Rendered at 23:20:03 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
shepherdjerred 2 hours ago [-]
ChatGPT/Claude are _amazing_ for learning.
On one hand, it is very tempting to use AI to just do your schoolwork for you.
On on the other, everyone now has access to an expert-level tutor in any subject they possibly care to learn about.
As an example I've been reading Build a Large Language Model (From Scratch) -- I take pictures of the book and ask Claude to help me understand concepts. I chat with it to ask questions and check my understanding. I don't get stuck because there's always "someone" to ask.
malfist 1 hours ago [-]
An expert level tutor who you'll never know if its wrong
doctoboggan 54 minutes ago [-]
This was an interesting practical concern a few years ago, then moved to a theoretical concern, and today is basically old tired trope.
Tar90 10 hours ago [-]
I've been experimenting with a lot of these learning skills (see also, e.g., https://github.com/mattpocock/skills/blob/main/skills/produc..., which I think addresses the progress tracking part better than this skill) and they're great for getting simple answers, getting unstuck on exercises, getting some intuition through simulations, etc, but I still feel some more structure is needed for effective self-teaching. Has anyone tried using a hard book as a sort of course structure and asking the LLM to teach you through it? You could of course just read the book, but a lot of advanced subjects don't have great learning materials.
goodmythical 8 hours ago [-]
>Has anyone tried using a hard book as a sort of course structure and asking the LLM to teach you through it?
I think LLMs are going to fall short for the same reason that the hard book itself does. At least with regard to genuine fluency in a given topic.
That is, even if the text includes challenge questions and an accurate rubric, the questions are always going to be a limitted application of the total use of the information.
Once you're getting to the "evaluate" and "create" portion of Bloom's taxonomy, the text can no longer tell you whether or not you're on the right track because the book cannot understand your appraisal/defense of an argument/implementation.
The LLM, too, will either "good, now you're getting it!" or "close, but not quite" based on system prompt if you submit a multi-paragraph or whatever answer.
All of which, I think, is why the link you posted ultimately directs the student to a community: learning in a vacuum is limitted.
jerkstate 6 hours ago [-]
> The LLM, too, will either "good, now you're getting it!" or "close, but not quite" based on system prompt if you submit a multi-paragraph or whatever answer.
I'm not sure why you think this. have you tried it?
lacedeconstruct 7 hours ago [-]
The biggest advantage of LLMs is they can produce the exact background information you need to get unblocked, I found it very useful while trying to learn DSP and the math required
myworkaccount2 51 minutes ago [-]
I like to think that i am decent at self studying, but i feel like i learn in a weird manner.
To understand anything new you need to first have a base of knowledge to build on top of.
I find it helpful to think of the new concepts as a high dimensional space where specific facts / concepts lie in.
The first step: is to jump right in, take a fact / concept and read a paragraph or two about it. You have not really learned anything at this point, and you may not even fully understand what you ready that's fine.
{Say you want to learn about dot product, you may learn how to calculate it mathematically / mechanically}
(Note if you did not understand anything, it may be a sign that something in step 0 is missing)
Next step: is to read about it again. But this time find a different angle (I call it a different "projection", we are projecting this unknown knowledge vector onto the basis vectors of your pre-existing knowledge base.) look at it form a different angle. It helps to rephrase the content in your own words, read different definitions for it, try to explain why different ways to describe the same thing can be correct at the same time.
{Instead of thinking the dot product as sequence of multiplication and addition, think about the classic visualization of the product of the magnitude of one vector and the magnitude of the projection of the other onto the first [1]}
Step 3: Repeat step 2 until you understand it. (Or even better you can come up with a completely new way of looking at the concept)
Step 4: Now you are already familiar with the idea, yet this step is the most critical for continual learning. Ask Why. Why do we need this? Why describe the same thing in 10 different ways? If you can understand this you have added a basis vector to your arsenal of knowledge! From the previous steps you made yourself familiar with the shape of the knowledge vector, step 4 will put more context around it, and show it where it fits in your pre-existing knowledge base.
(Pro tip: mix step 3 and 4 as needed! no need to stick to a strict sequence here)
step 4 is difficult, as there may be may applications for a concept, in such cases it helps to understand the patterns between the applications. What is causing all these different things to rely on this one fact?
---
edit: forgot to add the point of the comment lol
The point is llms are good at generating / regurgitating these different view points. They are also excellent quickly checking your ideas / view points.
ShinyLeftPad 6 hours ago [-]
Using llms to learn is a crutch because the best way to learn is by applying knowledge and doing and those things are offloaded to llms.
supern0va 5 hours ago [-]
If your goal is learning, then offloading those things to the LLM is a choice you can make, but a poor one. You can choose not to offload those things, do them yourself, and use the LLM for evaluation. You know...like a tutor.
