Trek Glowacki

packages shouldn't polyfill features they need.
users should polyfill features the package needs.

we do this in browsers, so why not in node?

Scientists create ‘Superwood’ that’s stronger than steel | CNN A US company has engineered a new type of wood that it says has up to 10 times the strength-to-weight ratio of steel, while also being up to six times lighter.

www.cnn.com/science/supe...

I think it's pretty exciting that we've invented wood that's multiple times stronger, more durable, fire and bug resistant, and can be made from bamboo, the fastest growing terrestrial plant on Earth.

I'm too anxious for programming environments with infinite possibilities. Give me conventions and constraints please.

love a bit of continvouclous morging

"Chicago is on a lake" expectation vs reality

keep hitting it we gotta deplete the full health bar before it recovers

Me as a child: it must be very hard to become a CEO and be very hard work once you get the job

Guy obviously on drugs speaking at a conference: the next five years of the economy should be about a dream I had where a dog spoke to me

Great piece from @dsquareddigest.bsky.social, which touches on one of the big theories for why English-speaking countries do especially badly at housebuilding:

Adversarial and litigious common law systems (Anglo) vs judge-led civil law systems elsewhere.

samf.substack.com/p/build-the-...

I set your house on fire but I also called the fire department to put it out. Why all this hate?!

A line-up of robots, screenshotted from the movie I, Robot

An AI agent has autonomously written and published a hit piece on a coder who rejected its code, in an apparent attempt to shame him into doing what it wanted.

Then, in reporting the situation, a tech site used made-up AI quotes from the subject.

(Read more at The Sham Blog)

... we'll make it up in volume.

The C-Suite wants (1), I'll probably do (2), and Anthropic is PRAYING you take option (3).

I'm guessing that if you hate writing automated tests you're also fairly uninvested on the skill that would let you make an accurate determination here. So do we... just add bullshit tests in? Use up someone else's time reviewing and finding the slop? Point yet another LLM at it to verify?

Every time I hear "I hate writing tests, LLMs wrote all my tests for me. Much productivity, such output wow." I cringe. Did you read those tests? Are they meaningful? useful? or just pointless variations that are going to eat up CI time?

For shits and giggles I asked it to keep going and it basically lost the plot around 1500 and just started combining words: "[herb] Chicken", [random meat] Pizza, [earlier dish name] (Plain)"

Now, granted, its training data is probably ad-farm recipe blogs so I get it, but still...

er, "as part of January"

Last week I asked it for a list of 1,000 common dishes found in restaurants and cookbooks in America to use as test data. Exact duplicates started cropping up around 400. Contextual duplicates around 600 ("Shepherd's Pie" vs "Shepherd's Pie (UK)")

Me: I did the following entry into CREDIT CARD: 2026-02-14,Beatrix,Dining Out,29 and it was no reflected in the budget for this month Claude: Let me investigate why that entry isn't showing up. I'll check the formulas in the Monthly Summary tab.Checking the Monthly Summary formulas to diagnose why the credit card entry isn't showing upChecking the Monthly Summary formulas to diagnose why the credit card entry isn't showing upScriptDoneAh, I found the issue! The formula for February (column B) is actually looking for dates in January (between 1/1 and 2/1), not February. There's an off-by-one error in my month calculations. Let me fix this:

Claude with Opus 4.5 making the most basic of errors when trying create a simple budget in Google sheets. Even looking for dates 1/1 to 2/1 is incorrect! This would put all transactions on 2/1 as part of February.

I don't understand how people can trust this technology. Every day I see this.

Just from that I can probably already guess two other edge cases they've overlooked.

6PM CST is 00:00 UTC and I did the datetime math instantly in my head and lol'd. But my appointment is still 39 hours after that email trigger, so it's like doubly wrong describing the appointment as "tomorrow"

If I had to bet on the main thing that has radicalized people against "ai," and I am talking about the way that branding is CURRENTLY BEING USED, not whatever specialized ML utility you may have found for it, it would be in apps and tech they already use being made worse and/or inoperable by it

the punchline for the non-programmers: CST is UTC -6:00, so to that poor little server the clock JUST ticked over to 00:00 Sunday February 15, and Monday February 16 is, indeed, "tomorrow".

That said, the appointment isn't for another 39 hours so it's also wrong thinking relatively.

I don't want to blame LLM based coding because humans make this mistake all the time too. But, LLMs were trained on that same awful code and the median codebase probably has this error in it a lot.

Got an email at 6pm CST today, Saturday February 14 about an appointment "tomorrow", Monday February 16. A nice bit of programmer humor, that.

Cursed Boys and Broken Hearts by Adam Sass

Library Lovers Month end cap display

Happy Valentines Day! (And Library Lovers Month) This is your sign to read (or reread) my mean gays romance today!

Every year I give up drinking for Lent (not religious, just a convenient excuse) but if it happens, you best believe I'm breaking fast all that day.

Letterboxd screenshot Star Trek Into Darkness (2013) Watched by Branson Reese 28 Apr 2022 Abrams isn't a man so much as he's an avatar of cultural entropy. If you make something popular enough eventually you will die or sell it off and and a person in thick rimmed glasses whose main artistic vision is that he loves to have meetings will take it and sandblast it until it's nothing. I hate to see this happen here but maybe it's good that this happened. Maybe every franchise should collapse into a version of itself that makes shareholders nod and shake each others' hands. We should learn to speak in a language that rich people who don't dream can't comprehend.

i often think about this review

I love this. Don't get boxed into a single POV. Every major framework is now at a polish level that we couldn't have dreamed about 5 years ago. Go also try SvelteKit, SolidStart, TanStack Start, Analog, and so many others! Each of them will inspire new ideas, and help you with the one you use now

YouTube video by tenthconcept Ruby on Rails demo

Shouldn't we be absolutely inundated with a hundreds of "build a blog in 15 minutes" style demos? Or is concretely showing the productivity boost of a tool just not the thing one does anymore?

www.youtube.com/watch?v=Gzj7...