I can only hope that there will be some fun class action lawsuits against companies that get refunds but don’t find a way to pass the money back

in four fours, two eights, or one contiuous burst?

(I intuitively go with 10: 5 then 5)

So a modern Name of the Rose?

How would you tie this to Kevin Lynch’s “Image of the City,” which explores how city designers can build their cities specifically to help make those mental images easier to build and keep?

The risk with an AI UI is that there's no external way for the user to know how to translate your marketing copy into a prompt.

The AI told me that I should pick an airport, a departure and arrival date, and that its "travel agent" functionality was because it could give me a list of things to see in my destination city.

Oh.

It's a chatty copy of Lonely Planet -- not what I think of as a travel agent.

"... I could fly into Madrid, but open to something else -- Barcelona, perhaps, or even a smaller airport in northern Spain if there's a good window. Figuring about ten days on the road before and after, and I'd love to add in a few days of rock climbing, if I can pass by a good area for it."

Not long ago, I used a travel booking tool that advertised itself as an AI "digital travel agent." I gave it the sorts of things that I miss doing with a real travel agent:

"Hey, I'd like to figure out a schedule that passes near either Picos de Europa or southern Spain near the eclipse..."

I want to talk a little more about discoverability. I often find that the creators of AI-driven services have a very specific set of scenarios in their minds -- but then advertise extremely broadly.

Unlike a button on a UI, there's no way to know what the creators have in mind!

On the other hand, a button or a short form fill is familiar, responsive, and expressive.

Don't get me wrong -- there are great times to use AI inputs, especially when the topic is confusing and ambiguous. But it comes with tremendous risk compared to a well-designed button.

This is a guessing game -- "can the user guess the scope of the tool?" It's the opposite of good discoverability! Worse, the user needs to type an entire sentence -- "please book me a plane flight on Wednesday from Dallas airport...", so its inefficient! (And it's almost always slow.)

Chat is great for expressing (and understanding) a broad diversity of possible messages, but many services don't do a broad diversity! They do a small handful of things.

My consultancy currently seems to have a solid line on customers who were pretty sure that AI meant that they didn't need a UI anymore. Then they're confused by why users don't get it.

Real-time collaborative editing *and* rebuilding for feature equivalence? It's two different Portents of Doom in one project!

(I'm only half joking -- good luck, and I hope this goes smoothyl and well!)

Ok, I'm curious. I'm watching several different teams and techniques, and I'd love to know where you're landing.

Dos Alito think it should be illegal to file your taxes before tax day?

It feels like the next step is for judges to order — at the moment when the first haves petition is filed — that the DOJ attest on penalty of contempt that they will have a response, or to release the prisoner.

That would help identify the baseline population, but doesn’t quite help with the deceptive placement of the dots: they look like locations. Perhaps a stipple pattern that doesn’t connote “place” as much?

Where does this leave us?

Anthropic's report is very interesting, and there's a lot to work through in the data. I'm excited to read it.

But the visualization suggests a little bit of playing fast-and-loose with the results -- and that instantly diminishes their credibility a little.

A map of South East Asia. There are no dots in China or Japan (but a smattering in Mongolia); South Korea is very densely filled; Taiwan and Japan slightly less so.

South American countries have a uniform level of dots ... but Nicaragua, Venezuala, Guayana, Suriname, and French Guiana are all completely empty.

The map also shows India with uniform dots. Apparently, while the US is broken into states, it's uniquely so.

We can also learn about the survey by seeing just a few dots in Mongolia, and none in China; Japan, Korea, and Taiwan each are densely filled. South America tells its own story.

Heatmap

Naively, I would expect this map to be a population map: more people live in populated places, and so respond to surveys.

The designers seem to have made a different choice -- one that sort of shows population, while also making lots of intentional decisions.

They give a sense of real science -- look, we got people from all over the place!

They're also deceptive.

See that clear California border? People are not uniformly distributed across the state of California -- they live mostly in the greater Los Angeles and San Francisco areas.

A map of the united states, purporting to show locations of survey respondents. The outlines of the states of Washington, California, and Florida are densely filled.

The caption: "Each dot represents 4 respondents"

I'm interested in Anthropic's Qualitative AI study, with lots of questions about both what they learn and how to use the results.

www.anthropic.com/features/81k...

But from a visualization point of view, there's something I noticed that's bothering me.

Those dots are impressive.

Not sure. East Germany was recognized by the UN, as was Yugoslavia and Czechoslovakia. Did they have a right to exist?

Palestine is recognized by much of the UN.

Students graduate based on grade, get jobs based on product, and are told to hurry though process because it gets in the way of the other two.

Guess they’ll just have to keep reprinting the old photos

Anyone who would say yes didn’t see this message

Also you probably should anyway

Paul Allen built a museum largely as a tax deductible way to show off his collection of sci-fi memorabilia, bought and restored an old movie theater, and largely seemed to spend his money on things that struck him as nifty — from science institutes to owning the Seahawks. You could do worse.

You worked for Joe Nand, or Bob Xor?

By this argument, no brief that speaks of morality, economic effects, social interests, or literally anything but “the law” should ever be considered