LLMs will wear whatever hat you throw at them.
nico 4 hours ago [-]
Super interesting, and thank you for the repo link
I haven’t done it with these questions, but I have done some experiments with a local language model I’m working with
I compare the model just doing typical RAG, vs the model using a tool to “pre-study” the material first. The tool breaks down the material into basic question/answer pairs, and then gives the model a way to access that knowledge on inference
In my specific use case (a small agent that can talk about Wikipedia articles), it works really well, and the approach seems mostly limited by two factors: first, the quality/variety of question/answer pairs produced by the study tool, and the ability of the model to use the stored knowledge coherently in conversation
So, on the point above about quality of question/answer pairs, I’m wondering if these learning skills could be applied
hxugufjfjf 3 hours ago [-]
Yes that’s what me and others are trying to find the best possible way for with LearnHarness. In the Discord someone just shared a plugin for this https://github.com/kerr-mwv/learning-path-plugin
I’m not the developer or anything; just really enthusiastic about this because I’ve already learned so much with LearnHarness. Link to repo: https://github.com/amosblomqvist/learnharness
I get what you saying but I still feel this is a methodology of how to approach learning. Put this into something like NotebookLM and add an option for reviews using spaced repetition and you could have an awesome product for students.
ezrabuenk 5 hours ago [-]
I'm a university student, doing a bachelor in CS and I think you are on to something. I've studied for a bunch of courses by working through the entire required syllabus with LLMs working as a sort of TA. It works wonderfully and makes learning by yourself a lot quicker and less frustrating
guiraldelli 7 hours ago [-]
> Has anyone tried using a hard book as a sort of course structure and asking the LLM to teach you through it?
@ivansavz has done that as a "teaching assistant" on his statistics lectures from his book No Bullshit Guide to Statistics.
I have particularly benefited a lot from his agentic skill when doing his lectures and learning about the subject.
Of course, his skill is a single pillar from the lectures: his book, the exercises he shares and his videos are more fundamental than the agentic skill.
deepsquirrelnet 7 hours ago [-]
I've tried to use it to create my own curriculum and tutor to learn a subject I didn't know.
You can clone the repo and load it in claude code and start working on it. I didn't finish the full course.
I just used this skill with a dense philosophical text this morning and I think it did a great job. I was exhausted after 15 minutes.
elxr 8 hours ago [-]
Interested in where this discussion goes as well.
I've always thought that using a book (or the complete documentation of whatever language/library is being learnt), plus having a custom app that integrates both the writing/building/note-taking step with the LLM guidance/follow-ups including the seeing the generated artifacts, etc. all in one window will be the best solution. Sort of like chatGPT desktop, but focused only on learning, and with a much smaller but more intelligent feature-set.
Too many of these skill.md based learning workflows just require too much context switching and having to use multiple apps.
Ebooks are a still great experience but not fast enough nowadays.
I have the socratic discussion with an agent, then via the mcp, I add resources to a structured path.
kaeluka 4 hours ago [-]
I’m doing this currently. Upload the ebook to chat gpt. Works quite well, especially in voice mode it’s cool
tristor 7 hours ago [-]
I have been thinking about this for awhile but haven't come up with any great answers. Ideally, I could use an LLM as a coach/tutor to go through an entire classical education in a structured way by combining reading assignments, graded/examined essay writing, and socratic style questioning. Possibly with a thesis / thesis defense at the end. The reading materials for a classical education are mostly old enough that they're available for free via Project Gutenberg, so it should theoretically be entirely feasible to do this in a way that its accessible to anyone with a library card and an internet connection, but at the same time the way most of the LLM tools work, they lack the necessary tracking/context in the harness to make this feasible and the way skills get applied is very clunky, to succeed here it needs to be very structured with strong tracking.
genxy 15 hours ago [-]
Socratic tutoring is the first thing I did when ChatGPT first shipped. Your liberal arts education is a "huge unlock" if you just apply it.
its offtopic but that Programmed learning wikipedia article is really bad. It's clearly written by someone with personal stake in it and isn't remotely encyclopedic in tone.
Not sure what to do about it because I don't care enough to learn everything about this topic in order to fix it.
Phemist 10 hours ago [-]
Comparing the opening sentences:
> Programmed learning (or programmed instruction) is a research-based system which helps learners work successfully. The method is guided by research done by a variety of applied psychologists and educators.
I have no idea what programmed learning is, only that it has an alternative name, that it has scientific street cred and that it is effective according to some nebulous definition of successful.
as opposed to:
> Mastery learning is an instructional strategy and educational philosophy that emphasizes the importance of students achieving a high level of competence (e.g., 90% accuracy) in prerequisite knowledge before moving on to new material.
which is gives me a more clear idea of what the core philosophy is behind it.
Can the programmed learning page be marked as low-quality, non-encyclopedic? That is already a good signal to have.
derangedHorse 6 hours ago [-]
Programmed learning just seems like active learning [1] with chunked [2] material
Programmed Learning (Programmed Instruction) goes back to instructional material created for the troops in WW2 so it has a history the predates both of those and was part of BF Skinner's research.
Socratic questions are not asked blindly. They are meant to help the student to realize the incompleteness or falsety of what they think they know. That us, to realise they do not actually know something and thus put them into a position of not knowing, which is the only way to gain understanding.
One man said, that enlightement (as in Zen) cannot be faked. Whoever had it, will be able to tell if someone else had it or not. There is no way to merely sound like it. Socratic method is like enlightement.
awongh 11 hours ago [-]
I think there could be some ways to fake this, using RAG or some other kind of agentic process, but in terms of the true meaning of the phrase, this is one of the things that is very far from being replaced by AI- an interaction between two people where a question that seems simple on the face of it is actually exactly what the learner needed to hear at the exact right time.
