Linkblog/2025/02/17

Interesting discussion about the significance, or lack thereof, of the date 1875 when it comes to Cobol.

Recent chatter has made fun of the DOGE crowd for not understanding Cobol, with the simple answer that: “0 in COBOL date is 1985”, or something along those lines.

Does not seem like this is the case, if loops is to be trusted.

Ideas are like fish. If you want to catch little fish, you can stay in the shallow water. But if you want to catch the big fish, you’ve got to go deeper.

We’ve all got hundreds or even thousands of ideas floating around in our brains. But the really big ones are few and far between. Once you catch a good one –because they’re so rare– write it down immediately, says Lynch. From there, ideas attract other ideas and start to grow from their initial seed state. The final job is to translate those ideas into a medium, whether it’s a film, a painting, or software.

Ideas are gold, part of this Linkblog / site is hopefully to put down more of them not just in my own private Logseq graph, but into the world, maybe it will make them more real.

RIP David Lynch

Despite my enthusiasm for uv, I insisted that I couldn’t recommend it before having seen it in a lot of different contexts at work.

That’s because the Python community is huge and diverse. You have students, data scientists, AI devs, web devs, sysadmins, biologists, geographers, plugin authors… They may work at university, in the administration, in a startup, in the army, in a lab, or in a big corporation.

I agree heavily with the points brought up in this post, as someone who did in time eventually learn how to live with Poetry, but still found its interface to be less than preferable (also horrible slow at SAT solving the lockfile for my last works project…), uv (& co) just do the thing, and do it really fast.

Saw mention of this post on Simon Willison’s Weblog.

Michelle discusses Robin Sloan’s Is It Okay, the discussion of if LLMs are ethical.

The discussion of AI for super science is interesting, Robin (who is a science fiction writer), mentions that AI used to cure cancer are obviously ethical due to the effect that it would have on society, but Michelle points out that “LLMs are not this. They synthesise text, which is not the same as data. Particularly when they are trained on the entire internet, which we all know includes a lot of incorrect, discriminatory and dangerous information.”.

I would also like to see super-science produced information spit out of markov chains gone insane, but I think I see the perspective of Michelle, where that simply just isn’t what they are meant to be, to extent beyond the latent space into the areas unknown, that probably has to be the human’s doing, not the LLM.

Found in Sophie’s roundup of links.

I contend that ai agents should be built exactly like local first apps/clients. every agent should be able to connect to a “server” that provides the context of the system, sync that state over to it’s own persistence system, and perform actions on the system.

I think I understand Sunii’s concepts here, I would hope so since I’m building agentic stuff at work at the moment.

But honestly, I am not grasping.

But it still feels like something I wish to make note of / share.

Brad is talking in the public space about gamedev ideas.

The first entry of a journal documenting the development of a video game. Coming up with an idea for a game.

His game idea seems big, but so is his digital garden, hopefully the type to see through such an endeavour!

Take meetup link, say https://www.meetup.com/ndevmeetup/

splash /events/ical on it!

A URL you can chuck at your calendar: https://www.meetup.com/ndevmeetup/events/ical

Very nice, thank you Taylor.

🇺🇸🦅✨

Very cutesy storage icons from Louie.

This is very disappointing to see, but also extremely understandable from Hector’s perspective.

His contributors to software over the last ~20 years have been great, you know you’ve made it when you’re seen as a software developer who deserved their own wikipedia page.

Hopefully whatever he moves onto next, is less toxic.

HTMX CEO coming in with some wisdom.

TLDR Three “dirty” coding practices I’m going to discuss in this essay are:

• (Some) big functions are good, actually

• Prefer integration tests to unit tests

• Keep your class/interface/concept count down

If you want to skip the rest of the essay, that’s the takeaway.

I am so team “prefer integration tests to unit tests”, Carson also dunks on mocks, mocks are evil, they are not real, you end up testing mocks, not testing your system.

Preach.

From Pierre’s collection of links

So many nice patterns.

I vibe with this gooey shape pattern from Brett Wilbanks.

Shop cat shop cat shop cat shop cat.

oklch() is a new way to define CSS colors. In oklch(L C H) or oklch(L C H / a), each item corresponds as follows:

• L is perceived lightness (0%-100%). “Perceived” means that it has consistent lightness for our eyes, unlike L in hsl().
• C is chroma, from gray to the most saturated color.
• H is the hue angle (0-360).
• a is opacity (0-1 or 0-100%).

New color syntax just dropped.

