So here’s a gross generalisation: people don’t know what they want, and when they do know they can’t explain it.

I think folks who live in spec-driven worlds just don’t appreciate how poorly understood the capabilities and processes of computing and networks are, to the point where even if you gave people hallucination free LLMs you’d probably find people just frustrating themselves because they can’t figure out what they want, what it should do, and how it should be used.

Quite apart from the huge chunk of the app ecosystem that relies on common protocols, centralised services and so on.

> rustrating themselves because they can’t figure out what they want

I have seen this with yet "AI-unexperienced" users - for simple questions or textreformating, they are absolutely fine with the Chatbots. But thats all they can get out of it: Above that level their experience is very often frustrating due to too less understanding while having high expectancy towards the model in reverse (like "its AI so it should know...")

And this people end up then not using any of the AI tools beyond that because they do not "believe" / trust that this thingy can get work done.

> * Based on our music discussions, make a playlist for me in Spotify. No songs older than 15 years.

It is easy to express yes, but is this how people think about what they want to listen to?

Someone can already do this in a chat bot, see the list of things and then find it on Spotify or whatever their favorite music service is.

What the chat bot won't know is what they have listened to, which is probably a much higher predictor of what you'd like than the discussions you've had about music. You've probably had those discussions because you are already listening to that music!

You just explained what you want far more clearly than most people do.

Only the fourth of those is not more easily accomplished using existing tools; the things you're describing as explainable are such because they've already been achieved.

Once the AI does something, I want to follow up. It's just the start.

All of your examples seem to require using another company's app.

I thought the parent's point is that the App Store will loose dominance (because you won't need apps).

That's right. I think AI agents will access services without you going through apps.

> No songs older than 15 years.

That’s a surprising requirement. Is it a proxy for “less likely that I’ve already heard it”? For “the band probably still tours”? For “I can discuss it with friends without sounding like I’ve been living in a cave”?

> Example two: Voter ID. In every country in the world, you need ID to vote. Democrats have pushed to completely remove it for years.

Would just like to point out that here in .au you don’t need ID to vote (as in to show up and cast your vote, not talking about registering).

Is also like to point out that in other countries the national government issues ID to their citizens. The US doesn't have anything like that.

You also don't need ID to vote in Britain, which like USA does not issue mandatory ID cards (at least not yet).

I saw something similar in Google's ToS update that went out the other day:

> 13.2 Customer Indemnification Obligations. Customer will defend Google and its Affiliates providing the Services and indemnify them against Indemnified Liabilities in any Third-Party Legal Proceeding to the extent arising from (a) any Customer Application, Project, Customer Data, or Customer Brand Features; or (b) Customer's or an End User's use of the Services or AI Agents in breach of the AUP or Section 3.3 (Restrictions).

(From https://cloud.google.com/terms)

I used the Excel example the other day. What Excel gave us is a tool that was simple enough that people without a lot of technical expertise could wield it in anger, and so folks who have the domain expertise can do something useful with it.

It’s easy to look down on it as fragile, ugly, insecure or whatever, but by and large it works for solving people’s problems.

I see vibe coding the same way. There’s going to be a lot of shit, just like there’s a lot of shit spreadsheets, but I also think there’s going to be a lot of itch scratching.

Ugh. Every time I accidentally put a HEIC on my Windows machine, or one of my students tries to upload a WEBP to our school's LMS :/

Most of teaching kids to code is them understanding terminology, structure, and getting them unstuck. This has bits of information mixed with unscaffolded terminology all over the place, some hidden in collapsibles, and some just missing.

For example the `range` example in the second activity doesn't show a `for` loop's structure, how `range` is called, or what it does. One of the hints shows you the syntax for a `range` call, but nothing about loop syntax.

The hints are on request rather than at-need, and aren't linked to what's going wrong for them - doing some structural analysis of the code to decide what hint might be needed would make hints more useful. The couple of activities I looked at didn't have particularly useful hints that targeted real problems and were more of the "here's a snippet of completed code" variety, which doesn't teach anyone anything.

I'd recommend going and looking into some teaching pedagogy before building stuff, tbh.

The actual activity objectives don't look too bad, but the UI is a mess of the wrong sorts of information to help kids learn the principles.

I was curious about the tutor, but not curious enough to hand over an auth token to it. I've been building a writing tutor using local models for my kids, and it's been an interesting experience trying to provide helpful, structured information at the time they need it, rather than just asking for answers, or parsing generic advice.

your feedback is really helpful! yes, you are right on the range exercise. the course is not self-contained and I put more work on the lab but did not check on the logic sequence of terms and concepts. on the hints, I am curious what works in your earlier work on writing tutor.

on the AI tutor, I originally allow a user ask a few qeustions without requiring login, but Claude advised against it for security reasons (possible abuse of unauthenticated model relay/call)

The writing tutor is part of a home hosted web app that I've got to try and encourage my kids to get a bit more fluent at intentional writing, plus specific types of writing (creative, reflective, persuasive, etc).

