xiaolai retweeted
Weekend hack: HackerTube. I was annoyed with YouTube's recommendations so I had Claude filter down 43M comments over 20 years of HN to 6,454 videos that keep being recommended. *Significantly* better viewing experience, ~2 hours with Opus 5.5:
hackertube.org/
- Excellent recs without any logins (state is kept in local browser session)
- Thumbnail is a representative frame from middle of the video instead of the super duper A/B test hyper-optimized user uploads
- Underlying data from the comment analysis as one zip file, point your Claude at it and ask for your own version
- HackerTube Longs ✊
A bit more on the process:
I use Claude Code (desktop app for Mac), Opus 5.5 on Medium effort as my default. I only talk to Claude and have it do everything, the only "manual work" I helped with in tandem was in the browser (reminiscent of MenuGen): 1) I had to create some kind of API key from Google Cloud UI to query YouTube for video metadata, 2) I booted up a VPS box for serving and 3) I bought the domain name. Once the box was up I gave Claude the IP and told it update, secure, harden, deploy and optimize the code, and it did the whole thing.
Many awesome projects that would have normally taken weeks/months even can now be few-shotted for breakfast. What else are we not thinking of.
"Do you even talk to real people anymore?"
Actually, human connection matters more than ever (AI just filtered out the noise).
The true magic of AI is breaking past our old limitations. Because of it, I'm now doing things I only dreamed of:
Coding my own apps (instead of relying on others)
Writing books in English (instead of just my native language)
Crafting fiction (instead of just non-fiction)
AI opened the cage. The only question is: where do you want to fly?
Power keeps the control. It gives away the blame.
From the ancient scapegoat to "the algorithm did it," my new book No One Did It names the move — responsibility laundering — and hands you the diagnostic to walk any laundered case back to a name.
amazon.com/dp/B0H3GDDL6H
Made a little typing practice app for the kids. They love it... What they don’t realize is that it’s not the app they love — it’s the clickety-clack of the keyboard. type.review
The Half Second is out today.
The book is about the gap between what happens to you and what you do next — the place willpower never reached, and where identity quietly does the work.
Years in.
amazon.com/dp/B0H1TKTMDD
If you’re a heavy Claude user like me, you probably feel this pain: * Switching between multiple accounts is a headache (CLI/desktop). * The 5h/7d limits are hard to track. * Can’t easily use third-party models.
Rename a project? Tons of “skills”… but which ones actually work? So I built ClauDepot. It’s still rough, but it works well enough for me. 👉 github.com/xiaolai/claudepot… (alpha v0.0.10)
github.com/xiaolai/north-sta…
A 260-token system prompt that overrides three structural presumptions every RLHF-trained instruction-tuned LLM inherits from training: that you want confirmation, that old scarcity still applies, and that best practices are ceilings.
The argument: lixiaolai.com/articles/2026-…
The Claude Agent SDK is pre-1.0 (v0.2.x) — APIs break frequently, functions get renamed, parameters change. A static skill would teach Claude outdated patterns that produce broken code. The pipeline keeps this skill accurate by tracking version bumps, researching new issues, and updating rules daily. Without it, advice written for v0.2.30 silently becomes wrong when the SDK moves to v0.3.0. So I've written this: github.com/xiaolai/claude-ag…
![GitHub - xiaolai/claude-agent-sdk-skill-autoupdated: [ARCHIVED 2026-05-18] Superseded by anthropi...](https://pbs.twimg.com/card_img/2116370790761558016/nMRJ-UEC?format=jpg&name=800x419)
While willpower is unreliable for behavior change, selftalking uses short scripts, carefully crafted (10 principles), spoken in our own voices with heavy repetition, to reshape our first reactions across contexts, rooted in the default network (System 1) and its familiarity heuristic. Against the pull of spin doctors, Selftalking restores our control and our freedom through a quiet identity shift that enables real progress.
Claude wrote shitty codes, and then tried to cheat ESLint: `// eslint-disable-next-line @typescript-eslint/no-explicit-any`
Have to curse at it: Is this some lazy-ass fucking hack? Ain't there tighter, balls-to-the-wall better ways? Get your shit together and think like a goddamn genius—no half-assed bullshit, no pussy shortcuts!
A few days ago, I started writing a Chrome/Edge browser extension (github.com/xiaolai/insidebar…) to solve some of my own pain points — for example, avoiding the hassle of switching between tabs, saving myself hundreds of mouse or touchpad clicks a day, and cutting down on countless copy-paste operations. Most importantly, it gives me a single place to store and search through important chat logs. (Too bad publishing this kind of extension to the store is such a pain.)
With Claude-Code, I revived a little domain-search tool I had loved years ago but that had since disappeared — impossibility.org. It only took me one day. Then I thought: why not just buy the domain and bring it fully back to life? So I did.
