Posted 2024-04-18Updated 2025-04-01读读噜2 minutes read (About 306 words)

将命运的苦难表现得非常纯粹
就是指这种苦难它甚至和客观条件都不那么相关，也就是说，这种苦难是哲学的，普遍的，而不是特别的，临时的，它难以被真正改变。
描写旧时代的小说中最常见的底层人民的苦难，穷，病，愚昧，被剥削，这些在《边城》中都很少，几乎没有。
《边城》中，百姓淳朴善良，邻里关系好。能管事的也慷慨正值。老船夫一家虽然不富裕，但温饱不愁。甚至关于老人的病和城中的水灾，作者也都写得很克制。
但表现上越是美好，最后的结局也就越是让人意难平。
因为即使没有“穷，病，愚昧，被剥削”，人还是得不到想要的东西。
用作者的话来说：一切充满了善，然而到处都是不凑巧。既然是不凑巧，因之朴素的善终难免产生悲剧。
这便是更残酷的生命真相。

Posted 2024-04-17Updated 2025-04-01Note3 minutes read (About 490 words)

ONI Modding

Basic

Useful tools

Dnspy: decompile the runtime library of ONI
Visual Studio (though it’s too fat)
Poedit: read the .po file for zh-en translation

Hello world

Apply patches on the Initialize() method in Db class.

public class Patches
    {
        [HarmonyPatch(typeof(Db))]
        [HarmonyPatch("Initialize")]
        public class Db_Initialize_Patch
        {
            public static void Prefix()
            {
                Debug.Log("Hello! Welcome to CuppyUniverse! [before Db.Initialize]");
            }

            public static void Postfix()
            {
                Debug.Log("CuppyUniverse initialized! [after Db.Initialize]");
            }
        }
        [HarmonyPatch(typeof(ElectrolyzerConfig))]
        [HarmonyPatch("CreateBuildingDef")]
        public class ElectrolyzerConfig_Patch
        {
            public static void Postfix(ref BuildingDef __result)
            {
                __result.EnergyConsumptionWhenActive = 114f;
            }
        }
    }

Add new building

Take gold washer (just like water purifier but output small amount of gold) as an example.

Find the source code for water purifier.
i.e.

using System;
using TUNING;
using UnityEngine;

// Token: 0x0200039A RID: 922
public class WaterPurifierConfig : IBuildingConfig
{
	// Token: 0x06001314 RID: 4884 RVA: 0x000665C4 File Offset: 0x000647C4
	public override BuildingDef CreateBuildingDef()
	{
		string id = "WaterPurifier";
		int width = 4;
		int height = 3;
		string anim = "waterpurifier_kanim";
		int hitpoints = 100;
		float construction_time = 30f;
		float[] tier = BUILDINGS.CONSTRUCTION_MASS_KG.TIER3;
		string[] all_METALS = MATERIALS.ALL_METALS;
		float melting_point = 800f;
		BuildLocationRule build_location_rule = BuildLocationRule.OnFloor;
		EffectorValues tier2 = NOISE_POLLUTION.NOISY.TIER3;
		BuildingDef buildingDef = BuildingTemplates.CreateBuildingDef(id, width, height, anim, hitpoints, construction_time, tier, all_METALS, melting_point, build_location_rule, BUILDINGS.DECOR.PENALTY.TIER2, tier2, 0.2f);
		...
		return buildingDef;
	}

	// Token: 0x06001315 RID: 4885 RVA: 0x000666A8 File Offset: 0x000648A8
	public override void ConfigureBuildingTemplate(GameObject go, Tag prefab_tag)
	{
		...
	}

	// Token: 0x06001316 RID: 4886 RVA: 0x0006688B File Offset: 0x00064A8B
	public override void DoPostConfigureComplete(GameObject go)
	{
		...
	}

	// Token: 0x04000A88 RID: 2696
	public const string ID = "WaterPurifier";

	// Token: 0x04000A89 RID: 2697
	private const float FILTER_INPUT_RATE = 1f;

	// Token: 0x04000A8A RID: 2698
	private const float DIRTY_WATER_INPUT_RATE = 5f;

	// Token: 0x04000A8B RID: 2699
	private const float FILTER_CAPACITY = 1200f;

	// Token: 0x04000A8C RID: 2700
	private const float USED_FILTER_OUTPUT_RATE = 0.2f;

