Ying Xiang geminixiang

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.

This is an OPML version of the HN Popularity Contest results for 2025, for importing into RSS feed readers.

Plug: if you want to find content related to your interests from thousands of obscure blogs and noisy sources like HN Newest, check out Scour. It's a free, personalized content feed I work on where you define your interests in your own words and it ranks content based on how closely related it is to those topics.

欸昨天貼了 claude meetup taipei 心得裏面提到 2026 大趨勢是用 skills 取代 mcp 結果引來許多疑問，趁早上踩飛輪時就順手寫了篇快速解釋希望對大家有幫助！

先說重點

context engineering 是 agent 開發最重要的事，其中關鍵就是防止 context 快速肥大導致 llm 失憶或注意力潰散，導致任務執行失敗。
由 mcp 升級 skills 的主要好處就是解決上述問題，讓你的 agent 變的更聰明能幹少出錯。

下面是俱體細節👇

speechSynthesis強制使用Chrome中的Google小姐中文語音

網路上的window.speechSynthesis教學主要都只有說切換指定語言

像是這樣就能切換成中文語音:

const synth = window.speechSynthesis;

const speak = (msg) => {

市區公車及公路客運API動靜態資料

動態資料傳遞延遲

公路客運六都市區公車動態資料延遲時間(來源端—PTX平台)：

	動態資料延遲時間
公總	即時推播，約2~3秒
台北市	約10秒

	from typing import Union, List
	from dataclasses import dataclass


	class Validations:
	def __post_init__(self):
	"""Run validation methods if declared.

	The validation method can be a simple check
	that raises ValueError or a transformation to

	# -----------------------------
	# USAGE
	# -----------------------------
	# chmod +x mrcnn_continous_run.sh
	# source mrcnn_continous_run.sh
	#
	# DESCRIPTION:
	# Allows matterport's MRCNN to run in sections so that every few epochs it re-executes
	# and, therefore, resets the memory solving provisionally the memory leak issue.
	#

	"""
	Mask R-CNN
	The main Mask R-CNN model implemenetation.

	Copyright (c) 2017 Matterport, Inc.
	Licensed under the MIT License (see LICENSE for details)
	Written by Waleed Abdulla
	"""

	# change related API to make this file run mark rcnn demo competible

	#! /bin/bash

	# NEWLY ADDED BACKUP FUNCTIONALITY IS NOT FULLY TESTED YET, USE WITH CARE, ESPECIALLY DELETION
	# Developed for DSM 6 - 7.0.1. Not tested on other versions.
	# Steps to install
	# Save this script in one of your shares
	# Edit it according to your requirements
	# Backup /usr/syno/share/nginx/ as follows:
	# # cd /usr/syno/share/
	# # tar cvf ~/nginx.tar nginx

	{0: u'__background__',
	1: u'person',
	2: u'bicycle',
	3: u'car',
	4: u'motorcycle',
	5: u'airplane',
	6: u'bus',
	7: u'train',
	8: u'truck',
	9: u'boat',