About This Blog

Introduction

This is a tech blog where istb16, who runs tk3.biz, writes as a software engineer about things I’ve tried and things I’ve built myself.

I started it in January 2026. As I kept writing, the topics became clearer, so I rewrote this page.

What I Write About

Local LLMs

I write about getting local LLMs to run at a usable speed on a mini PC with a Ryzen AI 9 HX 370 (Radeon 890M), measuring each setting along the way.

• Start here: The Complete Guide to Running a Local LLM at Usable Speed on a Ryzen AI Mini PC (2026 Edition)
• What each setting does: Local LLM Settings That Work and Settings That Don’t (with Measurements)
• How to measure: How to Benchmark a Local LLM (with a Checklist)

Development and Code Review with AI

I write about how I use AIs such as Claude Code, Codex, and Antigravity for development, and how to have one AI review code that another AI wrote. I built and released dev-orchestra, a plugin that runs reviews by several AIs automatically, and there’s a series about it.

• I Had Three AIs Review 1418 Lines I Wrote, and They Found 11 Real Bugs
• Running Multiple AI Reviews by Hand Was Too Tedious, So I Built a Plugin That Automates It
• I Released dev-orchestra: Design, Implementation, and Multi-AI Review from a Single Request

Setting Up Windows

Tools I installed for developing on Windows, and fixes for problems I ran into on a Japanese-language setup.

• I Installed Coreutils on Windows
• Launching Windows Terminal by Typing “terminal” on Japanese Windows 11

Apps I’ve Built

I introduce apps I’ve built, along with why I built them. You’ll find the others in the project list on the home page.

• I Built JAPAN RISK MAP
• I Built a Chrome Extension: Switch Browser Data

Rules I Follow When Writing

• I only write commands, settings, and numbers that I’ve actually run and checked. If I haven’t tried something, I say so
• When I measured something the wrong way or something didn’t work, I write that up as it happened (for example: I Measured My Local LLM 13 Times, and the Measurements Were Wrong 19 Times)
• Every post has an English version. You can switch with EN at the top right
• Technical posts are also republished on Qiita. The originals are on this blog

Summary

• I write about local LLMs, development and code review with AI, setting up Windows, and apps I’ve built
• What I write is what I’ve run and checked myself
• For comments on posts or other inquiries, send a DM on X (@istb16)