Very few teachers get to do this or are good enough, but it's just as much a question of relationships than anything else- how do you pose the exact right problem so that the learner can discover something for themselves.
zacian 14 hours ago [-]
Socratic questioning needs a lot of thinking to arrive at the realization (and be able to accept it) that one has incomplete knowledge.
I think about this, whenever I have questions which are related to morality and how people act.
gglitch 36 minutes ago [-]
From my memory of enthusiastically working maybe 1/3 of Plato in seminar 25 years ago, if you read enough Socrates, you start to absorb a few base patterns he uses over and over again, much as you do in Euclid. This is me agreeing with you. You start to recognize conversational/logical pathways that are likely, however annoying or even sort of BS they feel at the beginning, to lead you to an impasse that will show you the boundaries of your operating area. "Hmm, Meno. Interesting question, but I'm actually not sure I know what knowledge even is. Since you know so much, maybe you can teach me." [Buckle up]
andai 13 hours ago [-]
The problem is that everyone else can't tell, and that there's a lot of incentive to fake it. See Saints and Psychopaths by Hamilton.
I think this same phenomenon — faking competence, or good intentions — extends beyond spirituality as well.
15 hours ago [-]
fahrvrgnugen 14 hours ago [-]
[dead]
pkoird 10 hours ago [-]
Is it only me who's thinking what's so special about this and whys everyone oohing and aahing over this? Afaik, people I know (and me) we've been doing this ever since LLMs became capable enough.
baby 10 hours ago [-]
SKILLs are often underwhelming I find
miki123211 9 hours ago [-]
I think skills are useful in three contexts: for preferences which genuinely vary from user to user, for internal info / workflows / environment / harness descriptions which the model has no way to discover otherwise, and to teach it stuff that Claudex n doesn't yet know but that Claudex n+1 will.
IMO, prompt engineering tricks (which is basically what many skills are) indicate a weakness of RL. If the model has the intelligence to do something, but doesn't have the wisdom / judgment to know to do that in a given situation, that's something that the labs should (and will) fix with RL in generation n+1.
This is why prompt engineering used to be such a big thing circa 2023-24. GPT 3.5 Turbo was capable of much more than it was RL'd to do, e.g. the "think step by step trick" instead of it always starting with a CoT. Labs have gotten much better at this though, and so such tricks are no longer necessary.
pkoird 7 hours ago [-]
This. All "skills" and harness tricks will be internalized in the next iteration, specially if they are useful.
pizzathyme 6 hours ago [-]
Learning a complex topic over time with LLMs is a personal unsolved problem. Whenever I ask about a topic, they just spit out a long answer, which I read, and then ask a few questions, and that's the end. I don't really learn much at all other than one small factoid.
I haven't found a system yet that actually moves me along, makes progress over time, verifies my understanding.
Interested to try this out. If anyone has things they've found work, please share!
I was doing something similar before but - as a +1 - I find the clear and condensed description valuable, even if only to use as a foundation. It expresses cleanly what I've ended up rewriting in a prompt many times.
alephnerd 10 hours ago [-]
Pressure testing ideas and general discussion. You know - like the Socratic method.
A lot of people don't know about the Socratic method either so this is their first applied introduction to it.
As a meta-comment, this kind of reflexive dismissiveness is why HN has gone to shit. Zero curiosity, only dismissiveness.
pkoird 8 hours ago [-]
I suppose it surprised me that either a) people didn't know what Socratic method was (I thought it was famous and well known), or b) LLMs are pretty good at doing Socratic questioning.
Additionally, I don't even know if you need a "skill" for this. If you just say use Socratic method to coach me in X, or teach me Y from first principles, it already does a pretty great job.
alephnerd 7 hours ago [-]
[dead]
natebc 9 hours ago [-]
TBH I've always given Jane Street a little side-eye. They position themselves as some haven of academic research/study/growth when in reality they're just another HFT pouring accelerant on the economy.
frangonf 10 hours ago [-]
Is it wise to believe that every user of LLMs know about Socrates and his impertinent questioning he spent his time on around Athens? Is it possible that some millionaire quants math wizards found that llmwashing this ancient techniques is food for humanity in this uncertain times, or if anything, for their portfolio?
lifeisloving 10 hours ago [-]
[dead]
lemax 13 hours ago [-]
I’ve struggled to get AI to teach me things well without spilling hints or giving answers away. Assuming this was built on some solid evals this could be awesome.
> When the user is wrong, do not give the answer. Ask a narrower question that exposes the contradiction in their reasoning
vkgcr 13 hours ago [-]
It's a prompt - the LLM will always inevitably veer off path and stop following instructions / start hallucinating.
husseinhb 13 hours ago [-]
[dead]
jan_m_savage 14 hours ago [-]
Oh this is so good. I asked it to teach me programming in Logo; though the model is not the best (GLM 4.7, on pi), the tutor is asking very good questions and is not too pedantic about exact answers.
For those new (beginners), install pi from pi.dev and then your favorite AI model (ask your model how to set up pi and it). Then ask your pi-based AI how to set up the tutor (just give it this link, it will set it up for you).
