tuukkah

The New York Times is publicly using generative AI, but responsibly: "We should tell readers how our work was created and, if we make substantial use of generative A.I., explain how we mitigate risks, such as bias or inaccuracy, with human oversight." https://www.nytco.com/press/principles-for-using-generative-...

OKFN and OKF are the same thing: "Open Knowledge Foundation (OKF) is a global, non-profit network that promotes and shares information at no charge, including both content and data [--] founded by Rufus Pollock on 20 May 2004 [--] Between May 2016 and May 2019 the organisation was named Open Knowledge International"

You had started your AI project before 2004?

By coincidence, I came across this old book today: "2000 Tips for Teachers” https://www.taylorfrancis.com/books/mono/10.4324/97813157825...

Another useful keyword might be "patterns": https://teachingpatterns.org/ https://learning-patterns.com/ https://www.jeckstein.com/pedagogical-patterns/

Yes, I think it will be somewhat random how many cases each child needs to be taught before they generalize. One thing I forgot to mention is that for polysyllabic words, they use a learning aid here: in the beginning, the syllable boundaries are marked with dashes in all text and in the second year, only in longer words. I don't know if it works in Spanish too, something like this: "A-yer ce-na-mos cons-ter-na-dos." Again, I think it's unnecessary for some children but helps others to keep learning while gaining more experience with the syllables.

In English, you never know: "a unit"

I don't know if there's a book literally like that, because the topic spans multiple university degrees and you should have the theoretical background to recognize situation X as well as to apply intervention Y. Even small children's brains are very complex and yet they lack the self-awareness, reflection and communication skills that would help the teacher in "debugging" the (potentially entangled) issues.

I had a look around for resources in English and this site seems highly valuable in math and in general: https://www.understood.org/en/topics/math

Of course, you don't have to believe me; you could also read about the orthographic depth of English vs other languages on Wikipedia or something. "In shallow orthographies, the spelling-sound correspondence is direct: from the rules of pronunciation, one is able to pronounce the word correctly.[1] That is to say, shallow (transparent) orthographies, also called phonemic orthographies, have a one-to-one relationship between its graphemes and phonemes, and the spelling of words is very consistent. Examples include Japanese kana, Hindi, Lao (since 1975), Spanish, Finnish, Turkish, Georgian, Latin, Italian, Serbo-Croatian, Ukrainian, and Welsh. [--] English is unusual because it combines deep orthography, with multiple possible sounds for many letters.[2] This makes it among the most difficult languages in the world to learn to read." https://en.wikipedia.org/wiki/Orthographic_depth

Practical pedagogy is called didactics and primary school teachers (should) learn a lot of it: after all, a child's brain is still quite undeveloped and you cannot teach them like you would teach a peer or yourself. E.g. you cannot teach grammar rules but you can teach a foreign language through singing, learning games etc. https://en.wikipedia.org/wiki/Didactic_method

These approaches work for most children most of the time, but when they don't, you have special education teachers who have a different degree in diagnosing (debugging) learning difficulties big and small as well as implementing interventions etc. The service they provide is also called remedial education [and it's especially cool when a primary school teacher and a special education teacher work in a big classroom together, the latter immediately bringing back to speed anyone who grasped something slower than the others]: https://en.wikipedia.org/wiki/Remedial_education

No, comprehension is again orthogonal to reading as it is needed in spoken language too. Remember: everybody's (unless you have a severe speech or hearing impairment as a baby) first language is a spoken language, and to extend it to a written language you will have to learn (be taught) an artificial orthography to map to and from. You can know a lot of vocabulary before learning to read (if you ever do). In a language with a phonemic writing system, your sets of spoken and written vocabulary are the same, whereas in English they only overlap. In both cases, knowing a word is orthogonal to knowing its meaning(s).

I think that only applies if learning to read in English. In a more phonemic writing system, if you can read, you can read any unfamiliar word too. This way, even a young child can read anything to quickly acquire more vocabulary unrelated to what their parents ever acquired.

