The Hidden Cost of AI Agents: Tokens, Tools, Retries, and Latency
AI agents look simple at first. You take a model, add a prompt, maybe connect a tool, and it works....
Tutorials
Your AI Coding Assistant Isn't Reading Your Code, It's Mailing It Home
Hook An AI coding CLI that uploads your entire Git history — commit logs, secrets, and all...
Your AI Agent's Memory Is Now an Attack Surface, and Nobody Designed for That
Your AI Agent's Memory Is Now an Attack Surface, and Nobody Designed for That One email....
What's Actually Cooking in Azure AI Right Now (Notes from My MVP Prep Rabbit Hole)
So here's the thing — I've been buried in Azure documentation for the past few weeks, partly because...
Matrix Chess Club Bot
See my thinking index.html html Matrix Chess Club Bot ...
The Hard Part of Test Automation Isn’t Writing Tests Anymore
A few years ago, creating the automated test was usually the difficult part. You had to choose a...
The cheap open models are out there. Actually using them is the annoying part.
Open-weight models quietly got good this year — good enough that for most real engineering work I...
Model Showdown Round 9: Qwen 3.6 27B vs Qwen 3.6 35B-A3B vs Qwythos-9B vs GLM-4.7-Flash vs Nemotron-3-Nano
I put Qwen 3.6 27B, Qwen 3.6 35B-A3B, Qwythos-9B, GLM-4.7-Flash, and Nemotron-3-Nano through the same real coding task on my homelab RTX 5090. Along the way I had to live-patch two...
Beyond Scaling Laws: Why "Thinking Longer" Is a Systems Problem, Not a Prompting Trick
For five years, the answer to "how do we make the model better" was always the same: bigger model,...
How I Built an MLP Evaluator for a Novel Board Game
When you build a chess engine, you stand on the shoulders of giants. Piece values were tuned decades...
Access to Claude in China sells for 70-90% below the official API price
A 90% discount doesn't come from nowhere. Someone pays for it, and usually that someone is you: with...
Your Codex model shuts off July 23 — a 7-day migration map
I pin model IDs on purpose. Floating aliases have burned me before — a silent swap under a -latest...
We Checked Whether On-Site SEO Predicts AI Citations. The Data Says Mostly No.
Every GEO ("generative engine optimization") tool, including ours until recently, sells some version...
Reverse-engineering an MMO Aion 2's network protocol to build a real-time DPS meter (Rust + Tauri)
Disclosure: this is a write-up about my own side project — a combat analytics tool for AION 2. No...
Running an LLM agent entirely in your browser
A tiny LFM2.5 fine-tune that drives a storefront entirely in-browser with tool calls and RAG - trained on interaction patterns, not domain facts
Lattice: The Pure Rust LLM Engine Optimized for Apple Silicon
Native Local AI on Mac: Meet Lattice Running Large Language Models (LLMs) locally has...
Have You Outgrown Prompt Engineering?
For a while, prompt engineering felt like a search for the perfect phrase. Give the model a role....
My Demo Script Found a Production Bug on Its First Run: A Tiny Post-Mortem
How a narrated demo for an MCP image server caught an API-contract drift that 10 passing unit tests never could — root cause, fix, and lessons about mocked tests.
Build One AI Tool Server, Call It From Three Different Agents (MCP Explained)
A beginner-friendly tour of the Model Context Protocol: one Python server that generates images with Gemini, called by Claude Code, a Google ADK agent, and a Rust CLI.
SQL Injection in AI-Generated Code: Why Cursor Keeps Writing It
TL;DR AI editors love string interpolation for SQL. That is exactly how injection gets...
I audited my own AI-generated refactor and found 46 bugs. Here's what that taught me.
How a 1,920-line file turned into 410 lines, why "it works" was a lie twice, and why the fix wasn't a...
We audited our own "fail closed by default" claim. Here's what failed open.
SidClaw is a governance layer for AI agents. The whole pitch fits in four words: fail closed by...
Stop Guessing Your AI Performance: The Professional Guide to Edge AI Benchmarking on Android
If you have ever deployed a Large Language Model (LLM) or a complex computer vision model to an...
We Built a Super-App in 30 Days. Here's the Chaos Nobody Saw.
One team. Thirty days. Nine service verticals. Fifty-plus screens. And one question that kept us up...