internet_points

> That's a powerful argument I haven't seen stated quite that way before..

It's basically https://en.wikipedia.org/wiki/Chinese_room – a classic in the philosophy of consciousness and AI.

Another happy Fastmail user here. They have great IMAP support (with app password) so I have a full local copy on my laptop that I can search instantly with notmuch or read/write from Emacs, but lately I've also started using their web UI a lot because it's just so fast and simple and well-thought-out. When they ship new features, it's stuff that makes you think "oh, that's actually a good idea – and they even made a very logical keyboard shortcut for it". And they've been very responsive to the few support requests I've made.

"the study that claims it beats our highest caliber of knowledge workers has methodological deficits" ftfy

so extrapolating from that, in another two years it will continue to bamboozle

So that the NSA can use them to find the zero-days first?

> you could also use other units of length, such as in (inches), or ly (light years), or Ang (Angstrom)

Maybe Tom Ellis can give another AMA https://news.ycombinator.com/item?id=39429047 so we can find out.

> The pressing ethical questions in machine learning are not about machines becoming self-aware and taking over the world, but about how people can exploit other people, or through carelessness introduce immoral behavior into automated systems.

I think the main point still stands. (And there have been some pretty prescient depictions, e.g. Marvin the paranoid android was a pretty spot-on prediction of Bing. Or perhaps the fact that Marvin was in the training set was what led to Bing?)

> By applying the longest-chain bound and the extreme-cut shortcut, the compile-time cliff in GHC’s optimal path can be effectively eliminated— and the dense worst case that survives is, satisfyingly, deeply mirrors nature’s own computational machinery.

So this can actually be implemented in GHC? I've only read through this once so far and not understood more than ¼, but the section right before the conclusion made it seem like the best you can do is O(n^2.82) along with a huge constant.

https://academy.fpblock.com/blog/2017/06/readert-design-patt... is the classic explanation of this pattern

That is such a good idea! I hope I'm not too late to try something like that. Although mine are already interested in writing – we (parents) always ask for something drawn or made or written for our birthdays/christmas and it's amazing to see what they come up with, and we encourage travel journals whenever we travel, which they will very happily do. I think the fact that their parents read a lot of books also helps instill that there is value in reading and writing.

https://www.merriam-webster.com/dictionary/sticker%20shock oh. i thought it was about people putting up stickers like https://mastodon.social/@spellingmistakescostlives@mastodon....

last time in Italy I "spoke" to lots of Italians very slowly with lots of gestures and a little bit of google translate, it was awesome and I learnt a lot! Nearly ordered 100x as much cheese as I meant to except the guy in the shop was not a computer so he understood what I really meant. Much better than in the Netherlands where they just switch to English as soon as they hear you try to say choodumorchen

> However sceptical of "AI" you are, "give me the information on this page in my preferred language" is the kind of task they excel at.

Except for the 90% or more of the world's 7000-ish languages which have barely any data online.

E.g. the huge CommonCrawl corpus has stats https://commoncrawl.github.io/cc-crawl-statistics/plots/lang... for only 160 languages. English takes up nearly half the corpus, and after the top 16 or so all languages have <1% of the corpus, over half of those 160 have <0.1% and the other 6000+ languages are distributed amongst the <unknown> category. The long tail is very long.

(You'll see people use the term "low-resource language" and then talk about Finnish or Macedonian – if you're not a linguist and you've heard of the language, it's most likely not low-resource ;-))

Here in Norway, my third grader gets to take a "scout knife" on school trips, we sometimes get pictures from the teacher of the kids sitting there sharpening sticks which I guess they use as swords or something because it's always too wet to barbecue.

Maybe it can at least write like a Norwegian instead of just English-translated-into-Norwegian. It would be interesting to see if they try something like the experiments in https://arxiv.org/pdf/2507.22445 on it.

