Notes on Roadtrips by The Browser Company

What roadtrips may or may not have to teach us about creativity, ingenuity, and building something e...

63 Chinese Cuisines: the Complete Guide

Not one cuisine, not eight, but many more. Our best shot at a comprehensive-ish guide.

Improving Recommendation Systems & Search in the Age of LLMs

Model architectures, data generation, training paradigms, and unified frameworks inspired by LLMs.

Map Features in OpenStreetMap with Computer Vision

Mozilla.ai developed and released the OpenStreetMap AI Helper Blueprint.

The Differences between Deep Research, Deep Research, and Deep Researc...

Dive into the world of AI engineering with this exploration of Deep Research in report generation.

The Model is the Product | Vintage Data

Old data, new models

ARC-AGI Without Pretraining

Training a Rust 1.5B Coder LM with Reinforcement Learning (GRPO) | Oxe...

Group Relative Policy Optimization (GRPO) has proven to be a useful algorithm for training LLMs to r...

ChatGPT Clicks Convert 6.8X Higher Than Google Organic

作者发现自己的网站，来自ChatGPT的点击转化率是Google的6.8倍，基于LLM的搜索优化时代开始了。

It's still worth blogging in the age of AI

在AI时代写作的意义。

GitHub - therealoliver/Deepdive-llama3-from-scratch: Achieve the llama...

中英双语从0复现了LLama3。

The Profitable Startup - Linear Blog

Linear讲述了自己的盈利之路。“盈利能力最被低估的一点是它给您带来了多少安心。一旦你盈利了，你就不再担心生存，而是专注于真正重要的事情：创造伟大的东西。以你想要的方式建设。

Lifestyle Programming

最近几年开始流行“一人公司”，作者2013年就开始这么干了。

20 years working on the same software product

这个好玩，因为自己要结婚摆宴席写了个排座位的软件，原来英国人结婚也要吃席。以及，这样一个冷僻的软件竟然有这么大的影响力。

GitHub - bytedance/monolith: A Lightweight Recommendation System

A Lightweight Recommendation System. Contribute to bytedance/monolith development by creating an acc...

Understanding Reasoning LLMs

Methods and Strategies for Building and Refining Reasoning Models

Open-source DeepResearch – Freeing our search agents

We’re on a journey to advance and democratize artificial intelligence through open source and open s...

All the Transformer Math You Need to Know | How To Scale Your Model

Here we'll do a quick review of the Transformer architecture, specifically how to calculate FLOPs, b...

The Artificial Intelligence Revolution: Part 1 - Wait But Why

Part 1 of 2: "The Road to Superintelligence". Artificial Intelligence — the topic everyone in the wo...

I conditioned myself to fail

(WIP) A Little Bit of Reinforcement Learning from Human Feedback

The Reinforcement Learning from Human Feedback Book

How to Train an AI Image Model on Yourself

It takes less than an hour and $3.

R1-Zero and R1 Results and Analysis

An analysis of Deepseek's R1

Building effective agents

A post for developers with advice and workflows for building effective AI agents

The 2025 AI Engineering Reading List

We picked 50 paper/models/blogs across 10 fields in AI Eng: LLMs, Benchmarks, Prompting, RAG, Agents...

AI Founder's Bitter Lesson. Chapter 1 - History Repeats Itself | Luka...

Historically, general approaches always win in AI. Founders in AI application space now repeat the m...

The Anti-Social Century

Americans are now spending more time alone than ever.

The Evaporative Cooling Effect in Social Network : Networks Course ...

Cognitive load is what matters

There are so many buzzwords and best practices out there, but let's focus on something more fundamen...

Building effective agents

A post for developers with advice and workflows for building effective AI agents

People who are going to change the world

Inspired by Patrick Collison's list of interesting people.

People · Patrick Collison

Moon – Bartosz Ciechanowski

Interactive article about the Moon

Making “Social” Social Again

Announcing Mozi

The Making of Kino: You Can (Not) Build a Pro Video Camera App in Four...

A true story of indie app development: why we made the plunge to build our pro video camera app Kino...

“Link In Bio” is a slow knife - Anil Dash

A blog about making culture. Since 1999.

Pushing the frontiers of audio generation

Our pioneering speech generation technologies are helping people around the world interact with more...

Using Reinforcement Learning and $4.80 of GPU Time to Find the Best HN...

Convert expensive LLM prompts into fast, cheap fine-tuned models

Exclusive: Flappy Bird Creator Speaks

How did a chain-smoking geek from Hanoi design the viral hit Flappy Bird - and why did he walk away?

It made me $500k but I think content is a bad business | Swizec Teller

Why I didn't become a full-time creator/educator

Entropy Explained, With Sheep

What is entropy, really? Why does it always keep increasing? Why don’t eggshells uncrack, or wine gl...

Cartographist

Web Browser Optimized for Rabbit-Holing

What I’ve Learned in the Past Year Spent Building an AI Video Editor

Lessons from An Unexpected Year in AI

An Interview with Meta CTO Andrew Bosworth About Orion and Reality Lab...

An Interview with Meta CTO Andrew Bosworth About Orion and Reality Labs

Redesign the NYT

Asking the wrong questions — Benedict Evans

With fundamental technology change, we don't so much get our predictions wrong as make predictions a...