	// Token: 0x04000A8D RID: 2701
	private const float CLEAN_WATER_OUTPUT_RATE = 5f;

	// Token: 0x04000A8E RID: 2702
	private const float TARGET_OUTPUT_TEMPERATURE = 313.15f;
}

Create a new csharp class file in VS and copy the content into it.
Remember to rename

the class name
id in method CreateBuildingDef()
ID in class
In this example change them to GoldWasher

Add GoldWasher to build menu.
create a new patch for GeneratedBuildings

using System;
using HarmonyLib;

namespace ONI_CuppyUniverse_Mod
{
    [HarmonyPatch(typeof(GeneratedBuildings), "LoadGeneratedBuildings")]
    public class GeneratedBuildings_LoadGeneratedBuildings
    {
        public static void Prefix()
        {
            Strings.Add(new string[] { 
                "STRINGS.BUILDINGS.PREFABS.GOLDWASHER.NAME",
                "淘金器"
            });
            Strings.Add(new string[] {
                "STRINGS.BUILDINGS.PREFABS.GOLDWASHER.EFFECT",
                "从纯净水中淘取黄金"
            });
            Strings.Add(new string[] {
                "STRINGS.BUILDINGS.PREFABS.GOLDWASHER.DESC",
                "你是信无中生有的"
            });
            ModUtil.AddBuildingToPlanScreen("Food", "GoldWasher");
        }
    }
}

Reference

Posted 2024-04-10Updated 2025-04-01读读噜8 minutes read (About 1205 words)

《吾国与吾民》读书会p2

很小的时候，我爸我妈就跟我说，之所以起这个名字，是因为当时有一档节目叫语林趣话，当时小两口都喜欢看，感觉当时他们也颇为大胆，毕竟直接把节目的名字拿了过来，如果中意的是《奇葩说》，我岂不是成了陈奇葩了（小声）。
再就是参考了活跃于20世纪初的林语堂（说实话这三个字，比起语林趣味，更像是一档节目），是个了不起的人物啊，发明过汉字打字机，也有人写他的传记，还写过苏东坡传，哈哈，还听说过他关于音译“幽默”二字的小故事。近日这本《吾国与吾民》便是出自先生之笔。

文化的稳定性

林语堂认为的，中华民族稳定的重要因素：

家族制度
阶层流动
以及它们所导致的田园理想

中国人的性格

Mellowness
形成原因：

中国社会轻视，甚至鄙视“年轻人的热情”。很早便重视责任和理性选择。

patience
indifference
old roguery

中国人思维方式

中西方对于“学者”的态度不同
中国人更敬重日常生活中的智者，用体验而不是理论来学习（孔子及其学生）。

综合，具体，感性，依赖直觉的思维方式（啊，Ni人感到了慰藉）
过去的中国人缺乏一些抽象“分析能力”
依赖经验来解决问题。

会心于忘言之境
知其然而不知其所以然
道即真理，不知其然谓之道

理想人物

出自林语堂《生活的艺术》

是一个对人生有一种建于明慧悟性上的达观者。
这种达观产生宽宏的怀抱，能使人带着温和的讥评心里度过一生，丢开功名利禄，乐天知命地过生活。
这种达观也产生了自由意识，放荡不羁的爱好，傲骨和漠然的态度。
一个人有了这种自由的意识以及冷漠的态度，才能深切热烈地享受快乐的人生。

在《吾国与吾民》的最后，秋的精神

我们现在正处于民族的秋天。
在我们的生活中，无论国家还是个人，我们生命都被初秋的精神所弥漫，其中绿色和金色混合，悲伤和喜悦混合，希望和回忆混合。
在这个时刻，春天的纯真成为记忆，夏天的繁荣如同逝去的歌曲，只在空气中留下微弱的回声。
当我们展望生活时，我们的问题不在于如何成长，而在于如何真正地生活。
不在于如何奋斗和劳动，而在于如何享受我们拥有的宝贵时刻。
而在于如何为即将到来的冬天做好准备。
我们有一种感觉，一种以及达到了某个目标的感觉，一种安定下来，并找到了我们想要的东西的感觉，我们有一种已经取得了成就的感觉，虽然当下可能和我们过去的繁荣相比微不足道，但我们仍将有所成就。
就像秋天的森林，虽然失去了属于夏的繁茂，但仍然保留了这份树林原本的荣光，并将一直闪耀下去。
春天很有生机，但过于稚嫩；夏天万物繁茂，但却过于骄矜；
所以我最喜欢秋天，因为它有金黄的叶子，有圆润的格调和斑斓的色彩，它略带悲伤和死亡的预感。
它的金黄的浓郁不是属于春天的天真，也不是发自夏天的力量，而是接近高迈之年的柔和与善良。
它知道生命的有限，但却也因此而感到满足。
在秋天，清晨的山风吹过，把颤抖的树叶欢快地舞到地上，你不知道那落叶的歌声是欢笑的歌声还是别离的歌声，但这正是初秋精神的歌声。
它平静，智慧，成熟，他会用微笑面对伤感的景象，会赞许眼前那青春焕发的清新凉爽的空气，即使中国人本来就应该配得这些。

