“Football won today” Goddamn right, Zlatan.

Spain, about Argentina’s goalie:

My toddler has learned the word “chaos” and we’re entering a new era of self-description.

I’m getting frustrated with Fastmail’s spam filter. I’ve now had multiple VERY INCREDIBLY OBVIOUSLY spam emails land in my inbox, while my USPS Informed Delivery emails regularly end up in Junk. Never thought the day would come when I consider going back to Google for email.

If you’re not pulling the catalog cards yourself and then spending an hour wandering shelves and carts as you try to find the perfect tome to reference, are you even doing research?

Ah, ça ira, ça ira, ça ira!

“We need a Butlerian Jihad now” Yeah because the world created from that is an absolute paradise, definitely nothing abhorrent or fucked up at all.

so my toddler decided he didn't like sleeping anymore while we were on this trip so didn't have a chance to get to this then gonna try and squeeze it in some other time, but still excited about this one

Marathon on the Switch 2 would fix me

AWS Lambda written in C#

Seriously, I’m either doing something terribly wrong or everyone else is overthinking it.

I wish I could live in precedented times.

Them: “You’ve got to have a theory of mind for agents and understand how to prompt different models.” Me: “haha talk to model and agent go brrrrrrr”

Me: okay, let's figure out how to get this intense coding task for work done with Haiku. Gotta save the money and watch tokens. Also me: alright, I'm going to tweak my Emacs config, time to break out Fable.

“We’ve got an LLM that’s uncannily good at finding exploits in code and chaining them together for real cyberattacks” “Oh, wow, can we use it to maybe find and patch those vulnerabilities?” “lol fuck no, get rekt”

Football is a simple game. Twenty-two men chase a ball for 90 minutes and, at the end, you get cooked by two guys from the 28th-best team in MLS.

joining in on the atproto fun by making a rubik's cube time app i'm going to be at the beach all week, so let's see if i can get it done in that time lexicons and an appview, shouldn't be too bad

I feel like someone should write a book about this or make a movie, or something. Seems like a slam dunk for some classic science/speculative fiction.

I’ve got a safety razor from 1964 that I bring out every month or so.

A lot of people really enjoy hating on AI, thinking it's an unholy abomination that should be scrubbed from the earth. But they don't consider that it's helping me port Emacs to iOS.

I believe? I believe that...we can...win?

I’m not kidding when I say this might destroy my family.

Not the most complex thing, but GLM 5.2 one-shot a tic-tac-toe game in a self-contained HTML file for just under 17K tokens total. Prompt was one sentence. Not bad.

Marathon on the Switch 2 would fix me

So how likely is it that the US gov ends up with a decent share of Anthropic’s IPO and that’s what gets Fable/Mythos back? Maybe not IPO, but shares in the company.

I love that Anthropic has already trained me to tell Fable where the authn/z parts of my codebase are so it knows to delegate the work on those parts of my code to a less capable model, thereby aligning on security and safety.

I really want to know which sicko on the Bsky team implemented their 2FA.

There’s also the built-in MCP server in Xcode if you don’t want to deal with Xcode’s interface. developer.apple.com/documentatio...

About to make a subagent to review my prompts and modify them to not trip Fable’s safeguards.

wait wait wait...is Apple doing WSL? github.com/apple/contai...

Anthropic: you can have a little Mythos, as a treat

ayyy I'm getting some personality when claude is in "work" mode. it's just a basic smiley and a ✌️, but I'll take it.

context is weird, man. i open a claude session to work on configs or something, tons of personality and stuff. in a work repo? none of that. like, it doesn't get cold or anything, but you can tell it's got a tie on. just because of one word in the project's path.

I think a lot of y’all would have a better experience with AI agents if you kept sessions short and focused. Like, you know how god-objects are an anti-pattern in code? So go god-sessions.

Aww yea Apple Intelligence for all my HomeKit stuff, it's such a well-adopted ecosystem I can't see how this could possibly be pointless

Siri, come up with a snarky post about Apple re-hashing the 2024 WWDC keynote and acting like this is the first time they've talked about Apple Intelligence + Siri.

O365 is a thing that exists. I never even have to think about HR, but that shit? Every goddamn hour.

The fuck is happening in Monaco?

i think i'm getting old, i just turned on light mode and it's easier to read things

You know, you can just choose to not say things. That’s always an option.

Final boss of Stormlight Era 2?

FIRST LIGHTNING BUG OF THE YEAR LETS GOOOOOOO

Huh, this seems to be my impression too. The harder I try to prompt agents via Markdown files/hooks/rules/etc, the more time I spend churning as I rotate out models. www.seangoedecke.com/prompts-are-...

Hard to beat steak and scotch.

I think there are a lot of people who aren’t willing to go scorched-earth on how they interact with agents every once in a while.

I feel like this post is directed straight at me and my previous team at $EMPLOYER.

Got GPT-5.5 to use goblins as a metaphor. Apparently they haven't fully beaten them out of the model.

Debating setting Mythos a /goal to get O365 email working in mu4e once it drops. Probably cheaper and faster than convincing $EMPLOYER that letting me read email plain-text is actually a security win.

We all are, minute by minute.