Posts
RSSNotes on software, language, and the overlap between them.
Introducing agentpane
A local web UI for AI coding agents. Multi-pane, multi-session, streaming — run Claude Code and Codex side by side from your browser.
Introducing helm
A typed TypeScript framework for AI agents. Replace dozens of tools with two — search and execute — and sandbox LLM-generated code with granular permissions.
Introducing tokka-bench
A comprehensive evaluation framework for comparing tokenizers across human and programming languages.
Dialects for Humans: Sounding Distinct from LLMs
Humans are developing new linguistic patterns to distinguish themselves from AI-generated content, and the rate of change will accelerate.
Using HuggingFace Datasets Offline
How to save a HuggingFace dataset to disk and use it offline
Tips #1
Markdown detection in Google Docs, swiping between tabs in Brave Browser for iOS, and running TypeScript files from the command line.
Rebuilding Alpaca with the Hugging Face Trainer Class
Fine-tuning Llama-2-7B using the Alpaca dataset and Hugging Face Trainer
Introducing gom: GPU Monitoring across Containers
I published `gom`, a CLI tool for monitoring GPU usage across Docker containers.
Enroot on Slurm for Distributed ML: Part 2
How to use Enroot on Slurm for containerized multi-node training.
Enroot on Slurm for Distributed ML: Part 1
How to use Enroot on Slurm for containerized multi-node training.
Quick & Helpful Slurm Commands
A quick guide to using Slurm for distributed machine learning.
Setting Up Docker for Machine Learning
The Dockerfile I use to set up my machine learning environment.
Accelerate vs. DeepSpeed vs. FSDP
Which one should you use for distributed training?
LLMs Will Never Be Able to Do (Complicated) Math
Since contemporary LLM architectures lack recursion, they're fundamentally incapable of doing some math operations.