  Access to large-scale generalist AI systems that could be weaponized should be limited, meaning their code and neural network parameters should not be released open-source.

  Their words

  At the same time, in order to reduce the probability of someone intentionally or unintentionally bringing about a rogue AI, we need to increase governance and we should consider limiting access to the large-scale generalist AI systems that could be weaponized, which would mean that the code and neural net parameters would not be shared in open-source and some of the important engineering tricks to make them work would not be shared either.

  FAQ on Catastrophic AI Risksyoshuabengio.org

Writer and novelist. Publishes essays and book notes at nateliason.com; author of the novel Husk and of Crypto Confidential.

• Lovedrcmnd.app

  Salt, Fat, Acid, Heat

  Their words

  Such an incredible book on becoming a better chef, with little tricks and methods I use almost every day now. Plus the recipes in the back are delicious!

  nateliason.com

Mathematician and maker of the 3Blue1Brown YouTube channel, which explains mathematics through animation. Creator of the open-source Manim animation library and founder of the Summer of Math Exposition.

• Recommendsaffiliate linkrcmnd.app

  The Art Of Probability

  Their words

  Hamming was a practitioner, spending much of his career in Bell Labs, so unlike many books on probability, this stays grounded in the kind of intuitions that come up a lot in solving practical problems. It has countless helpful puzzles and problem-solving tricks which are organized in such a way as to give a sense of the deeper theory underlying probability and statistics.

  3blue1brown.com

American software engineer; worked on code-server at Coder and teaches TypeScript and open-source contribution. Publishes at joeprevite.com.

• Lovedrcmnd.app

  VS Code

  Their words

  I switched from Sublime to VS Code years ago and I can't see myself going back. I used to use the Sublime keybindings in VS Code. However, last year I switched to the Vim keybindings. It took some time to reach the same speed, but I would say I've surpassed my previous speed and added a few new tricks thanks to Vim.

  joeprevite.com

Programmer; builds Mac and iOS apps, wrote WatermelonDB, and writes about tools, focus and software craft.

• Readrcmnd.app

  A Guide to the Good Life

  Their words

  Podcast (Part 2). A modern, approachable guide to Stoicism — a philosophy of life whose goal is to maximize the experience of tranquility. This is a seriously useful bag of psychological tricks.

  radex.io