Then start pi by typing pi in the terminal; and to launch the tutor preface your question with: /skill:socratic-tutor. Like so:
/skill:socratic-tutor teach me Python.
koolba 12 hours ago [-]
This is a cool idea. Though trying this out using the claude CLI is kind of funny because by default it's suggestion for the next prompt is the answer to the question.
nathell 12 hours ago [-]
Indeed! To turn that off, open /config and turn off Prompt suggestions.
__rito__ 5 hours ago [-]
Has anyone been using the teach skill [0]? How does it compare?
I was playing with it for a short time, otherwise, I just tell the model to 'teach me X, and do not give me direct answers, and give me exercises'.
Been using this with Opus 5.5 on Medium for a week or so. I've used the same model for learning before and I definitely prefer using this skill because it stops the model from the usual handholding. Although I'd say most of the effect can be replicated with this line: "When the user is wrong, do not give the answer. Ask a narrower question that exposes the contradiction in their reasoning."
Imanari 14 hours ago [-]
Socratic tutoring with LLMs works great for me and I have used this skill and it also works great (can you do the same with a simple prompt? probably)
Using LLMs this way was really refreshing, I could really feel how I started to think on my own again.
pimeys 14 hours ago [-]
I also like this skill. It's one of the few I've used the longest: more than a year now which is a century in LLM years. Although you have to remember to ask the LLM to fill its context before starting the questions.
Especially nice when you are not so sure how something should look like, so it makes you stop and spend a few hours answering questions and thinking about what you want to do. Combine that with the RFC skill and you get a good basis for your feature and then a skill to split the RFC to issues on Linear.
Imanari 9 hours ago [-]
Which RFC skill are you using?
sigbottle 9 hours ago [-]
I don't think a simple prompt will exceed any kind of memory system. Nothing fancy, but you need the agent to tune to your quirks; I find that a prompt to try and distill everything a priori is too annoying.
Been just using less of chat recently and using more codex for learning - a direct shared interactable terminal is nice, it's still relatively cheap, and importantly you can log specific thoughts to say a git repo for memory.
It's not attempting to be some kind of knowledge ontology graph like many businesses want - you just want it to understand the specific quirks of your specific learning path, which may not be formalizable or worth be spending time to step back and formalize every little thing.
If you read them, the skills are completely different (with the Jane Street skill being more complex). Socratic questioning [0] is a well-known technique. Indeed, a Github search returns 92 repositories for "socratic tutor skill" [1], suggesting that the well-known technique has been added to LLM tools independently by many people.
Yep - just looked at my projects in Claude, and my first "socratic tutor" was registered on 2nd of December 2024. I forget where I got the inspiration though (probably X)
vintermann 16 hours ago [-]
Same with dialog-tuning in the first place. That you could manipulate the context of unicorn-text generators like GPT-2 in order to actually have something like a conversation with them, or that a model could generate "private thoughts" before answering, were ideas which weren't as self-evident as they seem now. But still ideas a lot of people had at the same time.
orbis-tertius 14 hours ago [-]
Yes, the comment was intended as a playfully hyperbolic nod to the lack of novelty and originality in the idea, but I guess the intended tone didn't carry through :)
nimchimpsky 14 hours ago [-]
[dead]
16 hours ago [-]
orbis-tertius 3 hours ago [-]
If anyone is interested we built a TUI to use this outside Claude Code, with integrated spaced repetition and anki flashcards export
What sort of topics are people using this to learn?
brlewis 8 hours ago [-]
As someone with a superficial understanding of how LMs work, I'm surprised that the ASCII art part of the skill can work. I guessed that the harness has to get involved there, as ASCII art generation looks to me like a different task than next-token generation. Can a knowledgeable human add any detail?
I asked Gemini and at first it claimed that generating ASCII art is business as usual for the harness and the LM. It didn't give any convincing evidence. I asked it to steel-man the opposite position, and after that asked for its current thinking. It had convinced itself of the opposite, but still no definitive sources.
jazzypants 8 hours ago [-]
I mean, I assume that it trained on ASCII art so I don't see how it's any different at all
poilcn 16 hours ago [-]
what am i missing, why does it have 434 forks and 6.3k starts? is it botted? what's the point
ngruhn 15 hours ago [-]
Pure markdown skills are trivially recreatable. I think the value of them is more the idea than the artifact.
michelb 14 hours ago [-]
I think a lot of people use 'fork' to bookmark a repo
Turneyboy 9 hours ago [-]
It's really just because it's Jane Street.
zkmon 16 hours ago [-]
This is really cool. I think we will see more of these, inching towards 'refined' thought and expression percolating from AI models into human minds.
I noticed that the quality of email text has gone up around my work these days. everyone is sounding like a master of drafting emails. Ofcourse Ai has a role in it. But real cool thing about this is, this is pulling up people's thinking to have a structure, minimalism and sequencing of expression. Same as how textile industry has enabled people to have an style and expression in clothing.
maurits 10 hours ago [-]
> But real cool thing about this is, this is pulling up people's thinking to have a structure, minimalism and sequencing of expression.
I don't agree. Writing introduces friction that forces you to think and reflect. If you let AI do that part, you are not only skipping the writing, but more importantly, the thinking.
Christopgr 10 hours ago [-]
Also, having a system hone people into the same writing style kills creativity and diversity.