"We used to think dyslexia was related to IQ but it's largely orthogonal." More specifically, dyslexia is not a generic learning disability, it's a highly specific one (and mostly preventable).

I think you misunderstand. In a largely phonetic language, almost everyone learns to read in one school semester, after which it's a fully solved problem - no spelling bees or anything. Peculiarly, you don't need spelling bees either when learning English later. ("Contronyms" and "words" are orthogonal to reading as they apply to spoken language too (and it's very much automatic).)

You should think "the step from single sounds to syllables", and the way to do that is to begin with the easy syllables like "tu", "mi", "el" (not unlike the multiplication or addition tables) before moving to longer ones. [And note that M alone is not "em", it has to be "m" when learning to read - a common pedagogical mistake! M + I makes "mi" not "emi", so M must be "m".] At least that's how children are taught in Finnish schools since sometime before the 1980s, and since then almost everyone learns to read during the first school semester. Also, one simple and efficient protection against dyslexia is to play the Graphogame (or similar) to get a lot of repetition with the sound-letter correspondences while learning to read (for various reasons, some brains take longer to build the necessary connections and you want to avoid the negative affects of learning slower than your peers if you can).

Quote from a sibling comment:

  - macOS: Accessibility API
  - Windows: UI Automation
  - Linux: AT-SPI

Precisely! This makes me think that in 1930 there was a big difference between what had been written by then and what was generally available and known. Today, we have access to letters that scholars wrote to each other etc. Also, a 1930s encyclopedia was no Wikipedia (which still has gaps) and had no entry on Ada Lovelace (for multiple reasons).

If I ask the same questions, sometimes it answers with information on "computing machines", "arithmometers" and even Charles Babbage - but it does not know Ada Lovelace.

When it doesn't feel enjoyable anymore, you're supposed to get out of the sauna and cool down - preferably in a lake. Then repeat as many times as you like.

If you want to help orgs who have come to the conclusion they need to diversify to EU services, it does not mean you have to have come to the same conclusion! Also, it's not the same kind of dependency if you get or buy something one-off from a website like this, as if you build your org on top of a single platform/jurisdiction.

That's great news. Thank you for sharing the resource!

Has the WebVTT story changed? I once tried to customize the subtitle rendering but it seemed too difficult.

AIs will reverse engineer the processing algorithms based on observing a few example inputs and outputs...

Isn't that same as a jpeg then?

Edit: There are some parameters:

> FilmKit communicates PTP (Picture Transfer Protocol) over USB, the same protocol that X RAW STUDIO uses. The camera does all the heavy lifting: it receives the RAF file and conversion parameters, processes them, and returns a JPEG.

In their article, AMD chose to use the built-in 5Gbe NIC, but I read you could use the two USB4 @40Gbps ports as interconnects too (or for USB NICs).

Meanwhile, the corresponding "non-standard" desktop PC is the Framework Desktop, which with the Ryzen AI Max+ 395 can use 120GB of its 128GB RAM for the GPU: How to Run a One Trillion-Parameter LLM Locally: An AMD Ryzen™ AI Max+ Cluster Guide https://www.amd.com/en/developer/resources/technical-article...

Here's BART: https://www.traintrackr.co.uk/product/bart3

Thanks! Perhaps one of the sites that track train delays can give a statistic?

This list doesn't have anything that looks relevant: https://www.rijdendetreinen.nl/en/disruptions/archive?date_b...

The day does not appear as an outlier in the monthly statistics: https://www.rijdendetreinen.nl/en/statistics/2025/10

I don't find a detailed statistic on the overall delays, but the per-station statistics for Amsterdam Centraal say 5% of trains were cancelled and 17% were delayed by 5 minutes or more (mostly by 10 minutes): https://www.rijdendetreinen.nl/en/train-archive/2025-10-29/a...

Could you provide a reference about the train disruption? I tried but couldn't find anything in English.

Considering that it's an open source tool, I don't know if it's that bad to be shilling for the commons, basically.