211
I caught Google Gemini using my data and then covering it up
I asked Google Gemini a pretty basic developer question. The answer was unremarkable, apart from it mentioning in conclusion that it knows I previously used ...
212
Compiling LLMs into a MegaKernel: A path to low-latency inference
TL;DR: We developed a compiler that automatically transforms LLM inference into a single megakernel — a fused GPU kernel that performs…
213
Tools: Code Is All You Need
The solution to agentic flows was code all along.
214
Rob Pike got spammed with an AI slop "act of kindness"
Rob Pike (that Rob Pike) is furious. Here’s a Bluesky link for if you have an account there and a link to it in my thread viewer if you don’t. …
215
All AI models might be the same
what can language model embeddings tell us about whales speech, and decoding ancient texts? (on The Platonic Representation Hypothesis and the idea of *universality* in AI models)
216
Highlights from the Claude 4 system prompt
Anthropic publish most of the system prompts for their chat models as part of their release notes. They recently shared the new prompts for both Claude Opus 4 and Claude …
217
When did AI take over Hacker News?
One in three top 10 Hacker News posts in August 2025 were AI-related. Can we pinpoint when this trend started?
218
Local LLMs versus offline Wikipedia
How do the sizes of local LLMs compare to the size of offline Wikipedia downloads?
219
Scaling LLMs to Larger Codebases
Where to focus investments to best leverage AI tooling
220
With AI you need to think bigger
I have noticed something in the past 12 months, something a little profound. I have noticed that I am no longer scared that a project will be too big or too complex for me, or that a project will use a technology or programming language I don’t know.
221
GPT-5: Overdue, overhyped and underwhelming. And that's not the worst of it
A new release botched … and a breaking research new paper that spells trouble
222
223
224
The bitter lesson is coming for tokenization
Highlights the desire to replace tokenization with a general method that better leverages compute and data. We'll see tokenization's fragility and review the Byte Latent Transformer arch.
225
I fed 24 years of my blog posts to a Markov model
Yesterday I shared a little program called Mark V. Shaney Junior at github.com/susam/mvs . It is a minimal implementation of a Markov text generator inspired by the legendary Mark V. Shaney program from the 1980s. Mark V. Shaney was a synthetic Usenet user that posted messages to various newsgroups using text…
226
NVIDIA frenemy relation with OpenAI and Oracle
I’ve spent the last 48 hours completely falling down the rabbit hole of NVIDIA’s Q3 Fiscal 2026 earnings report . If you just skim the headlines, everything looks perfect: Revenue is up 62% to $57 billion, and Jensen Huang is talking about a "virtuous cycle of AI."
227
The Deep Research problem
OpenAI’s Deep Research is built for me, and I can’t use it. It’s another amazing demo, until it breaks. But it breaks in really interesting ways.
228
Important machine learning equations
A reference guide to fundamental machine learning equations with correct implementations and clear explanations of how they connect to each other.
229
I read all of Cloudflare's Claude-generated commits
Max Mitchell's personal portfolio website showcasing his photography, YouTube videos, coding projects, and work history.
230
231
A History of Large Language Models
Large language models (LLMs) still feel a bit like magic to me. Of course, I understand the general machinery enough to know that they aren’t, but the gap between my outdated knowledge of the field and the state-of-the-art feels especially large right now. Things are moving fast. So six months ago, I decided to close…
232
We need a clearer framework for AI-assisted contributions to open source
A framework for reasoning about AI-based changes you encounter or create.
233
AI killed the tech interview. Now what?
How can we do better interviews in the age of AI
234
235
Agentic Coding Recommendations
Current recommendations of agentic coding.
236
Compiling a neural net to C for a speedup
A cozy little corner of the web.
237
Designing NotebookLM
Designer, builder, and visual storyteller. Now building Huxe. Previously led design on NotebookLM and contributed to Google AI projects like Gemini and Search. Also shoot photo/video for brands like Coachella, GoPro, and Rivian.
238
Understanding Transformers Using a Minimal Example
Visualizing the internal state of a Transformer model
239
The RAG Obituary: Killed by agents, buried by context windows
I’ve been working in AI and search for a decade. First, I built Doctrine, the largest European legal search engine, and now I am building Fintool , an AI-powered financial research platform that helps institutional investors analyze companies, screen stocks, and make investment decisions.
240
AI Meets WinDBG
Because manually squinting at hex dumps is so last century. Let me show you how AI-assisted debugging is leaving WinDBG's command line in the dust.