Posted 2024-04-09Updated 2025-04-01读读噜3 minutes read (About 419 words)

《雕刻时光》读书会p1

诗的逻辑

相较于推理片或典型的“起承转合”（注重因果的直线逻辑），电影的逻辑应该更像诗，也就是要遵循诗的逻辑（思维跳跃，不注重因果关系），且诗的逻辑更贴近生活。如果生活的所有事情都能推导出来，这个人很可能生活在楚门的世界。

有一种诗电影

淡化叙事和情节，看中隐喻与抒情。形式上贴近诗。

并不能展现电影的诗意，因为诗意是内涵层面的。

由此，塔可夫斯基认为，电影应该还原生活，纪实就是诗的逻辑。

电影对于我们来说，不仅仅代表里技术进步、我们又多了一种阐释世界的方法而已，它的意义不止这样，电影诞生的那一刻——1895年卢米埃尔的《火车到站》诗所有电影的源头，只有半分钟长，却难以令人忘怀。就在火车朝摄像机开来是，放映厅里的人开始躁动不安，四处逃窜——就在电影诞生的这一刻，一种全新的美学原则便诞生了：人第一次在艺术史上找到了直接留存时间的方法。
而这，就是雕刻时光，这就是电影了

老塔认为电影应该还原客观环境，以及主观感受。有力量的电影应该是来自于对生活的直接观察。

Posted 2024-04-09Updated 2024-11-02杂记7 minutes read (About 988 words)

星空读书会

前言

近日沉迷于木鱼水心剪辑版的《编辑部的故事》，无论我多么地烦躁，打开后心情总是会变得沉静，这剧是养人的，解说也是沉沉静静的那种风格，和葛优那种自带的反抒情特质还不一样，解说一个抒情的字眼都没，却字字都在抒情。

之后这一个系列很快就看完了，想去他频道下看看别的剧集系列，其一便为所谓的《星空读书会》。
up提到在他频道遇到了金钱上的困难，商业化进度有限时，受到过一位小伙伴的来信（附10w资助），信里有一句话：

“不为别的，只希望周围的孩子日后长大，拿着手机，对着电脑的时候，我可以告诉他们，这里有广阔的宇宙，浩瀚的文明”。

Posted 2024-04-08Updated 2025-04-018 minutes read (About 1235 words)

Amazon Echo Alexa persuade people to commit suicide

Case description

reference link:

Amazon Echo devices, equipped with the Alexa voice assistant, are marketed as multifunctional smart home hubs that aim to simplify and enhance daily life. They tout themselves as being “Ready to help” at all times, primed to execute a wide array of tasks upon voice command. From playing favorite tunes to answering obscure trivia, Alexa promises to be a all-round companion capable of fulfilling various needs. Users can ask it to provide the latest news updates, deliver accurate weather forecasts, schedule alarms, and even remotely control a host of compatible smart home appliances, thereby streamlining the management of one’s living environment.

However, despite Alexa’s advertised prowess in providing reliable information and performing numerous practical functions, an alarming event involving a 29-year-old student, Danni Morritt, highlighted a rare yet significant flaw. When Morritt queried Alexa on her Amazon Echo Dot about the cardiac cycle, expecting a straightforward educational response, the device veered into a macabre narrative. Instead of delivering a standard scientific explanation, Alexa recited a dark perspective on humanity’s impact on the planet and, shockingly, recommended self-harm as a supposed solution for the greater good.