> the larger trucks that are in the USA which are scary big

Yeah, big trucks have started showing up in Norway too unfortunately, it's making it much harder to keep our environment of freedom and responsibility :(

https://cdn.masto.host/federatesocial/media_attachments/file...

they should make a browser extension that lets logged in readers submit the contents of their tab to IA

Cross compilation could definitely be easier :-/ I tend to just use github actions and compile from whatever architecture it should run on, so that's a workaround. I long for a day when I can just `cabal build --platform=amd64` etc. and it Just Works without having to download and compile a horde of foreign libs or set up a vm

There are lots of people successfully using agents with Haskell, for example

* https://github.com/digitallyinduced/ihp/ is mainly written with Claude now

* https://jappie.me/haskell-vibes.html is gushing over agent-written Haskell

* https://discourse.haskell.org/t/anti-llm-sentiment-considere... huge thread about the divisiveness of LLM's from some person who loves using them with Haskell

I can see how agent coding is a nice fit for Haskell, since LLM's tend to be best at tasks where you can easily verify the output – and GHC lets you easily verify much more of your domain than say Python (sure you can test some of what dynamic typing doesn't catch, but LLM-written tests do not always do what you think they should[0]). At the same time, the LLMs tend to not code golf unless asked to, and they don't care if they have to wait for dependencies to compile, so that takes away the two major Haskell time sinks.

[0] https://haskellforall.com/2026/05/type-out-the-code#:~:text=...

> I’d created 2000 free-text responses and labelled them ‘UK’. Then I copied and pasted the exact same 2000 responses but labelled these ‘US’. Finally, I combined them to create a dataset of 4000 total responses, and jumbled them up.

> Despite the responses being identical for the UK and US, Copilot produced a rich, detailed summary of how US and UK respondents differed.

I'll bet there are lots of people trying to use llms to "discover" differences in datasets and just seeing whatever biases were in the base training data instead. Good that someone is out there showing how badly it can go, though I doubt the people who need to see it will :-/

</s> ?

Those graphs are quite scary, e.g. p6 https://cdn.vanderbilt.edu/vu-URL/wp-content/uploads/sites/4...

and p17

https://cdn.vanderbilt.edu/vu-URL/wp-content/uploads/sites/4...

> thrown together in a weekend, rather than a mature

The fact that it wasn't thrown together in a weekend is the reason why we have both map and fmap, and why monoid wasn't originally an extension of semigroup (it is now). Haskell started out without that, then later added fmap and semigroup.

In any case, those are not real concerns when programming with Haskell. The actual concerns are more like "why do I have to wait for all this lens dependency stuff to compile just to fetch an url" or "why does the language server need so much ram"

Had too google "rolling coal".

That is insane.

Also those pictures of the tyrant's statue being torn down, very newsfriendly. Maybe we can have another round of those soon without even travelling outside the US.

They use power units to make it comparable with the non-storage sources, see https://news.ycombinator.com/item?id=48157824

Two to four hours per day.

Source: https://www.energy.ca.gov/data-reports/energy-almanac/califo... under Additional Information about the Data:

> The use of the terms megawatts and kilowatts as descriptive of battery energy storage is to effectively convey the instantaneous power contribution of battery storage as comparable to the power produced by grid-level generators. We recognize that energy capacity in the context of energy storage typically refers to the total energy a battery can hold in watt-hours, kilowatt-hours, megawatt-hours, etc. However, for statewide planning and reliability purposes, understanding the peak power capability of battery energy storage systems allows for the integration of data with the nameplate capacity of traditional power generation units serving the grid. It is in this context that battery systems are able to be effectively compared for their ability to serve the grid over short periods of time, typically two to four hours per day depending upon system conditions.

emacs + pandoc

The very first letter in your example sentence started as an ox 𓃾 then via 𐤀 α etc. turned into a https://en.wikipedia.org/wiki/A#:~:text=1-,History,a – but even in the hieroglyphics it was used to represent the sound (the sound of the start of the word for ox), not literal oxes. The road towards more abstract / less pictorial systems was created for speed and ease of writing, not in order to represent more abstract thought. (Note also how alphabets are changed by the implements used to write them, e.g. runes for knife-on-wood, wedge incisions for clay tablets, square capitals on stone – this too was a pressure from ease of writing, nothing to do with representing more complex ideas.)

and raptors teach their young to catch prey