Mhm. Is it just me or did Claude usage _significantly_ skyrocket recently?

I know this is probably all subjective and likely due to a poorly written harness or whatever, but still.

Priorities man!

This, completely made my day.

*Random guy on the Internet raises the hand for a crisp high five*

The cooking levels is, over 9000!

Fabulous.

There sure are a lot ”companies” that ”really would like my expertise” and pay me for ”insights into AI”. 🤖

Anyone who’s actually taking anyone up on one of these?

Sounds like the more recent version of the email stating you’ve inherited a small fortune, just as long as you input the bank info.

Haha oh wow, hard no.

I don’t doubt it one bit!

@kitlangton.bsky.social's effect.institute is freaking _insane_. This, this is how you do an interactive course and make it fun. Like, really fun.

The level of detail and polish here makes this look deceptively simple.

But there is true craftsmanship here.

Bravo sir. Bravo.

Apache Iggy looks so good, and that blog post on moving to io_uring is just a great read.

Alright friends.

Does one need a printer at home in 2026?

”It never resonated with me”

Yeah no shit.

I can’t get over how fucking brilliant Shōgun is.

Hiroyuki Sanada is scary good as always. But the whole cast is… exceptional.

I don’t need a MacBook Neo. I don’t need anything really.

But something about it is appealing. It’s calling to me.

I was even trying to argue the case with my wife about getting it instead of a used iPad for the kid when she’s older.

A bag of dates with ”cinnamon bun” flavoring, which might just be the most Swedish thing ever.

Tell me you’re in Sweden without telling me you’re in Sweden.

Right. Don’t let perfect get in the way of getting something done, or even started.

Can LLMs Be Computers? | Percepta We build a computer inside a transformer — executing arbitrary C programs for millions of steps with exponentially faster inference via 2D attention heads.

”We answer this by literally building a computer inside a transformer. We turn arbitrary C code into tokens that the model itself can execute reliably for millions of steps in seconds.”

You did what now?

Ah. Glorious.

This made my day.

I love the internet.

And Rust.

And esoteric bugs.

As long as I don’t get them. Or well, that’s not entirely true. Finding them and solving them is _the best_.

This is interesting, just for the sake of argument (not trying to prove you wrong or anything) – in what indirect way?

Isn’t the ease of working with, and maintaining the code directly correlated to the quality of the resulting product?

Is it just me or did the Amazon iOS app somehow turn into even more hot garbage? I can’t even tap recommendations.

FLIPPENDO ! - Comet ( Chiptune / DNB / Gameboy ) YouTube video by FLIPPENDO !

Okay what the hell how do you even

www.youtube.com/watch?v=27Ib...

Hell to the yes. Is how.

I thought that went without saying, but I have now made a note of it.

”We moved from waterfall sign-offs to continuous integration. We can move them again.”

Ugh, what? Move it where? Back to waterfall? What?!

I’m so confused Sam.

Ah, well I respect that! Hope I don’t pile on here.

That’s just it right? There isn’t really a golden solution, because I’ve definitely worked with/on teams where the backlog has really worked, but that came down to the team members being just excellent humans I think, not the process itself.

I don’t have a good solution to this mind you, so I’m really curious if there are any takeaways or actual practical applications of ”things to do” as opposed to ”things to say”.

The real challenge lies in identifying the most critical tasks that truly matter, while simultaneously managing and addressing other planning or refinement sessions, requests for updates, syncs, and the need to actually work, review the other 15 LLM-generated PRs that have been released.

Looking forward to this one. As I have strong opinions here.

The backlog is a lie. A safety net we tell ourselves represents the upcoming world. A ”groomed” backlog, or whatever - is mostly for anyone that isn’t a an engineer, and even then it barely gets read, even less now with LLMs.

I know I can dismissing by tapping but, I don’t know. So far not a fan? Will need to give it a longer whirl though.

I do seem to make fewer mistakes while typing?

Hover typing in iOS is already infuriating because it positions itself on top of the UI element that you need in order to send the message.

”Luckily” BlueSky has that button on the top right, so not an issue in that regard, but it hides character count - so I’m guessing right now. Made it.