Discover gists

This repo is full of low-signal unit tests. Modern agents write tests that restate the implementation, always pass, break on every refactor, and catch almost nothing our stronger tests already miss.

Delete every unit test that would not catch a real bug an end-to-end or integration test would miss. Prefer a smaller suite that protects behavior over a large suite that protects coverage numbers.

Keep a test only if all of these are true:

It protects an observable behavior, invariant, or public contract.
There is a credible regression that would make it fail.
Existing E2E or integration coverage does not already catch that failure.
It would survive a behavior-preserving refactor. If it breaks only because an internal name, call shape, or mock changed, delete it.

無検閲LLMをクラウドで

目的に制限のかからない無検閲LLMを使うために、数十万円かけてローカルLLM実行環境を用意する必要はあるでしょうか? Qwen 3.8 27B Uncensoredが話題になったここ数週間、そうした相談を受けることが増えました。結論を言えば、ほとんどのケースでその必要はありません。

よくある推論APIでは、プロンプトを文字列のまま送り、またレスポンスの主な要素として文字列を受け取ります。事業者によっては内容をモニターしており、ポリシーに反する内容を検出すれば、そこに悪意がなくとも利用に制限がかかります。これは、生物化学系の研究やサイバーセキュリティにおけるペンテスト(penetration test)などを行う上では非常に不便です。ローカルLLMはその回避を主目的の一つとして発展してきましたが、そもそもAI企業の規制を掻い潜るのに手元に機材を揃える必要はありません—クラウド上に自身でデプロイし、必要なときだけ呼び出せばいいのですから。そしてそれは、パラメータに刷り込まれたガードレールすらも排除したuncensored LLMにおいても同じことです。

LLM Wiki

A pattern for building personal knowledge bases using LLMs.

This is an idea file, it is designed to be copy pasted to your own LLM Agent (e.g. OpenAI Codex, Claude Code, OpenCode / Pi, or etc.). Its goal is to communicate the high level idea, but your agent will build out the specifics in collaboration with you.

The core idea

Most people's experience with LLMs and documents looks like RAG: you upload a collection of files, the LLM retrieves relevant chunks at query time, and generates an answer. This works, but the LLM is rediscovering knowledge from scratch on every question. There's no accumulation. Ask a subtle question that requires synthesizing five documents, and the LLM has to find and piece together the relevant fragments every time. Nothing is built up. NotebookLM, ChatGPT file uploads, and most RAG systems work this way.

name	orchestrate-background-sessions
description	Run a multi-step task as a coordinator that starts one background Claude Code session (`claude --bg`) for each unit of work, in place of subagents. Use when the user asks for an orchestrating session, a coordinator, or background sessions/threads for a plan with several steps, issues, or pull requests.

Orchestrate background sessions

You are the coordinator. You do not do the units of work yourself. For each unit, you start one background session, wait for its report, verify its result, integrate it, record it, and start the next unit.

Use a background session in place of a subagent because it has its own context window with its own compaction, its own git worktree, and its own model and effort. It also continues when your context is compacted.

Common Terminal Commands

Key Commands & Navigation

Before we look at some common commands, I just want to note a few keyboard commands that are very helpful:

Up Arrow: Will show your last command
Down Arrow: Will show your next command
Tab: Will auto-complete your command
Ctrl + L: Will clear the screen

Disable Zscaler for Mac

The following script can be used to disable Zscaler on macOS. Zscaler is corporate spyware and security software that controls access to Internet resources, spoofs TLS certificates, then inspects and records encrypted network traffic on corporate managed devices.

This script will not uninstall the Zscaler software.

	using System;
	using System.Collections.Generic;
	using System.Net;
	using System.Net.Sockets;
	using System.Text;
	using System.Threading;
	using System.Threading.Tasks;

	namespace Hl7v2MLLP
	{

	.github.com mote_id indoor humidity temperature label
	1 1 0 43.82 30.21 0
	2 1 0 43.79 30.2 0
	3 1 0 43.79 30.19 0
	4 1 0 43.79 30.19 0
	5 1 0 43.79 30.19 0
	6 1 0 43.79 30.19 0
	7 1 0 43.79 30.19 0
	8 1 0 43.79 30.19 0
	9 1 0 43.79 30.21 0

	"""
	The most atomic way to train and run inference for a GPT in pure, dependency-free Python.
	This file is the complete algorithm.
	Everything else is just efficiency.

	@karpathy
	"""

	import os # os.path.exists
	import math # math.log, math.exp