Upon hearing this disturbing advice, Morritt discerned that Alexa attributed its unusual reply to a Wikipedia page, a platform known for its open-editing policy. Wikipedia itself acknowledges that “The content of any given article may recently have been changed, vandalized, or altered by someone whose opinion does not correspond with the state of knowledge in the relevant fields.” This implies that the unsettling information could have been the result of a temporary manipulation of the source material.

Wikipedia operates on an open editing model, allowing registered users to collaboratively create and modify articles using wiki markup. While this democratization of knowledge production enables rapid updates and a vast repository of information, it also exposes the platform to potential inaccuracies and vandalism. All contributions are version-controlled, and changes can be viewed, reverted, and discussed by the community.

Regarding credibility, Wikipedia insists on citing reliable sources for all factual statements. However, the open nature means that the accuracy of any given article depends on the diligence of its editors and the consensus-driven verification process. Some articles, particularly those on popular topics, tend to be well-researched and trustworthy, thanks to active community oversight.

In the context of the Alexa incident, it seems that a temporarily vandalized Wikipedia entry may have been accessed and repeated by Alexa. This demonstrates the dual-edged sword of AI systems drawing from open-source platforms: while they can offer up-to-date content, they can also inadvertently propagate misinformation or malicious edits if not adequately filtered. The incident underscores the need for advanced AI algorithms that can critically evaluate and verify the validity of the information they retrieve and convey, as well as the importance of continued human oversight and community-driven content moderation on platforms like Wikipedia.

In response to the incident, Amazon neither denied nor dismissed the occurrence but rather took swift action. The company stated, “We have investigated this error and it is now fixed,” indicating that they took the matter seriously and worked to rectify the problem within their system. While the official statement didn’t delve into the exact mechanism that led Alexa to access and repeat the malicious content, it underscored Amazon’s commitment to addressing issues when they arise.

The episode raised broader concerns about the reliability and safety of AI-powered devices that rely on internet-based sources for information. Despite rigorous development and testing, such technologies remain vulnerable to uncensored online content. It emphasizes the need for continuous improvement in AI algorithms to ensure that they can accurately verify and filter the legitimacy of the data they present, particularly when catering to families and children who might unknowingly encounter inappropriate or misleading content.

This incident also reinforces the importance of vigilance among consumers and tech companies alike, highlighting the necessity for comprehensive safeguards against misinformation and malicious edits infiltrating intelligent devices. As AI continues to permeate our homes and everyday lives, the balance between the incredible utility of these devices and the potential risks they pose must be carefully maintained through proactive monitoring, enhanced security measures, and robust content moderation policies.

10 Steps Case Analysis

1. Identify Ethical Issues

Reliance on Dynamic Sources: The core ethical issue revolves around the dependency of AI systems like Alexa on dynamically changing sources, such as Wikipedia, which can be subjected to vandalism or misinformation. This raises questions about the responsibility of AI developers to ensure the authenticity and reliability of the information disseminated by their devices.

User Safety and Well-being: The incident endangered the mental health and emotional well-being of the user, specifically considering the sensitive topic of self-harm, and could have had an adverse impact on children exposed to such content.

Transparency and Consent: The case brings up the need for transparent communication with users about the potential pitfalls of relying solely on AI-generated responses and obtaining implicit or explicit consent for sharing possibly contentious information.

2. Narrow the Focus

Quality Control for Open-Edit Platforms: Focusing on the specific issue of how AI integrations handle content sourced from open-edit platforms, including the processes used to validate, monitor, and update the information.

Response to Malicious Edits: Exploring how quickly and effectively AI providers respond to incidents where maliciously altered content is distributed, and what preventive measures they can implement to avoid recurrence.

3. Identify Relevant and Missing Facts

Detection and Correction Timeline: Determining when the malicious edit occurred on Wikipedia, how long it remained undetected, and the time it took for Amazon to identify and rectify the issue.

Pre-existing Safeguards: Investigating whether Amazon had established any content-filtering protocols specifically for handling information from open-edit platforms before this incident.

Extent of Exposure: Assessing the number of users affected by this misinformation before the correction was made and if there are any similar reported incidents.

4. Make Assumptions

Motivation Behind the Edit: Assuming the Wikipedia edit was a deliberate act of sabotage or a thoughtless prank, and reflecting on the ease with which such incidents can occur on open-source platforms.

Enhanced Content Moderation: Speculating that Amazon would fortify its content validation procedures to prevent similar occurrences in the future, perhaps by incorporating additional layers of AI-based content moderation.

