Sida Chen

It is year 2026 and Copilot still refuses to be half as decent as real LLMs at generating text

Seeing their latest blog post, I at least partially take it back. These people do care about security, and probably for the good of all of us.

Anthropic does great research, makes great products, and has awesome aesthetics in their UI design and model names. It's such a shame that such a company could be run by total assholes.

I really like cleaning up the pile of crap on my computer once in a while. Recently, I ran a complete audit of every software and program I have installed, and compiled them into a list: joshcena.com/notes/tech-r.... This is going to grant me eternal peace of mind

Après moltbook, voici molthub, un site de p*rn pour les LLM. Vous n'êtes toujours pas prêts.

(je crois que tout est devenu totalement hors de contrôle 🤣)

moithub.com

"What is the next number in the series? 1, w, 4, 8, 16..." Choices: 32; 31; 30; not enough data

Midwit move: "Lagrange interpolation" "floored polynomial" yada yada

Try instead: search it on OEIS

In this case the obvious correct answer is 31, for the maximal number of regions formed by joining n points around a circle. Anyone past elementary school should know that by heart

I still browse X occasionally because it's like alcohol/smoke: a little and it amuses you; too much and it poisons your brain

Bsky is the only platform where I don't find my brain poisoned, at the cost of it being absolutely mundane. Oh I need those midwits even if it's engagement bait

FYI the code snippet doesn't use the method at all

Not going to name call, but there's a language in TIOBE top 10 (January 2026), that has *all* of `stopifnot`, `setNames`, `data.frame`, *and* `seq_len` in the same standard library.

A language doesn't need to be Haskell-pilled to be enjoyable though. Rust, being a "C-like", FP-inspired language, brings me far more joy than most other languages. Ultimately I think the syntax, libraries, and semantics of OCaml have all failed me.

I don't write enough Ruby to form an opinion, but I write Python daily so it's grown on me. I think both are fair. Some of my problems with OCaml: small stdlib; verbose syntax (and not in a useful way IMO); impure things aren't stashed away; declaration order matters. Just death by a thousand cuts

I've been writing little OCaml programs in and out for a year, and it's still on the "disgusting" side of the spectrum for me. I don't know how a language managed to combine the unwieldy parts of FP *and* C. (For reference Haskell and Rust are two of my favorite languages)

I also don't think you have the correct responsibility analysis here. Are you aware that require-yield is recommended by ESLint core, or that require-await is also a core rule, or that, in general, linters aren't just to "catch bugs"?

That's a very skewed depiction of the rule. It requires async to be licensed by an await, the same way require-yield (which is a core rule) requires generators to be licensed by a yield, or no-extraneous-calss requires classes to be licensed by an instance property.

Sign that says "maximum permitted occupancy of this space shall not exceed 73 persons"

So we are putting a maximum on a maximum now

Today I realized this is an a and not an o

Bun is great
Until it OOMs

Bon courage !! Ton français rattrapera très rapidement celui de tes contemporains dès que tu seras là-bas 😃

This is a truly awesome post!

Screenshot of the HTML spec. Text: An end tag whose tag name is "sarcasm" Take a deep breath, then act as described in the "any other end tag" entry below.

The HTML spec keeps on giving

html.spec.whatwg.org/multipage/pa...

Canny machine-translates post content (written by user) to the browser's preferred language, but the UI labels remain in English—has to be the funniest l10n strategy

I don't remember if this is my first time (maybe Docusaurus 2 too?), but it's good to be involved in a news that stayed in HN#1 for so long!
(Interestingly, exactly 10 out of the 300+ comments ever mentioned or linked MDN.)

It will be in Firefox 135 behind a flag, and other browsers are still working on it. If all things go well it should be GA late this year.

Screenshot of MDN showing the landing page of Temporal

The Temporal docs are now live on MDN! After watching it evolve for ~5 years I finally sat down and wrote docs for it—one of the biggest additions to the JS docs area ever.

Isn't that what Yarn PnP does?

Screenshot of a GitHub PR page with title "Reference for stage 3 temporal" to the mdn/content repository. It has 247 files changed and 11,492 lines added.

And it starts cooking...👀

A beautifully rendered error message from LALRPOP demonstrating a particular parse path that leads to ambiguity, with an ASCII diagram showing two possible derivations.

We don't show enough appreciation for what a piece of art parser generators are. They are the ultimate forms of software design. Here I'm taking one second to appreciate this epic error message that LALRPOP produces.

You want 4. Those who can't but know how to tickle ChatGPT the right way to get a solution. They actually deliver anything you ask for.

Out of curiosity is stage 3 a sufficient condition for implementation issues? I imagine the "research phase" should start before stage 3 and stage 3 should be for "experiment phase" and "shipping phase" already?

Would be useful to have links to implementation issues too, so we know how close we are to stage 4 (and to documentation inclusion)