Very interestingly I was looking for the word "diversity" and so I was asking my mother for the word as she happens to be visiting and also told her what I was writing. She said that it's like taking a rainbow and making all its color gray, she is the artistic type. :)
javier123454321 7 hours ago [-]
That's been my conclusion trying to dive into the learning science for a learning tool I am building. Much like going to the gym, learning requires putting effort into producing an output. The fact there is a motorcycle next to every running track in every application is good for seeming like you did more effort, but not for actually getting better at the subject.
I completely agree. The use of AI in programming and writing has enabled us to produce highly professional-looking output, yet we are losing our capacity for professional judgment.
elcritch 13 hours ago [-]
> Same as how textile industry has enabled people to have an style and expression in clothing.
I'd tend to agree with what you said, except this part. Mass produced clothing is fantastic for many reasons, but not for style and expression. Look at the traditional clothing from almost any part of the world and it was gorgeous and colorful. Now it's all just a t-shirts.
ben_w 12 hours ago [-]
> Look at the traditional clothing from almost any part of the world and it was gorgeous and colorful.
Was it?
There's a lot of survivorship bias in old things; we don't see the things which weren't recorded. Old things had their style, that much is unavoidable, but before mass production normal people inherited their clothes because it was just that hard to create.
Also, a lot of the pigments we take for granted today didn't exist at any price in the pre-industrial era, and some which did involved trekking to a mine in Afghanistan: https://en.wikipedia.org/wiki/Ultramarine
This also means (in general, not specifically clothing) a lot of "traditional" things are post-industrial recreations or re-imaginings of the past.
ineedaj0b 12 hours ago [-]
spoke to a clothing historian in Europe years ago. Pigments, shoes, and washing have improved… But all clothes being tailored and more craftsman made the experience better.
She said she’d only choose the modern era because of the washing machine.
pbh101 10 hours ago [-]
Clothing was extremely expensive historically relative to wages due to high labor requirements, so until the Industrial Revolution most clothing we now look back to and admire was the domain of the relatively wealthy. It might have had better craftsmanship, but many were literally in rags.
9 hours ago [-]
williamdclt 13 hours ago [-]
It was gorgeous and colourful, but highly impersonal. Traditional clothing, almost by definition, isn't really an enabler of personal style and expression. And you can't really compare traditional clothing to day-to-day clothing, these are entirely different purposes and codes.
Day-to-day clothing is definitely not "all just tshirts" nowadays, even in the mainstream. Women fashion is very varied, and while men's fashion is a lot poorer there's still a huge spectrum of expressive clothing that's available. I don't think personal style and expression has ever been as available and encouraged as it is today.
watwut 10 hours ago [-]
Huh? At least where I from, you got cloth decorated by a person who did not produced copies the way machines do. That person was frequently you yourself.
> And you can't really compare traditional clothing to day-to-day clothing, these are entirely different purposes and codes.
I would interpret traditional clothing as "clothing they traditionally wore". It was not a uniform.
troupo 12 hours ago [-]
> It was gorgeous and colourful, but highly impersonal.
So is mass-produced clothing. Go into any store, and you will see the same shit with very few variations. For actual personality you have to search for niche/hand-made/small-productions etc.
zkmon 11 hours ago [-]
Before the stitching machine arrived in my village, all you can see was large rectangular pieces of cloths, delivered directly from the weaver craftsmen of the village, for men and women alike (only color difference - white for men and mostly dark green for women). The washermen still delivered clothes correctly across dozen's of families everyday, after washing. The famous Indian Saree is an example of unstitched clothing. Dhoti, lungi, turban and towel are other forms rectangular clothing.
People had cultivated an art in how the rectangular cloths are wrapped around them and they clothes are secured at their waist, or around their head. Talk to a Sikh or Afghan about turban.
DaiPlusPlus 8 hours ago [-]
> Before the stitching machine arrived in my village...
Please elaborate.
zkmon 2 hours ago [-]
Sewing machine that makes stitches. Early 70's. Somewhere in India.
12 hours ago [-]
11 hours ago [-]
zanellato19 7 hours ago [-]
> I noticed that the quality of email text has gone up around my work these days. everyone is sounding like a master of drafting emails. Ofcourse Ai has a role in it.
I find that everyone _looks_ like they know how to write now, but the good writing has become even rarer and a big text is now indicative of laziness more than of work, which means that I get annoyed with big emails and messages, they are almost always purely AI
Lithuanian 13 hours ago [-]
It's not really good that email (and written content overall) quality has gone up. The problem is that there's tons of high quality content but no way to know which content is useful or good. Before we could know someone's personality from their style, now everyone's just wearing Balenciaga.
prolly97 16 hours ago [-]
I love this positive take :)
11 hours ago [-]
ineedaj0b 12 hours ago [-]
I have Ai read those emails and condense into one sentence summaries.
I’m not reading all that. I want one word answers.
Also noticed huge uptick in “call me” emails.
iugtmkbdfil834 15 hours ago [-]
And, hilariously, as an AI booster, I personally dislike it, because now everyone sounds like they have a PR company ( and the same PR company come to think of it ). It was bad before in corporate land, but the level of standardization for optimized response is aggravating to me, when I am simply trying to have a short conversation.
Case in point, I reached about syllabus to my kid's teacher. Long story short, email ended with their boss, who sent me long winding non-answer that I had to resist to not knock down.