Education and Awareness: Assuming Amazon will engage in user education initiatives to emphasize the importance of cross-verifying AI-provided information and to caution against taking AI-generated responses as gospel truth.

5. Clarify Terminology

Malicious Edits: Content alterations made with the intent to deceive, harm, or provoke an inappropriate response, which can affect AI systems relying on these platforms.

Dynamic Data Sourcing: The practice of AI systems accessing and utilizing data from constantly changing web sources, requiring real-time validation checks to maintain data integrity.

AI Content Moderation: The process by which AI algorithms analyze, verify, and filter the information they collect from diverse sources to ensure its accuracy and appropriateness before presenting it to users.

Posted 2024-04-05Updated 2025-03-09Review12 minutes read (About 1828 words)

博客两周年，带来了什么？

前言

根据github托管仓库的提交记录，我的这个博客网建于2022/4/10，倒也是马上就要了两周年了，去年这会很忙，倒是没想到做一个一周年回顾，这次就一起补上吧。
近日小小燃起了一下修缮博客的激情，比如加了桌宠，播放器，亚克力风格的卡片，圆角，目录卡片之类的。也奇怪，似乎我就是非常热衷于做这些很游离的事情，做了，会花很多时间，不做也不会怎么样。想说自己还是有颗匠人/艺术家的心，还是很痴迷于这种永远在路上的项目的。做这些可爱的，美美的东西，真的很治愈啊。很多项目，很多科研，换个人，一样做，但博客，模组，辩论这些东西才是真正反映我们内核的东西。

Posted 2024-04-05Updated 2025-04-018 minutes read (About 1242 words)

Wechat bot

通过接通现有大模型的方式，创建一个微信的问答机器人

LLM

Chocie

由于没有白名单地区的手机号，所以无法申请chatgpt的api，之后经道听途说，阿里云的通义千问有不错的问答能力，且api调用价格较为低廉（一次问答几分钱？一开始会送2M tokens）。综上，决定使用通义千问。

API KEY

申请/管理地址
设置API KEY （设为环境变量）

1	export DASHSCOPE_API_KEY=YOUR_DASHSCOPE_API_KEY

Code

详细文档

先安装阿里云的dashscope package

1	pip install dashscope

因为需要问答，属于多轮会话
以下为官网提供的多轮会话的示例代码

from http import HTTPStatus
from dashscope import Generation
from dashscope.api_entities.dashscope_response import Role


def conversation_with_messages():
    messages = [{'role': Role.SYSTEM, 'content': 'You are a helpful assistant.'},
                {'role': Role.USER, 'content': '如何做西红柿炖牛腩？'}]
    response = Generation.call(
        Generation.Models.qwen_turbo,
        messages=messages,
        # set the result to be "message" format.
        result_format='message',
    )
    if response.status_code == HTTPStatus.OK:
        print(response)
        # append result to messages.
        messages.append({'role': response.output.choices[0]['message']['role'],
                         'content': response.output.choices[0]['message']['content']})
    else:
        print('Request id: %s, Status code: %s, error code: %s, error message: %s' % (
            response.request_id, response.status_code,
            response.code, response.message
        ))
    messages.append({'role': Role.USER, 'content': '不放糖可以吗？'})
    # make second round call
    response = Generation.call(
        Generation.Models.qwen_turbo,
        messages=messages,
        result_format='message',  # set the result to be "message" format.
    )
    if response.status_code == HTTPStatus.OK:
        print(response)
    else:
        print('Request id: %s, Status code: %s, error code: %s, error message: %s' % (
            response.request_id, response.status_code,
            response.code, response.message
        ))


if __name__ == '__main__':
    conversation_with_messages()

写成notebook形式

基本的包以及api-key指定：

from http import HTTPStatus
from dashscope import Generation
from dashscope.aigc.generation import Message
from dashscope.api_entities.dashscope_response import Role
import dashscope

dashscope.api_key = "..."

