Blog
32 articles
01 / The writing
Follow via RSS Articles
From Bag-of-Words to Jev: How Text Classification Got Here
Jev and System One Models: The Decision Without the Paragraph
Smarter Dumb Systems: An AI Agent Walks Into the Cheese Aisle
Inside DeepSeek-V4.1-Flash: Encoder-Decoder Backbones, Sparse Attention, and FP4 Cache
The History of LLMs, 2023 to 2026
Write the Policy at Inference Time: Shieldstral 1.0 3B for Local Text and Image Moderation
What Anthropic's J-space Actually Is (and Isn't)
metra-monitor: Self-Hosted Realtime Commute Alerts from Metra's GTFS Feeds
A Self-Hosted Garden Sensor Agent with Ecowitt, Claude, and Telegram on GCP
Running a Headless Claude Code Trading Agent on GCP with the Robinhood MCP Server
Kokoro-82M: Running a Local Text-to-Speech Model on One GPU
Running Ideogram 4 Locally: Quantized Inference and Structured JSON Captions
Simple Things Scaled Up
X-Token: Distilling Knowledge Across Tokenizers That Don't Speak the Same Language
Building an Agentic Movie Recommender on Cloudflare Pages
It Started as a Chatbot
LLM Glossary (In Progress)
DFlash: How Block Diffusion Breaks the Speculative Decoding Ceiling
Laws of LLMs and Agents
The Fragmented Researcher: Engineering Focus Around a 10-Month-Old
Large Language Models are beautiful.
27,000 Tokens Before Hello: The Agent Harness Tax
Gemma 4: Everything You Need to Know About Google's Most Capable Open Model
Embeddings are beautiful.
TurboQuant: The cheat sheet that ate your GPU (and how Google fixed it)
Doc-to-LoRA & Text-to-LoRA: How Sakana is teaching LLMs to learn instantly
BlinkThink: Self-Hosted Camera Snapshots with FastAPI and Gemini
Building a Browser Agent with Gemini and Playwright
The 10-Million Token Paradox: Decoding the Logic of Recursive Language Models
OpenClaw: 98% Plumbing, 2% Revolution
VerbalVista: Talking to Your Own Data with RAG, FAISS, and a Bit of Stubbornness
Exploring Model Quantization for LLMs
No articles match.
Try a different topic or search term.