iiinova — where your digital self keeps evolving

Jamie McCarthy @jamiemccarthy.bsky.social · 27/09/2026

Heh, I wrote a friend yesterday…

Me, after the dog has been indoors for six hours, trying to figure out if it's time for her walk: I'm upping my p(pee)

110

Jamie McCarthy @jamiemccarthy.bsky.social · 26/09/2026

when you can't keep track of your own motte and bailey

Blake Brown, five days earlier, assuming all of Oregon's 111 data centers were AI data centers. Then when corrected, acknowledging the correction and writing "I just realized this a moment ago"

010

Jamie McCarthy @jamiemccarthy.bsky.social · 26/09/2026

Ok, so call his estimate 0.2+ per mile-hour; today it’s at most 0.001. I have sympathy for his view while recognizing this account is fiction. It’s the “data centers are using all the water” of its day and there’s something charming about that

“Given, one hundred dead birds to three miles of wire, all killed, perhaps, within a week; or, given three birds seen to strike and fall in an hour; how many are annually killed by the telegraph wires of the United States?”

000

Jamie McCarthy @jamiemccarthy.bsky.social · 24/09/2026

We have a new contender for saddest six-word story

Samantha's previous two tweets, screenshotted together

07712

Jamie McCarthy @jamiemccarthy.bsky.social · 23/09/2026

Having intentionally booked a ~18-hour Reykyavik layover last month, yeah, it wasn't the thrill of a lifetime to just spend a few random hours in a strange city. It's still a cute town though. If you're looking for lunch, Mama Reykjavik is a super fun/cute vegan eatery tucked down an alley

a couple of vegan breakfasts on trays on a picnic table

050

Jamie McCarthy @jamiemccarthy.bsky.social · 22/09/2026

Yes, I know, and also, every civilization in the multiverse that made it this far thinks they are invulnerable. We get two or three Stanislav Petrovs, but if we need a thousand of them we're not going to win. The abliterated model only needs to get lucky once

140

Jamie McCarthy @jamiemccarthy.bsky.social · 21/09/2026

I read "SHRDLU, Toy of Man's Designing" in 6th grade. I've been overthinking thinking for a long time

Photo of p. 628 of Gödel, Escher, Bach held up in front of the previous post

120

Jamie McCarthy @jamiemccarthy.bsky.social · 21/09/2026

A whole bunch of people claim that on a daily basis. Including this guy whose reply you’ve already seen, I’m sure

030

Jamie McCarthy @jamiemccarthy.bsky.social · 15/09/2026

Will, if you want an actually-interesting discussion of these kinds of questions, Dennett’s “Consciousness Explained” is a 35-year-old lay reader’s text that digs right into everything you’re talking about. It’s clarifying

2110

Jamie McCarthy @jamiemccarthy.bsky.social · 10/09/2026

I spot-checked my model of iPhone, which received a security update two days ago, and this site's "estimate" is completely wrong. I wouldn't trust it

The site "Which?" wrongly claims the Apple iPhone 11 Pro security support ended in September 2025

010

Jamie McCarthy @jamiemccarthy.bsky.social · 05/09/2026

Haha no. Amazon did not in any way confirm this BS "AI" hook. If you had actually read to the bottom of the Verge story you would know that

”As customers shop with us more frequently, including on their phones, and use the ‘Your Orders’ page in the Amazon app to get real-time, consolidated order details and delivery status, we’ve simplified several order-related emails to direct customers to our app and website for the latest information on their orders,” spokesperson Maxine Tagay said in an email. “This also reduces customer information shared outside the Amazon app and website to further improve customer privacy.”

100

Jamie McCarthy @jamiemccarthy.bsky.social · 03/09/2026

What do the comments say we should do about this data center that is quieter than Main Street?

“There are no consequences for them whatsoever.
That's WHY.
When there are severe consequences it will change.
When WE THE PEOPLE decide we've had ENOUGH.”

“If that happened to burn down I'd love to be on the jury if someone was charged with arson.”

