Building a real-time AI tutor for 5-year-olds • Hacker News Board

Re: Building a real-time AI tutor for 5-year-olds

#321

Post by zero-sharp » Fri, Jul 10, 2026, 4:22 PM UTC

ndriscoll wrote: post #275

zero-sharp wrote: post #269

Earlier quoted context omitted.

Are you asking me to believe that the educational system underserves kindergartners because the average kindergarten teacher can't provide an education on your narrow technical interest? The responsibilities of the average kindergarten teacher probably include 1) making sure your kids don't swallow glue, and 2) making sure little johnny doesn't throw another tantrum.

[flagged]

That's one interpretation: the system doesn't take of your specific needs and therefore it's neglectful or inadequate. Another interpretation is that the teacher is doing exactly what the school needs them to do because most 5 year olds aren't concerned with how air conditioners work. And perhaps it shouldn't be the responsibility of the kindergarten teacher to provide technical education?

There's a conversation to be had about the educational system underserving the intellectually curious. Trying to make that point in the context of kindergarten is a little absurd to me.

Re: Building a real-time AI tutor for 5-year-olds

#322

Post by 27183 » Fri, Jul 10, 2026, 4:24 PM UTC

duskdozer wrote: post #300

wizzwizz4 wrote: post #286

Earlier quoted context omitted.

I would suggest that your ability to discern when the model is wrong might be the thing that has changed.

This is what it seems to be. I've had to review the LLM output from coworkers who have insisted that the "old" models are terrible but the new ones are very good. They are always discussing the latest models and their tweaks and things. Their output is just.... not very good. For example, it may not fail tests, but it does so for wrong reasons and will fail under different conditions. Or it checks or guards for thing…

It would be extremely valuable to be able to reliably identify this kind of behavioral tendency in an interview setting. Some kind of reliable test that screens for likelihood to go AI psychotic.

Re: Building a real-time AI tutor for 5-year-olds

#323

Post by cognitiveinline » Fri, Jul 10, 2026, 4:25 PM UTC

drayfield wrote: post #208

This is one of the most insidious projects I've ever seen and despite the touted credentials of the backers looks to be completely divorced from the realities of how children should be taught.

[flagged]

Re: Building a real-time AI tutor for 5-year-olds

#324

Post by duskdozer » Fri, Jul 10, 2026, 4:27 PM UTC

Kaggbac wrote: post #215

The async planner is the part I'd love more detail on. If it's reasoning ahead while the kid is still talking, some chunk of that work gets invalidated every time the kid says something off-script, which at age 5 is presumably constantly. Is the speculative planning you throw away a real cost line or a rounding error? I build streaming LLM stuff (nothing this ambitious) and the "right move at the right moment" framin…

LLM-generated comments are not permitted on this site.

Re: Building a real-time AI tutor for 5-year-olds

#325

Post by blitzar » Fri, Jul 10, 2026, 4:36 PM UTC

duskdozer wrote: post #308

blitzar wrote: post #288

Earlier quoted context omitted.

Welcome to silicon valley. Enjoy your stay.

I need to stop coming here; I keep "dead dove, do not eat"ing myself.

Would you prefer someone wrote "5 Michelin star meal, save for lunch" on the bag?

Re: Building a real-time AI tutor for 5-year-olds

#326

Post by duskdozer » Fri, Jul 10, 2026, 4:42 PM UTC

blitzar wrote: post #325

duskdozer wrote: post #308

Earlier quoted context omitted.

I need to stop coming here; I keep "dead dove, do not eat"ing myself.

Would you prefer someone wrote "5 Michelin star meal, save for lunch" on the bag?

No, no, it's on me. I guess I need a HUMANS.md file to pop up and remind me every time I sit at the computer....

Re: Building a real-time AI tutor for 5-year-olds

#327

Post by senordevnyc » Fri, Jul 10, 2026, 4:51 PM UTC

TonyAlicea10 wrote: post #267

Marha01 wrote: post #246

Earlier quoted context omitted.

> It's a classic example of the kinds of questions LLMs get wrong. There's plenty of others. Not sure your point here. We can easily find things a 4-9 year old will talk about that an LLM will get wrong, hallucinate, etc. My point is that the hallucination rate of modern LLMs, while not zero, is so low that this is no longer an issue in practice. > There's a lot of work being done in 'could'. And it's entirely ignori…

> My point is that the hallucination rate of modern LLMs, while not zero, is so low that this is no longer an issue in practice. That's just not true. Talk to an LLM about a subject you know deeply. They make up things all the time.

With stuff that a 4-9 year old is going to be learning? Color me skeptical.

Re: Building a real-time AI tutor for 5-year-olds

#328

Post by senordevnyc » Fri, Jul 10, 2026, 4:55 PM UTC

fourside wrote: post #280

havercosine wrote: post #164

Earlier quoted context omitted.

Thanks for taking time to write this. HN is showing (expected) dismissing attitude towards this idea. That tells me it might work :) ! Folks here are wildly overestimating (or ignoring!) how many adults are qualified to be good teachers and how many of them further have enough incentives (money, time, resources) to do it well. Its a _very_ small number. In my part of the world, "become a teacher" is often a job advis…

> HN is showing (expected) dismissing attitude towards this idea. That tells me it might work :) ! I recently learned the name of this logical fallacy: Galileo gambit. It’s one of my pet peeves. Yes, we all know Dropbox was infamously dismissed by the top comment on HN. No, just because someone criticizes your project on here doesn’t mean it’ll be a big success. Instead of fixing the education system and giving it th…

The Galileo gambit is the idea that because experts reject an idea, it must be true. HN is the furthest thing from experts when it comes to early childhood education.

Re: Building a real-time AI tutor for 5-year-olds

#329

Post by bluefirebrand » Fri, Jul 10, 2026, 4:59 PM UTC

lern_too_spel wrote: post #274

bluefirebrand wrote: post #203

Earlier quoted context omitted.

> Most critiques of AI are based on experience with the cheapest AI available a year ago, by midwits You don't even need to use AI at all to have valid critiques of it Plenty of other people are showing its flaws daily by replacing their entire personalities with AI slop Oh it can do some impressive stuff sure. It is still bad for society

To make my earlier critique of your argument clearer, some applications of AI might be bad for society, but it does not follow that the tutoring application is also bad for society.

A responsible society would make sure that the applications have good results with adults before even thinking about applying it to children

And despite the hype, the jury is still out for how the results are for adults using AI

Re: Building a real-time AI tutor for 5-year-olds

#330

Post by senordevnyc » Fri, Jul 10, 2026, 4:59 PM UTC

imp0cat wrote: post #130

m4rtink wrote: post #126

Earlier quoted context omitted.

Yeah, hard to get a children army going without cheap personalized AI tutors. ;-)

Yeah, it's like Diamond age, except it also destroys your kid's vision. But then again, that's a potential upsell just waiting to happen - do you want eyeglasses with that? Your kid is gonna need them.

This is totally false. https://www.aao.org/eye-health/tips-prevention/screen-use-ki...

Unless you're also advocating that kids shouldn't read books?