I dunno, generic executive summaries can be canned, but I would like to hope I pay them enough for an person to person exchange.
Arkhaine_kupo 14 hours ago [-]
I have resorted to asking some people "just send me the prompt", like as messy as it might be, I will get the gist much faster than reading through Claudish or Gptinglsh which my eyes gloss over half the text
iugtmkbdfil834 18 minutes ago [-]
I swear I may need to use it soon. I won't make any friends, but maybe that is where are now.
tarkin2 13 hours ago [-]
If they are copying and pasting the answers and replies how will it "pull up people's thinking"?
close04 14 hours ago [-]
> everyone is sounding like a master of drafting emails
> this is pulling up people's thinking to have a structure, minimalism and sequencing of expression
I see the same for the former but the opposite on the latter. Everyone sounds like expert drafters because the AI is doing all of that.
True to the "use it or lose it" principle, the people's thinking skills about the emails have weakened. They rely on giving the AI some bits of info, a document they never read, some vague instructions, and the AI molds the message with structure and sequence. Then I get to experience live the joy of discussing with these people face to face and they still just spit out disparate bits of information with no fluid structure and sequence for the ideas because the only time they thought about the topic was when the AI wrote an email about it.
gregorygoc 10 hours ago [-]
> because the only time they thought about the topic was when the AI wrote an email about it.
Calling this “thought” is an injustice.
14 hours ago [-]
11 hours ago [-]
conartist6 13 hours ago [-]
[flagged]
TonyAlicea10 8 hours ago [-]
No matter how many of these are tried (and I’ve tried making many iterations), LLMs are generally terrible at building a progressive curriculum.
They inevitably structure things poorly, often including topics before they’re fully explained later.
The bigger the curriculum the worse it gets. They can be amazing once a human instructor provides an outline but not from scratch.
game_the0ry 8 hours ago [-]
First of all -- hi Tony. Big fan, I bought 2 of your courses way back when I was starting my career. I think you are great teacher so I respect your opinion on this matter.
But could you elaborate? I find ai to be a pretty competent tutor. Could your experience be related to context rot bc of large topics?
TonyAlicea10 2 hours ago [-]
Hi, thanks! Always nice to talk to a past student. :)
I think AI can be very effective at what I call "top down" education. You pick a topic, and you can ask deeper and deeper questions about it.
I find AI is very poor at building "bottom up" education. By that I mean putting yourself in the shoes of the student, anticipating their questions, and teaching what they need to know before teaching what they want to know.
For example, when I teach HTML/CSS, I first teach about tree data structures and how they're generally stored in memory. Hours later that pays off when talking about CSS and performance, and gives the vocabulary to talk about the subject the student is actually learning. LLMs wouldn't likely do this unless the student asked "what's a tree?".
I think folks that read Hacker News are on average going to ask better questions, and so will have a better experience with top down education via LLM.
bemtevi 8 hours ago [-]
If these become really good, I think it's sad that people will interact with other people less and less and just learn from machines. It might make people smarter and more well-versed in some arcane topic, but I wonder if that will make the world a better place? I think probably not.
esafak 8 hours ago [-]
If it makes the average person more educated that would surely be a good thing.
bemtevi 8 hours ago [-]
Would it? Maybe up to a point. But I doubt average people are going to start randomly learning new topics in-depth, agent or no agent. People don't just start learning randomly.
But even if they did, people not relying on others at all leads also to a high degree of narcissism. More education is good, but not if that increase comes at a tradeoff of less socially-developed and empathic people. People here seem to forget this.
I find it odd that you seem to take your statement as dogma rather than something to be clarified carefully with edge cases such as these.
i_am_a_squirrel 42 minutes ago [-]
works well! thank u!
amelius 12 hours ago [-]
I want an AI that I can give a paper, and then sits with me to understand it, maybe even implement it (though this is the easy part for the LLM).
Love this. There is so much value when using LLMs this way, it needs effort and consistency from the user but the results are really good.
A vanilla version of this is, at the end of a session, ask the model to generate a couple of "ultimate" questions to really flex your understanding. Obviously, the quality depends on the model but I've found it useful.
tigitouchdown 14 hours ago [-]
I’ve been building a harness for just this all year, they’re great. One of the main use cases for LLMs I think is largely overlooked.
slaser79 10 hours ago [-]
Good thing with skills being just markdown text you can use them in any harness. Just imported into my phone harness (I already had a Socratic tutor like skill), but this crispier.
flipbrad 14 hours ago [-]
Do you have something you could share, and can it be repurposed for use with kids?
bilalabdullahi 13 hours ago [-]
I'd be curious to see whether learners can still solve similar problems on their own a week later
bsenftner 10 hours ago [-]
I've written an entire Socratic AI based office suite with project management and media creation, current count over 1500 agents:
It's comprehensive, it was written at an immigration law firm, so it has all kinds of language and other perspective understanding tools, as well as a complete media suite to aid in the creation of visual explainers, courseware, and marketing media.
Has anyone tried ChatGPT's study mode? I thought it was supposed to be just like this, but I was underwhelmed when I tried it, and they ended up removing it. (It might be back now?)