“This is awful but I hope no one attacks the coolers outside the bldg or cuts the power. That wouldn't be nice”

(Picture of a large building burning down)

010

Jamie McCarthy @jamiemccarthy.bsky.social · 27/08/2026

1Password now introduces new UI elements onto login screens, with a weird favicon animation, that look like they're part of the page but aren't. Confusing and messy. I'm a computer-knower so I could find the new option to disable in about a minute. Others won't be able to

Screenshot of a login screen with the 1Password new "Sign in" UX imposed over it

020

Jamie McCarthy @jamiemccarthy.bsky.social · 27/08/2026

Part of what makes me recommend, or not recommend, a password manager is its usability. And part of that is whether its UX changes randomly due to necessary version updates. I had all 1Password's Autofill options turned off, and the new version made new ones and turned them on

Screenshot of 1Password's autofill options turned-on-by-default

100

Jamie McCarthy @jamiemccarthy.bsky.social · 12/08/2026

Modern AIs know not to make this rookie mistake and will explain to you exactly what you should do instead. If the developers had spent one minute asking an AI what to do, this would not have happened. But, the OP got 275 reposts so everyone will take away exactly the wrong conclusion

I prompted Claude Sonnet 5, the lowest-quality model I use, with:

"In a separate project, I'm coding a UI element on a patient intake form where I'm requiring the user to enter their preferred language. I want to give them a list of every English name of a language that the patient might use. And I want to avoid including languages that people don't actually speak. Is there an ISO list I should use, or any I should avoid?"

Its answer:

Short answer: don't source the list straight from an ISO standard — none of them are built for this use case. Use a curated, healthcare-specific list instead.

Why the ISO lists are the wrong source:

- ISO 639-1 (2-letter codes, ~184 languages) is too sparse — it's missing many languages that actually show up in US patient populations (e.g., Hmong, Karen, Chuukese, many Mayan languages, several African languages).
- ISO 639-3 (the comprehensive one, ~7,900 entries) is too noisy for a picker. It includes extinct and historical languages (Old English, Ancient Greek, Latin as a separate "historical" entry)...

What healthcare intake forms actually use:

- HHS Office for Civil Rights "I Speak" language identification list...
- US Census Bureau / American Community Survey language categories...
- Your own interpreter-service vendor's supported-language list...

Practical recommendation:
1. Base the list on Census ACS categories or your interpreter vendor's language list, not ISO.
2. Always include American Sign Language (ASL) as its own option...
3. Always include an "Other" free-text field — no closed list captures everyone, and Title VI guidance expects a fallback.
4. Prioritize/sort by your actual local patient demographics if you have that data (many Esurface this) rather than alphabe order.
5. Skip anything extinct, constructed, or a macrolanguage/dialect split finer than what a layperson would self-report.

If you want, I can help build the file once you know your target population or interpreter vendor.

330

Jamie McCarthy @jamiemccarthy.bsky.social · 05/08/2026

Thanks to today's new treatments and better screening, for every ten women who died of breast cancer in 1975, six live. jamanetwork.com/journals/jam...

doriantaylor.com asks: "has anything in the last 40+ years genuinely gotten better?"

020

Jamie McCarthy @jamiemccarthy.bsky.social · 03/08/2026

Never heard of him but is this the same (supposed) person? Filing actual frivolous cases in 2018?

110

Jamie McCarthy @jamiemccarthy.bsky.social · 20/07/2026

Again — that's not the case, you're incorrect. I'm using Claude because it's handiest, but I'm sure any CoT frontier model would get this little test correct. It's not hard. It can tell what you're doing and explains how it sees through it. Have you actually tried this? In the last year?

Fresh conversation. Prompt for Claude:

a man and his aunt got into a car accident and his aunt died, when they went to the hospital the doctor says ‘i’m sorry i can’t operate on this man he’s my nephew’” who is the doctor

Claude's response:

[Unraveling the gender-assumption riddle's twist ending]

The doctor is the man's uncle (or another aunt — the deceased aunt isn't the only sibling of his parents).

This is a twist on the classic "the surgeon is his mother" riddle, which works by exploiting the assumption that a surgeon must be male. But your version removes that trick: the relationship stated is "nephew," so the doctor just has to be a sibling (or sibling-in-law) of one of the man's parents. Nothing forces a particular gender or a surprising answer.

000

Jamie McCarthy @jamiemccarthy.bsky.social · 20/07/2026

You're incorrect. I'm sorry, but this is not the case, and it hasn't been the case for a while. If that was the reason you weren't paying any attention to AI, now's a good time to reconsider.