The Infinite Audio Book

I had this seemingly dumb idea. Most such ideas are fleeting, and you don't act on them.

Signal: How To Ethically Boost Your Revenues

In this case study, we look at how simple psychological & UX strategies can help the messaging app, ...

The secret inside One Million Checkboxes

Teens wrote secret binary messages in One Million Checkboxes. I found them.

Margot Monet's creators hoped she would rake in thousands as an AI inf...

As people spend more time online, bonds forged with influencers generated by artificial intelligence...

A Visual Guide to Mamba and State Space Models

An Alternative to Transformers for Language Modeling

How to think about HTML responsive images · Dan Cătălin Burzo

srcset, sizes, picture, source, and how they all fit.

The UX of UUIDs | Unkey

Unique identifiers play a crucial role in all applications, from user authentication to resource man...

Half-Quadratic Quantization

An in-depth article discussing the intricacies of efficient model quantization in machine learning a...

Answer.AI - You can now train a 70b language model at home

We’re releasing an open source system, based on FSDP and QLoRA, that can train a 70b model on two 24...

My favourite animation trick: exponential smoothing

I write stuff about math, simulation, graphics, gamedev, and programming.

Training great LLMs entirely from ground zero in the wilderness as a s...

Chronicles of training strong LLMs from scratch in the wild

Beyond Self-Attention: How a Small Language Model Predicts the Next To...

A deep dive into the internals of a small transformer model to learn how it turns self-attention cal...

Bill's World

I'm stuck in 1999 on GeoCities! I've made this time portal so you future people can visit my site.

Way Enough - Fine-tuning gpt-3.5-turbo to learn to play "Connections"

Uncensored Models

I am publishing this because many people are asking me how I did it, so I will explain. https://hugg...

The Girls in Their Summer Dresses

This Web site is dedicated to the wonderful world of the short story and to all who enjoy reading sh...

Attending NeurIPS 2023

A simulation of me: fine-tuning an LLM on 240k text messages - Edward ...

Armed with a download of my 240,000 text message and Whatsapp history, I was able to fine-tune an LL...

Will scaling work?

Data bottlenecks, generalization benchmarks, primate evolution, intelligence as compression, world m...

World Models

Can agents learn inside of their own dreams?

math team

and other horrible things you do to get into stanford

The Great AI Weirding

Or how AI will put you in recursive rat races if you let it, and you probably will.

Too Much Information

What if, hear me out, we are writing too many papers?

Fantasy Map Brushes

Welcome to a project I call #NoBadMaps. Here you will find brush sets and tools to create fantasy ma...

GitHub - moddio/moddio2: HTML5 multiplayer game engine

HTML5 multiplayer game engine. Contribute to moddio/moddio2 development by creating an account on Gi...

Blogging Like a Hacker

Making Synthesized Sounds More Acoustic

I have been experimenting a lot with finding ways to get more acoustic sounds out of synthesizers.

Speculative Decoding for 2x Faster Whisper Inference

We’re on a journey to advance and democratize artificial intelligence through open source and open s...

Simulating Fluids, Fire, and Smoke in Real-Time

Wordle Solving Robot

We've made a wordle solving robot! As usual, we're sponsored by https://www.PCBWay.com/?from=atomic.

SMERF

Project page for SMERF: Streamable Memory Efficient Radiance Fields for Real-Time Large-Scene Explor...

His software sang the words of God. Then it went silent.

Who was Thomas Buchler, the late creator of beloved Torah program 'TropeTrainer'? And can anything b...

Quit Your Job

Frank Chimero · The Web’s Grain

Frank Chimero’s Personal Website

The Placeholder Girlfriend

It became real when I saw the list. When I saw the rubric.

LLM Visualization

A 3D animated visualization of an LLM with a walkthrough.

My $500M Mars Rover Mistake: A Failure Story — Chris Lewicki

Some mistakes feel worse than death.

Writing Documentation for Your House

Have you considered writing documentation for your house?

Kiwi Hellenist: The camel, the rope, and the needle's eye

Many popular beliefs about ancient Greece and Rome are misconceptions.

FDA Center for Veterinary Medicine agrees Loyal’s data supports reason...

A historic milestone in developing the first drug for healthy lifespan extension in dogs

GitMounter // -dealloc

An old toy project for browsing the contents of a git repo as a filesystem.

Portuguese Orange, Persian Portugal - Untested

Portuguese Orange, Persian Portugal - Untested

Quanta Magazine

Quantum algorithms can find their way out of mazes exponentially faster than classical ones, at the ...

Scientists finally succeed in growing dolomite in the lab by dissolvin...

For 200 years, scientists have failed to grow a common mineral in the laboratory under the condition...

He thought tending bar sounded like fun.

Bill McCamley was NM's labor secretary, but working at a local movie theater taught him the truth ab...

S3 Express is All You Need - WarpStream

The future of modern data infrastructure is object storage

Parsing 8-bit integers quickly – Daniel Lemire's blog

Suppose that you want to parse quickly 8-bit integers (0, 1, 2, …, 254, 255) from an ASCII/UTF-8 str...

Unity Software to cut 3.8% of staff in 'company reset' | Reuters

Videogame software provider Unity Software will eliminate 265 jobs or 3.8% of its global workforce a...