I have the leaked prompt around here somewhere...
anonymousDan 11 hours ago [-]
Copilot has a study and learn mode. It's ok.
burrish 13 hours ago [-]
Socratic tutor skill isn't new at all, why did this reach the front page
igleria 14 hours ago [-]
Trying it right now. it's pretty decent and it does sort of make my brain work. Pondering right now when the "bitter lesson" will make these skills useless.
Paracompact 14 hours ago [-]
"These skills" being making your brain work, or this Socratic tutor SKILL.md?
igleria 13 hours ago [-]
hahaha, I meant the skill.md from Jane Street. But I'm currently fighting for my brain every day now too!
spwa4 13 hours ago [-]
First one, then the other.
0815beck 15 hours ago [-]
man people need to talk to people this is depressing
cbg0 13 hours ago [-]
Do you have readily available access to professionals in every field you're curious about? I'm interested in learning more about electricity but don't know anyone who's certified.
rwbt 11 hours ago [-]
I wish I could just go to any university and talk to professors and have a socratic discussion about many subjects. Unfortunately doesn't work that way in reality.
Traster 15 hours ago [-]
I see a lot of this at my company too - everyone re-inventing the same ideas over and over. I wonder how much people actually adopt other teams' implementation of this. In my experience I'm bombarded with a thousand different iterations of these things that everyone wants to advertise as a sort of "Look at me! I'm so AI!" but I simply don't have time to evaluate every different persons variation on these things, I certainly don't have time to really thoroughly benchmark them, and from experience a lot what I needed a carefully tuned skill to achieve 6 months ago is trivial for a better tuned agent today. I'm literally getting to the point where even my own skills are beginning to bit-rot.
thibsrn 13 hours ago [-]
Sensei, please teach me the ways to reach alpha just like you.
Jokes aside, pretty cool. This is just a skill, and it'd be easy to dismiss it. But I do want to become a better horse whisperer. Prompt engineering used to be mocked a lot, especially on HN, and skills really is just that but with a new branding.
I think Duolingo's days and the likes are counted, AI is just a great way to quickly get up to speed on a subject or a language.
Paracompact 13 hours ago [-]
Duolingo maybe or maybe not, but even having God himself at your beck and call behind a chat interface will not guarantee you learn a language. Every stage of language acquisition requires multimodal stimulation and long sustained effort that can't be shortcutted by having a smarter teacher.
sjdh76 13 hours ago [-]
> Sensei, please teach me the ways to reach alpha just like you
If it was this easy, educators would have written up sets of standard Socratic teaching prompts long ago. Like perhaps in the time of Socrates himself. Or the British empire or John Dewey etc. There is probably some great research in the field of Education on this approach.
resters 7 hours ago [-]
socratic tutor was one of openai's sample prompts a year or two ago, i'm curious if this one is better.
imagine being an educator and realizing that an LLM is actually a better teacher than nearly all who are employed in the profession!
This should be as big a headline as agentic development reducing demand for highly paid software engineers.
amelius 12 hours ago [-]
LLMs should first learn how to write English for humans.
Right now what they produce reads like legalese, or patents.
I suspect this is because the AI labs are optimizing for agents talking to agents, rather than AI talking to humans (which is harder to optimize, and does not improve intelligence as fast).
mathieuh 16 hours ago [-]
I assume this isn't going to cause the LLM to start jumping me with questions about the nature of piety or justice
bob1029 16 hours ago [-]
This comes off as AI consultant slop from a year ago. I wonder who is in charge of AI at Jane Street.
My client just got done firing someone who thought filling the codebase up with files like this would be a good idea.
It would seem the AI revolution is happening at wildly different rates depending on where you are standing.
dimator 5 hours ago [-]
This is not really something to plop into a repo, is it?
This is for the case where you the human want to learn something, instead of spewed out for you.
mw888 15 hours ago [-]
I think it's a great idea, but I'm somewhat anti-skill.
Every single skill I see just strikes me as a prompt. In every case it's easier for me to create a description of what is needed specific to the task. If I need a longstanding process I will discuss with the agent and create a document.
This is one of the most flexible tools ever made, and I frankly am not going to accept unnecessary systematization of how I use it.
bluerooibos 15 hours ago [-]
> Every single skill I see just strikes me as a prompt
Both of which are just blocks of text being stomped into an LLM.
I've got a feeling that prompting as we're doing currently (huge blocks of text with very little structure), at least when looking for a deterministic result, will be looked back on as if we were living in the Stone Age. Not sure what the alternative will be, but I'm imagining something like a CI pipeline definition in YAML.
wusel 14 hours ago [-]
A skill is usually auto-invoked and this one here has a bunch of trigger phrases. I personally dislike that. For this one especially I'd prefer explicit invocation.
jsw97 10 hours ago [-]
That is configurable in a good harness. (Or should be.) It's not intrinsic to the skill idea.
cemkum 13 hours ago [-]
> If I need a longstanding process I will discuss with the agent and create a document.
Isn't this just what a skill is?
jkhdigital 15 hours ago [-]
Nodding vigorously to that last comment. If I want to systematize something I put it in code. Natural language is not code.
casey2 5 hours ago [-]
Rather than saying stuff like 4-10 or 3-5, make that an explicit program (rng) and tell it to add or subtract based on a decision. Otherwise it is very likely to just always pick the same number.