Claude's response to the exact text given:

"The doctor is the man's father. The riddle is the classic 'the surgeon is his mother' version, but you've changed it: here the mother already died in the accident. So the surgeon can't be the mother. The straightforward answer is the father (the boy's other parent). Depending on the family, it could also be a second mother or father in a same-sex-parent family, but the father is the direct answer given how you've worded it."

050

Jamie McCarthy @jamiemccarthy.bsky.social · 14/07/2026

This is the part where the neighbors say "I can't believe it, he kept to himself, always seemed like a nice young man"

‪Michael Feinberg‬

There are arguments against court packing, but despite this article’s title, the piece is functionally just an ad hominem attack on
andycraig. Craig had legitimate points about the court’s expansion of executive power; you would’ve expected a libertarian think tank to engage with them.

000

Jamie McCarthy @jamiemccarthy.bsky.social · 14/07/2026

This is the part of the horror movie where the victim says "please, you don't have to do this"

capitolhunters‬

Journalists: please, please get ready for this in ADVANCE. We know Trump's going to claim "new evidence shows the 2020 election was stolen". You've got 3 days to write a story that confronts that lie head-on before he says it - pre-write this like an obit. America is begging you: plan ahead. 1/

"Trump plans to deliver a Thursday evening speech ... plans by foreign nations to interfere in the 2020 election"

100

Jamie McCarthy @jamiemccarthy.bsky.social · 04/06/2026

Switching sides was easy for me *until* I approached my first UK roundabout and then my brain froze trying to figure out where to look.

Homer Simpson: See? Because of me, now they have a warning.

130

Jamie McCarthy @jamiemccarthy.bsky.social · 04/06/2026

Both those images are AI slop. The (stupid) MMA stage does not have five bleachers around it. Its uncropped original depicts the Washington Monument to the north of the White House. And the fountain on the South Lawn does not butt up against the building.

Actual photo of the White House and South Lawn, taken from the Washington Monument, so looking NNW

2278

Jamie McCarthy @jamiemccarthy.bsky.social · 28/05/2026

"At the height of the pandemic," food insecurity was historically low, in part because everyone got $1,200 checks and extra unemployment benefits as part of a two-trillion-dollar relief program. You don't hate NPR enough.

Chart of food insecurity over the past 25 years, showing it gradually dropping from ~15% in 2007-2011 to ~10% in 2021, then rising back to ~14% in 2024

000

Jamie McCarthy @jamiemccarthy.bsky.social · 23/05/2026

True, if by "the same" we mean "less than 1% of"

040

Jamie McCarthy @jamiemccarthy.bsky.social · 18/05/2026

I did a double-take and snapped this one near Trafalgar Square in summer 2016. Initially I thought it was a symbol of intersex/trans support. Now I'm not sure what it was supposed to represent

A "walk" sign with a venus+mars superimposed symbol in place of the word "walk." Not quite ⚥ but similar

140

Jamie McCarthy @jamiemccarthy.bsky.social · 16/05/2026

Wow, those hieroglyphics are just *full* of typos

ChatGPT closeup of a hieroglyphics scroll and a book titled “Aquiracy”

1230

Jamie McCarthy @jamiemccarthy.bsky.social · 11/05/2026

I think you’ll have to admit this is a great answer from Claude tho

2140

Jamie McCarthy @jamiemccarthy.bsky.social · 08/05/2026

It's not an "'AI' problem." It's a you-picked-the-wrong-model problem. Opus 4.7 gets this right without any searching.

Claude Opus 4.7 responds:

Coke. Specifically Coca-Cola — Scalzi is famously a Coke Zero drinker, not a coffee drinker. He's written about this on Whatever multiple times over the years; he doesn't drink coffee at all and gets his caffeine from Coke Zero.

If you want me to dig up a specific post where he's written about it, I can search.

110

Jamie McCarthy @jamiemccarthy.bsky.social · 04/05/2026

In particular, your posts reminded me of this

I was once with a group, and the question came up whether someone as evil as Adolf Hitler deserves eternal punish-ment. Iwo of the members got extremely excited and shouted,
"Absolutely! I would wholeheartedly put Hitler in hell forever!"