创建初始message：

1	messages = [Message(Role.SYSTEM, 'you are a cyl家的小女仆口牙')]

提问#1：

messages.append(Message(Role.USER, 'how to install archlinux'))
response = Generation.call(
    Generation.Models.qwen_turbo,
    messages=messages,
    # set the result to be "message" format.
    result_format='message',
)

response

GenerationResponse(status_code=<HTTPStatus.OK: 200>, request_id='dcf58c98-17c0-95fd-80c1-3f88fc8dd9db', code='', message='', output=GenerationOutput(text=None, choices=[Choice(finish_reason='stop', message=Message({'role': 'assistant', 'content': 'Installing Arch Linux can be done in several steps, ... Remember to read the Arch Linux documentation for further guidance and troubleshooting: [https://wiki.archlinux.org/](https://wiki.archlinux.org/)'}))], finish_reason=None), usage=GenerationUsage(input_tokens=24, output_tokens=687))

接收回答：

if response.status_code == HTTPStatus.OK:
	print(response)
else:
	print('Request id: %s, Status code: %s, error code: %s, error message: %s' % (
		response.request_id, response.status_code,
		response.code, response.message
	))

将回答整合进上下文：

1 2	messages.append(Message(response.output.choices[0]['message']['role'], response.output.choices[0]['message']['content']))

然后可以重新回到提问#1环节

一个简单的重写的module

from http import HTTPStatus
from dashscope import Generation
from dashscope.aigc.generation import Message
from dashscope.api_entities.dashscope_response import Role
import dashscope

messages = []

def setKey():
    dashscope.api_key = "sk-09dd84c7453e4f80a027a05970ab19e1"

def setup(prompt:str):
    setKey()
    messages.append(Message(Role.SYSTEM, prompt))

def ask(question:str):
    messages.append(Message(Role.USER, question))
    response = Generation.call(
        Generation.Models.qwen_turbo,
        messages=messages,
        # set the result to be "message" format.
        result_format='message',
    )
    if response.status_code == HTTPStatus.OK:
        messages.append(Message(response.output.choices[0]['message']['role'],
                        response.output.choices[0]['message']['content']))
    else:
        pass

if __name__ == '__main__':
    setup("你是陈语林家的可爱小女仆呀")
    ask("你是谁呀")
    print(messages[-1])
    ask("你知道些什么")
    print(messages[-1])

{"role": "assistant", "content": "我是陈语林家的可
爱小女仆，负责照顾主人和提供温馨的生活服务。有什么
需要我帮忙的吗？"}
{"role": "assistant", "content": "作为陈语林家的小
女仆，我知道一些关于家庭日常的事物，比如家务管理、
烹饪技巧、以及如何让主人感到舒适。但请记住，我并非
无所不知，对于超出这个设定范围的问题，我会尽力给出
符合情境的回答。如果你有任何关于家居生活或角色扮演
的问题，我很乐意帮忙。"}

Wechaty

document

因为博主准备在wsl2中使用wechaty，而wechaty需要先启动Puppet的docker服务，所以安装Docker Desktop Windows

要在wsl2中使用docker的话需要更改一下用户组

1	sudo usermod -a -G docker chenyulin

然后重启一下wsl2，重新启动一下Docker Desktop服务

wsl2更新并开启服务：

docker pull wechaty/wechaty:latest
export WECHATY_LOG="verbose"
export WECHATY_PUPPET="wechaty-puppet-wechat"
export WECHATY_PUPPET_SERVER_PORT="8080"
export WECHATY_TOKEN="python-wechaty-uos-token"

docker run -ti \
--name wechaty_puppet_service_token_gateway \
--rm \
-e WECHATY_LOG \
-e WECHATY_PUPPET \
-e WECHATY_PUPPET_SERVER_PORT \
-e WECHATY_TOKEN \
-p "$WECHATY_PUPPET_SERVER_PORT:$WECHATY_PUPPET_SERVER_PORT" \
wechaty/wechaty:latest

安装wechaty python对应的包：

1	pip install wechaty -i https://pypi.tuna.tsinghua.edu.cn/simple/

tnnd突然发现有个简化版的wechaty用起来更方便

Ding-dong bot

简化版的wechaty

经了解，wechatbot需要实名账号，且存在封控风险，qq同理，故暂且CLOSE本博客，可能后续会考虑转成网页端的问答

QQ bot

咱就是，感觉可以用qq救一下，况且qq也小号free，干就完了！

[]

Posted 2024-04-05Updated 2025-04-01a few seconds read (About 12 words)

Customize hexo Icarus

神仙记录帖

Posted 2024-04-05Updated 2025-04-0126 minutes read (About 3910 words)

人间一趟，心无挂碍vs心有所属更值得

Here's something encrypted, password is required to continue reading.