146 posts
A personal blog that started in the grassroots webmaster era of 2012 survived tightening regulation, the app wave and the collapse of search traffic — and finally lost its reason to exist in the age of AI.
Wire a local LLM into IntelliJ IDEA with LM Studio, LM Link and the Continue plugin — an AI coding setup that needs no API key. Runs Qwen3.5-27B over CUDA on an RTX 2080 Ti, so the tokens are all yours.
After 1Password raised prices by 33%, I nearly switched to Bitwarden — but one critical problem made me stay. This post compares the mainstream password managers, analyzes the risks of Passkey lock-in, and shares how to boost security with a YubiKey and successfully negotiate a 25% renewal discount.
Can you still mine with a GPU in 2025? What are the returns? I recently tried mining with an NVIDIA RTX 2080 Ti, and after a few days of mining I'm here to break down whether it's still viable.
I quit smoking on March 25, 2025. As of this writing I've been smoke-free for 158 days — roughly an 83% success rate (17% chance of relapse). Here I share both my failures and what finally worked.
Microsoft never publicly released a VL edition of Office 2019, so the standard edition has to be converted to VL. I could never remember the steps, so this post is my operating notes for the next reinstall. If you spot a mistake, contact me.
For the past few days my Aliyun CDN has been alerting constantly — traffic and request counts suddenly spike, triggering automatic shutdown. After analyzing the access logs, I found the requests mainly came from IPs in the following areas, all hitting a single image on my CDN.
With open source LLMs and the tools the community keeps shipping, AI is finally within everyone's reach. Plenty of people can now run large models on a local laptop — but laptops are underpowered, and they have other work to do; you can't keep a model running on them around the clock. I prefer PVE, so here's my deployment walkthrough, a guide for my future re-deploys.
If you work in IT, you know the 'e-waste scalpers' crowd. I'm one of them — used enterprise hardware is cheap and performs well, and nothing epitomizes it like the Xeon E3/E5 lines: multi-core, multi-thread CPUs for a few dozen yuan, sometimes just a few. But by the law of board-and-CPU pricing, the cheaper the CPU, the pricier the motherboard. So I put together a rough compatibility table of Intel chipset generations, the CPU generations they take, and the DDR memory generations they support, for my future junk-picking reference.
Recently, Chinese developers logging into GitLab Global (gitlab.com) were shown a notice titled 'Important Change to Your GitLab.com Account'. The gist: individuals and organizations in Mainland China, Macao and Hong Kong need to migrate their accounts to GitLab China (JiHu) within 60 days, after which the accounts will be deleted.
