gotta say I do enjoy watching the leopards eating *each other’s* faces

I believe the phrase is "Hang it in the Louvre"

Trump is making China very rich

God, I love her so much. Just banger after bager.

She's in it purely for the love of the game and I respect that.

Lovely example of how to use the agent effectively! BTW a great thing to add for a tool like this would be SARIF output, allowing you to pipe it to GH code scanning tools, reviewdog, etc.

All but three (!) of the 1,651 refugees admitted into the United States in the three most recent months we have data are white South Africans.

For comparison, in 2024, the U.S. admitted 100,060 refugees. It's an explicit white supremacist project.

www.rpc.state.gov/documents/Re...

A text message that reads: Hey [object Object], if you could bring your electric bill down to 0 for the rest of the year with solar, would you want to find out how?

Please, call me Child Object, object Object was my father

The brain worms are clearly levelling up.

wat

Highlighted section of the Copilot ToS which says "Copilot is for entertainment purposes only"

www.microsoft.com/en-us/micros... is it good to build an entire economy and software infrastructure on this

in 2025 I worked for people who thought “AI is cool” was a company value

man… remember 2025??

Why on earth is there new Harry Potter? The trailers even make it look almost identical, artistically speaking, to the films. So… what’s the point? Please do not feed the TERF.

“Hello, babies. Welcome to Earth. It’s hot in the summer and cold in the winter. It’s round and wet and crowded. At the outside, babies, you’ve got about a hundred years here. There’s only one rule that I know of, babies—God damn it, you’ve got to be kind.” Kurt Vonnegut

Language models are an incredibly powerful tool for many tasks, including tasks many people would not have predicted computers would be good at just a few short years ago. But they are not machine gods, and they are not capable of replacing humans in high-consequence systems.

Are the tools I use to do my job changing? Absolutely. They’ve been changing my whole career, albeit usually with less attendant hype, less capital investment, and less environmental damage along for the ride.

If the chain of reasoning you need is in the training dataset, you might be in luck. If it’s not…? All I’ll say is: have fun with that :)

And almost always it involves reasoning about things that language models cannot reason about. (Spoiler: it’s because they are not reasoning about anything.)

But the process of debugging large systems is much more complex than efficiently processing quantities of text. Sometimes it’s about understanding that something should be in the logs but isn’t. Sometimes it’s about remembering a story someone told you over a drink at a bar 15 years ago.

Some of these new tools can also help with figuring out what’s going on. I recently discovered that Claude is excellent at figuring out what’s happening in an strace dump!

And if the way we adopt language models involves shipping large amounts of machine-written code to production, I find it hard to see how that doesn’t increase the demand for skills of production debugging and system comprehension.

Heck… if anything, language models’ limited context makes them substantially more prone to misunderstanding how their code will work in production than most humans.

All complex systems operate in ways that differ—often substantially—from their creators’ intent. Anyone who has used language models to assist in writing software knows that they are not exempt from this law of systems.

A large part of my job is figuring out the differences between software systems “as designed” versus “as deployed” (and helping others to do the same). I have never been more confident that the demand for those skills isn’t going away.

reverse engineering is the process of finding out after somebody else fucks around

You are 100% right and I’m still pissed 😄

I mean that problem goes back wayyyyy further. But yes.

the council regrets to inform you that “ship everything our customers ask for” is not in fact a strategy

How is it 2026 and we are still falling into such obvious traps as measuring inputs?

Tech as an industry learned that DORA metrics were what we should be looking at to understand capability, refused to actually measure them, and then jumped headlong into a new hype cycle where the prime metric appears to be “how many lines of code can we write.” (Not ship, write.)