Joel Uckelman

Joel Uckelman

@uckelman.bsky.social

vassalengine.org Working on V4... | Github: uckelman | Logician | CS PhD | OSS contributor | vim | Wargamer | Short-stack developer | Complete the Five Year Plan in Four Years | Trolley opportunities

301 followers157 following393 posts

Joel Uckelman reposted

eBay users! Don't ignore today's email, even though it looks like a boring standard update to the User Privacy Notice. They want to use your data to train AI models. If you click the link over the words "the right to object", it takes you to a page with a toggle button letting you opt out.

Almost every time I've ended the day with a problem on the current project, I've deduced the cause at 3am and in the morning turned out to be right. It's reassuring.

This is just to say I have eaten the data in your HTTP request and which you were probably expecting would go into the database Forgive me it was delicious so 200 and so OK

What I love about this is that a senator from Missouri is being corrected by a senator from Minnesota about the name of a college team from Iowa while the chair of the committee---a senator from Iowa---sits there in silence. That is some kind of metaphor for Missouri, Iowa, and Minnesota.

Ben Collins@bencollins.bsky.social

Gobsmacked by the elite levels of not knowing ball in the Senate today, thinking an NBA team is called the Hawkeyes then, refusing to Google it, then accusing a guy of treason because you didn’t know the Iowa Lady Hawkeyes aren’t in the NBA. Somehow I don’t trust these people to Save Girls Sports.

If this is accurate, it's hostile action by OpenAI. Why are you sending your data to third parties again?

Eoin Higgins@eoinhiggins.bsky.social

OpenAI claimed their model had a breakthrough in a longstanding mathematical problem, but this scientist says that they basically stole his work

The warning text for the plaque marking long-term nuclear waste storage would make banger code comments. Consider your reaction if you were reading a C header file and you saw this: // This message is a warning about danger. // The danger is still present, in your time, as it was in ours.

Nah, that could never happen. Right?

Maia@maiamindel.bsky.social

i am a tech company in the year 2026. i am called nazgul after the evil soulless ghouls from the lord of the rings that go around killing people. my ceo is a crazy drug addict who keeps talking about war and jesus christ on tv. give me a trillion of your tax dollars

This is your reminder that computational complexity is swamped by cache locality when n is small. I just now benchmarked checking a six-element array for duplicates by brute force (6 choose 2 = 15 comparisons) versus using a hashset.

"I propose to fight it out on this line if it takes all summer." -- Grant, at the start of the Overland Campaign, 1864 Feeling this, combined with the theme from Gilligan's Island, in a work project...

Some interesting analysis about returning Result from frequently-called functions: purplesyringa.moe/blog/you-mig... I like the trick for keeping the ergonomics of Result without the cost; makes me wonder if this could be done automatically in conjunction with PGO.

purplesyringa.moeYou might want to use panics for error handlingRust’s approach to error handling comes at a cost. The Result type often doesn’t fit in CPU registers, and callers of fallible functions have to check whether the returned value is Ok or Err. That’s a...

There is a lip on the threshold between our living room and hallway that the Roomba sometimes can't get over, not without my help. Every time it makes me think of the Voight-Kampff Test from Blade Runner.

How tf am I supposed to know that the thing labeled "Trusted entities" on an IAM Role in the console is "AssumeRolePolicyDocument" in boto3? Thanks for the consistent naming, AWS.

View full thread

It turns out that there's a simple solution to this: make do types, which are consumed by applying them and return an undo type, and undo types which are consumed by applying them and return a redo type. 3/

This setup means that you can't have a double undo or redo, because you can't even express a double undo or redo. As soon as you call apply, the input is consumed. The object is gone. You can't get into an invalid state because the code which would lead to an invalid state won't compile. 4/

Calling undo or redo twice consecutively on the same action object is a bug---this will corrupt your state. If I undo a move, I should not be able to undo it again, as it's already undone. 2/

Some days you spend the entire day trying to determine where an error is being swallowed, just because somebody didn't take sufficient care in a language that doesn't force you to.

I spent the past three days fighting with import of some data which was malformed, but neither the exporter nor the importer did any type verification, so there was nothing to suggest what the problem was. Why in 2026 do we still have to suffer with shoddy engineering like this? STATIC TYPING NOW

Got some keys cut a few weeks ago, because we changed the lock. They were not three for the price of one, but were guaranteed for a lifetime's use. And you can't have one, you know, the other two. I think this will pop into my head any time I get keys, for the rest of my life.

This has never been a problem for the US Army before... such as when dozens of men died of the flu on the troop ship my grandfather took to France in 1918.

The Washington Post@washingtonpost.com

Breaking news: The military will no longer require U.S. troops to receive the annual flu vaccine, Defense Secretary Pete Hegseth said, rolling back what he described as “overreaching mandates that only weaken our war-fighting capabilities.” https://wapo.st/4eAG2sx