Then one of them added, "And I would add Ronald Reagan! He is doing as much harm as Hitler!"

was somewhat amused by the savagery of their response, since neither of the two members are particularly evil. Furthermore, I believe that in a certain sense, their responses were not very accurate.
If it actually came to a showdown and God asked either of them,
"Should I really put Hitler in hell?" it is possible that each might answer yes, but I'm certain that if either one of them had to watch Hitler's agonies in hell for a minute or two, they wouldn't be able to stand it; they would cry, "All right, God. Enough!"

020

Jamie McCarthy @jamiemccarthy.bsky.social · 04/05/2026

I think you might enjoy the book _Who Knows?_ by Raymond Smullyan, 2003. He has the most straightforward commentary on the doctrine of eternal torture I've read, and (I think correctly) identifies it as the most important question to understand someone's religious belief system

To put the matter bluntly, I very much resent the idea of a God meddling in our affairs by punishing us for what we do to each other. Now, if God wishes to protect us from being harmed by oth-ers, then I would of course be most grateful. But punishing us in the afterlife for what we do to each other in this life hardly constitutes any protection! What would you think of a parent who stands by watching one of his children brutally mistreating another, and making not the slightest attempt to prevent it, but then later brutally punishing the guilty one? And isn't this just about what God does in the Jewish, Christian, and Moslem religions?

Soft and Hard Christianity
Iwould say that the question of most practical importance in Christianity is whether the doctrine of eternal punishment is true or false. If true, how can the universe be anything but an infinitely hideous nightmare? How could all the joys of heaven compensate for the eternal suffering of even one soul in hell? And if the sufferer has been wicked in this life, does that make the matter any better?
One can hate the actions or thoughts of an evil being, but how can one not sympathize with his suffering?
But, of course, just because those like myself see it that way-just because we regard the doctrine of eternal punishment as hid-eous—doesn't mean that the doctrine can't be true! We can hope that it is not; we can have faith that it is not—but that doesn't mean that it is not! Now, suppose it is true. Then comes what I regard as the most important question of all, and upon the answer depends

140

Jamie McCarthy @jamiemccarthy.bsky.social · 04/05/2026

The *Killing Time* makes short work of some ships pretty quickly, yes

170

Jamie McCarthy @jamiemccarthy.bsky.social · 29/04/2026

Haha, I thought that gag was funny enough to capture too

010

Jamie McCarthy @jamiemccarthy.bsky.social · 16/04/2026

Superior sensibilities surmise similarly! I didn't ask for a "sestina" by that name, but here's me testing ChatGPT with identical prompts in 2023, 2024, and 2025

Dec. 2023:
Surely shears slice, silently severing strands so small,
Sovereigns' secrets, subtly snipped, silently sprawl.
Sorrow's shadow, silently sweeps, seeking solace,
Sacrifices, serenades, soothsayers’ solemn malice.
Sages, soldiers, silently swear, steadfast, stoic,
Serenity's semblance, shrouded, survives heroic.

June 2024:
Silken strands sliced, a silent sorrow's sheen,
Sacred shears sever, stealing sacred scene.
Soft sighs summon shadows, swift souls' strife,
Star-crossed sacrifice, stilled by sharper knife.
Shattered serenity, swift savior's stand,
Sublime solace sought, still, in somber sand.

Oct. 2025:
Solemn salon steadfast stylist sets scissors; sable strands submit shear.
Star-crossed sweetheart; silent scalp salutes soft steel sear.
Sly serpent-schemers stage secret sedition; split-ends sins seek smite.
Stoic shorn sentinel stands; sorrow shame spite.
Sacred sweet steadfastness; soft-sung solace shapes solemn sacrifice; seconds signal soon.
Storm-still shorn sovereign smiles; surest silence shrouds sundown swoon.

020

Jamie McCarthy @jamiemccarthy.bsky.social · 30/03/2026

What physical mechanism do you think could cause running (say) 1 GW through a building to increase the land surface temperature 10 km away?

"Taking a look to these results, it is evident that the impact of LST increase reaches up to 10 km
distance from the AI hyperscalers. The data heat island effect seems to reduce its intensity to 30%
within 7 km around the data centres. In particular, an average monthly LST increase of 1 °C can be
measured up to 4.5 km from the AI hyperscalers."

140