hyperpape

Saw someone predict that AI would enable exponential gains in hardware performance. It would truly be unprecedented. en.wikipedia.org/wiki/Moore%2...

You can be 95th percentile at recognizing that 95th percentile isn’t that good, but you’ll still be frequently surprised by examples where 95th percentile just isn’t that good.

It is a truth universally acknowledged that a nerdy guy must have a page where he tracks his predictions. I have one prediction. I think it is obvious. justinblank.com/notebooks/pr...

They should form a subgroup dedicated to games. Slogan: “it’s about ethics in game development”.

Someone mentioned a philosophical view, and I realized I had a very relevant paper from grad school, which I went to reread. Do not do this. I’m taking psychic damage here, my writing was so bad then.

I have never once seen a human drawn picture of a pelican riding a bicycle. I wonder if we’re even capable of it.

A friend and I took Analysis 2. On the three tests and final (in person) our scores were roughly: Her: 95 95 65 95 Me: 95 65 95 95 Completely opposite senses of what was difficult and what was easy. But yeah…going from 100 to 46? Odds are you cheated.

My childhood hero was Socrates, and I think he’d be proud.

I have a love/hate relationship with the Google AI snippets. I don’t exactly think they’re good, but they do make the world more surreal. Today, I was inspired by a shitpost, and got the AI to tell me that neuroscientists caution against thinking.

Reviewer #2: This does not remotely meet the bar for a novel contribution to modal logic. See Chapter 1. of Hughes and Cresswell. Reject.

By the intermediate value theorem, there must be a programmer whose work is neither better nor worse.

Time to read “Languages” by Dr Book.

In just four years, more than 300% of the HN front-page could be based on incorrectly extrapolating exponential trends into the future.

Feeling the asymmetry of bullshit. Someone posts an arXiv article. Have they read it? Who knows? I read 25 pages of 50, and think that the article’s methods seem bad. Do I 1) say nothing, and let the bullshit stand? 2) say something, and feel guilty for commenting on an article I haven’t finished?

Local reasoning? Training on examples pulled from the web, and it’s hard to fix in post-training? I don’t really know the cause but I have guidelines about not repeating constants.

Twice I’ve heard this line. Once, spoken by a criminal who’s trying to make simple murder sound badass and righteous. The other time it was Samuel L Jackson, and he delivered it better.

Super valid, but the part where he’s saying it from the Native Americans’ homeland is equally wild.

I think a necessary condition for “an emergency” is that congress ratifies the existence of the emergency and the use of emergency powers by law within 5-10 days. “But congress might not…” In an emergency, they would.

Be a captive audience listening to someone drone when the alternative is talking to your friends, vs sitting on your couch when the alternative is some reality TV dreck.

The proper response to “I’m not aware of” is always “let’s watch it now” or “review it and return to give your response to us.” Anything else is depressing accountability theatre.

I knew a dev working on a product where end users could see stack traces. He specifically rewrote null pointer exceptions, because his users were somewhat technical, but not programmers. They would see “null pointer exception” and try to guess what data was missing in lieu of a bug report.

In all seriousness, it’s pretty clear what happened here. In someone’s head they defined 2010s liberalism = woke = censorship, and now that right wing censorship is in the news, it feels right to them to mock it as “woke.”

There are hard bugs, but the fact that it’s possible to see this screen indicates a sort of trivially dumb failure mode.

I saw someone citing a well known philosophy paper to make a point about AI, and I sometimes wish there were disclaimers about which bits of philosophy were real, and which are strike dummies for training undergrads.

One my life’s tiniest regrets is that I never wrote up my review of Kingsman: The Golden Circle, based on the screen of the guy one row ahead of me. It’s hard to believe that any dialogue could’ve made it make sense.

Showing up to the pro-Western rally with my Marx/Sartre/Derrida poster…”guys, are we going to the same event?”

Posted notes on retries, circuit breakers, and other tools that are 100% guaranteed to make your systems more reliable or break them worse. justinblank.com/notebooks/re...

Heraclitus: I closed another bug with “could not reproduce”

There are whole projects dedicated to transforming java bytecode into something that can't be decompiled. Meanwhile, I've done it by accident. Tried a tweak to the regex compiler. It didn't work. Went to debug: public int indexForwards(int param1, int param2) { // $FF: Couldn't be decompiled }

“by far the easiest way to make a compelling, contrarian, data-driven argument is to misunderstand something about the data you're looking at.” www.thediff.co/archive/long... (Have not read the piece being commented on, just enjoyed the comment).

This. You don't even have to agree with any version of US civic belief, but the best version of the US-as-an-idea makes all of us born here citizens. And citizen as our highest title.

When I hit the “Substack won’t let you close cookie notifications” bug, I assumed it was just me. Then two friends said they also saw it, so I figured “that’s a nasty bug, surely it’ll be fixed soon.” That was four days ago.

I’m generally pretty bearish on AI codegen but if we’re trying to make Webpack do something different I’m absolutely hiring a robot to suffer in my place