Because I'm using pretty small models (biggest option I've been working with is a 14B) and primary age kids writing is really different from adults, there's a lot of very targeted prompts around mechanics, vocabulary, sentence and paragraph structure, general encouragement and so on. I've also got LanguageTool feeding spelling and grammar analysis into it, since small language models just make up so much dumb stuff when it comes to spelling (it helps that we have an Amazon Echo and they're used to dumb AI so don't expect too much of it).

The kids can "submit" their draft to get a bit of summative feedback which runs it through the whole suite and they can go through the different feedback contexts (highlights relevant passages, suggests word choices, gives general feedback related to the writing purpose and prompt). Each draft takes a snapshot so my wife and I can see what the kids change about the draft after the feedback. This part actually works pretty well.

There's also a "nudge" mode where occasionally the LLM tutor will surface a bit of feedback with a subtle notification while they are writing. This aspect sounded pretty good in my head but since it's analysing a writing fragment sometimes it's not super helpful and even though the notification isn't too disruptive, it still breaks into the flow.

The more I think about it, the more useful this sort of model sounds useful for code too, since it doesn't need to just be about "does the code run, are the outputs correct" but could reason a bit more about program structure, naming of variables and functions and whether they are appropriate to the purpose, quality of comments and whether they indicate an understanding of the code and so on. Most of that you could get with a decent AST analysis, but relevance of comments etc is something that could be quite useful.

Good luck with teaching Python to your own kiddo. I'm finding my own are more interested in programming microcontrollers like the micro:bit than graphics. Gets you really comfortable with fairly simple code and immediate feedback about whether you gets things right. See the micro:bit Python editor for a really well scaffolded IDE too (https://python.microbit.org/v/3).

If only they had some sort of marketplace to be able to sell them off at a reasonable price.

Im sending my Muse AI Agent to go negotiate a deal

I'm here to deliver in person. I'm wearing a blue blazer and red tie.

Hahah, smart boards. I remember pushing back hard against my school on those. The fact they evolved into short throw projectors that occasionally have ok touch monitoring is a win, even if it took a bunch of $5-10k surfaces that no one could draw on, but inevitably ended up with a few whiteboard marker smudges to get there.

Open gradebooks are amazing. You just need to have school leadership that sets realistic expectations with students and parents and backs up their teachers rather than caving to helicopter parents. I've had nothing but good feedback from parents and students about timeliness and transparency on marking feedback. Standardised testing on the other hand is goddamned menace that has done more to damage the value of assessment than transparency ever has. The value of open gradebooks is the feedback, not the numbers.

AI in schools feels like more of a fait accompli. Who provides most of the software infrastructure to schools? Microsoft and Google. Who is busy stuffing AI into all the things? Oh look, it's Microsoft and Google.

Sounds like it's just focused on shows and games from the description in the paper:

> At ages 1 and 1.5 years, which was conducted in a subsample, parents were asked to report the average amount of time that the child spent daily on television (TV) on weekdays and weekends within the past month. This was collected in 2010 to 2011 when TV was the primary source of screen viewing time for 97% of Singaporean families [38]. At ages 2, 3 and 6 years, parents were asked to report the amount of time, in 5-min increments, their child spent using screens on weekdays and weekend days. Three types of devices were considered separately: (1) TV and TV-related games consoles; (2) computers (desktop or laptop); and (3) hand-held devices including tablets, smartphones, and hand-held games consoles. At age 8 years, parents were asked to report on their child’s time spent sitting or lying down while watching screens on the most recent weekday (school days) and on the most recent Saturday and Sunday. Four types of devices/media were considered separately: (1) watching videos on a TV (TV programs, DVDs, Blu-ray); (2) watching videos or playing games on computers; (3) watching videos or playing games on tablets or smartphones; and (4) playing video games on consoles.

We’re having this sort of conversation at the school that I work at, which serves remote communities and other disadvantaged students. One of our states is currently making noise about reducing “screen time” for schools without much distinction about what it involves.

And here’s us, an online only school that serves kids that otherwise wouldn’t be able to access education at all or some types of classes at best, left wondering where that might leave us if our state decides to zombie its way into similar policy decisions.

Edit: Downvotes for pointing out that government policies without nuance around what "screen time" means hurts kids who have no alternative education options, apparently?