OKLCH frees designers from the need to manually choose every color. Designers can define a formula, choose a few colors, and an entire design system palette is automatically generated.

This is very cool to see in action, I’ve seen some clips of people just changing the hue angle, and things just sort of work without having to modify the other values.

Dax posted this great clip to Twitter showing this in action:

I recently realized that there’s a whole generation of new programmers who don’t even know what StackOverflow is.

This is a sad but true realization.

Depending on how you prompt, ChatGPT can be great at telling you the how without the why, and the worst of it, is even if you ask it to explain itself, you’re likely not able to sus out if its lying to you.

But, if Claude / OpenAI was a drug, I am definitely an addict, I tell myself that I still have been in the workforce before all this was a thing, therefore if I disconnected from the internet and ran no local LLM but still had access to GitHub / StackOverflow, I’d survive, but honestly, I haven’t even tested what my skill level would be without the tools I am now extremely accustomed to using.

Randy shipped a game, a real game!

He made it over a busy busy busy 32 day period, made using Odin + Sokol, with a demo you can play right now on the web.

A lot of the development happened live on YouTube as well, oh how I would love to see a supercut of the ‘best of’ of these livestreams to extract the good good knowledge on his approach to game design / working with Sokol / Odin / etc.

Love to see it.

Users attempting to add their Signal.me URL to posts, DMs, and bio descriptions are being met with error messages.

Love the war on encrypted chat from every angle, even the free speech absolutist Nazis.

If you’re new to my annual gear post, I travel indefinitely (often 30+ days at a time) with a small backpack, and I maintain a list of everything I travel with every year. I’ve been doing it since about 2008, and people on onebag have said that I was the first one to make this format.

Love these sorts of what do you have on your person type posts, with reviews, and links of where to cop.

Also Custom IEMs with your name on them? Dope.

Vercel actually cooked with this one, what seems (at least to me) like the best option to use whatever LLM you want (local or remote, Claude or OpenAI, Gemini or Deepseek, etc.) with a very ergonomic TypeScript interface.

Like the Drizzle for LLMs.

This makes a lot of sense, I have a feeling making a SBOM accessible with some payment would be a great incentive for companies to pay into open source.

“Instead of scanning for copyright notices and license texts yourself, just sponsor us on GitHub and get access to always up-to-date SBOM information by the people who really know what‘s inside”.

Better than “Also, the author of core-js is looking for a good job”.

Some thoughts on the latest terrible advertisement for Meta’s metaverse and on the man responsible, CTO Andrew Bosworth.

If you haven’t already seen the ‘ad’, go check it out.

Seeing what billions of dollars gets you when you have no soul as a company is hilarious.

Jack Grimes brought up something I’ve always felt to be the case over on Bluesky:

Impossible to care about Facebook VR at this point but it’s so funny how something like VRChat is free and fully community-maintained and it has like, endless weird stuff to explore and sign language support and actual games and Meta Worlds is like becoming a Sim so you can sit in a meeting

Lovely components, peep the calendar:

The React team (many who work at Vercel) as pushing people towards frameworks over libraries, and honestly making the wording seem like you’re a bit silly if you’re even considering a library…

Mark Erikson opened up a pull request to react.dev proposing changes to the published documentation.

Thankfully some different changes were merged in, which incorporate some of Mark’s content, and overall just better wording in a few places.

But still, this is sort of frustrating to see, I am the first in line to say that frameworks like React Router v7 (Remix) / Next.js / SvelteKit / SolidStart / TanStack Start are the way to go if you have the choice, but, the sheer amount of people who just want to slap a React app up on a GitHub page, or already have a perfectly good SPA they’ve been maintaining at their job for years, cannot be forgotten about as just simply ‘the old path’.

Evan You (of Vue / Vite fame), spoke about the hesitance of speaking fondly of Vite:

Based on what I gathered, it feels the React team’s hesitance on Vite (and build tooling in general) is how it aligns with their vision of React (i.e. RSC being the recommended paradigm), and how much influence / connection they have with the said tools to make sure the integration works as they designed and can adapt to design iterations.

I’m of the mind that Vite is perfectly poised where it currently stands, curious to know where this would even go.

But, in a follow up Tweet, Evan says:

Update: someone from the React team has reached out to me saying I am egregiously wrong.

I apologize if this is not what the React team is thinking, but it’s my honest guess because I am just as puzzled as many React users by what took it so long for Vite to be better acknowledged as a recommended tool.

I don’t like it when my parents fight.