SkyBelow 12 hours ago [-]
I've tried something like this and ran into a problem which I'm not seeing a fix for. Generally, to really apply this method, you need to understand the person asking the question.
>Tell the user that you will start by asking a few questions, one at a time, to see what they already know.
This puts it into the session each time. Do this once and it seems to work. By the 5th time, you are getting tired of reconfirming your mastery. This ends up depending upon the user to have a pretty solid foundation of what they know and then use this skill to expand the boundary a bit, and even then the AI can end up making so many assumptions that a person has trouble following it along.
I was thinking some sort of concept mapping kept long term might help, but such a mapping can become really difficult to build. Which should be expected when we consider that teaching isn't just its own discipline, but has many sub-disciplines targeted to specific areas (even if some of it, like classroom management, doesn't apply to this scenario).
I made a few attempts at a knowledge tracking tool for personal use to track how much I knew Japanese beyond a simple JLPT level or similar with the plan to tie it into something like this, but it wasn't great. I could 'cheat' and just tap into my Wanikani API to see my progress, but that still skips so many other areas of Japanese, and even when WaniKani says something is burned, I really want AI to still use it until I truly internalize it like I do words in English. I suspect other areas of knowledge are similar.
pbronez 10 hours ago [-]
Yes, lots of LLM Skills quickly turn into applications. The skill is the program and the LLM is the compute, and that leaves the harness to manage the state. I’ve found that these skills need explicit instructions on how to manage state across sessions.
In my work, this typically means an opinionated local folder structure and a dedicated /init skill that implements it. Immediately this means we have more than one skill so we’re talking about a plugin now.
Then I think about meta-state. Not just “what is on disk” but “where are we in this application’s multi-session lifecycle?” For Socratic tutoring, that means checking to see if we’ve completed an initial intake, then what skill levels the user has demonstrated, and finally what’s next on the menu.
For this specific skill, I’m surprised it doesn’t say anything about curriculum. It just assumes the model knows the topic and can improvise both a curriculum and credible assessments of progress in that curriculum. That can’t be true for all things people want to learn, so we need a /build-curriculum skill. And probably a /validate-curriculum skill for actual experts to use before launching a given curriculum.
Anyway, I’ve been thinking a lot about this kind of thing as I roll out skills to my firm.
orbis-tertius 12 hours ago [-]
This works really well.
A positive take: this works much better than the many projects on gh that did this before Jane Street. This would be evidence that creativity and execution still matter, even when the product is just an md file
A cynical take: network effects are everything. tens of projects doing the same thing and this only gains traction when an already popular org does it
keybored 7 hours ago [-]
Looks brilliant. I’ve recently been beating around the bush, not telling what the actual problem is, in order to research something without being handed a solution. It’s like intentional YX problem framing.
16 hours ago [-]
bearjaws 10 hours ago [-]
How is this even remotely new?
I remember with ChatGPT 3.5 people were saying "LLM socratic tutor will revolutionize education"
Here we are 3 years later (almost to the day) and well, education isn't revolutionized in the way people imagined.
Pragmata 14 hours ago [-]
This is honestly pretty damn good actually.
Running it with my local models it is a pretty helpful way to study up something where you already have some bases.
proc0 12 hours ago [-]
Ah yes, I recall all those times as an engineer where my managers kept insisting I communicate like Socrates to improve my work. Software is solved.
Gokul_e 14 hours ago [-]
can i use this for learning DSA
lifeisloving 10 hours ago [-]
[dead]
DeathArrow 13 hours ago [-]
At some point the context will degrade and the questions might not be useful.
zacian 14 hours ago [-]
One more interesting test I'd like to do with this skill is. "Is justified true belief knowledge?" and arrive at Gettier through Socratic questioning.
IshKebab 16 hours ago [-]
I dunno if you really need this. I asked AI to generate me a Lean tutorial with exercises and it did a really great job, one-shot with no skills. My 4th attempt to learn Lean and I'm actually making progress this time. (Mostly because you can say "I'm a bit stuck here, can you give me a hint" - it's really good at Lean even without access to the Lean compiler.)
Maybe it improves things a bit though. I might try it next time.
kittikitti 9 hours ago [-]
The Socratic method is meant for humans that are educated enough to filter out misinformation in a serious platform. Otherwise, the Socratic method is basically manipulating someone by acting dumb. On the other hand, there is value in asking follow up questions in the learning process.
On one hand, it is very tempting to use AI to just do your schoolwork for you.
On on the other, everyone now has access to an expert-level tutor in any subject they possibly care to learn about.
As an example I've been reading Build a Large Language Model (From Scratch) -- I take pictures of the book and ask Claude to help me understand concepts. I chat with it to ask questions and check my understanding. I don't get stuck because there's always "someone" to ask.
I think LLMs are going to fall short for the same reason that the hard book itself does. At least with regard to genuine fluency in a given topic.
That is, even if the text includes challenge questions and an accurate rubric, the questions are always going to be a limitted application of the total use of the information.
Once you're getting to the "evaluate" and "create" portion of Bloom's taxonomy, the text can no longer tell you whether or not you're on the right track because the book cannot understand your appraisal/defense of an argument/implementation.
The LLM, too, will either "good, now you're getting it!" or "close, but not quite" based on system prompt if you submit a multi-paragraph or whatever answer.
All of which, I think, is why the link you posted ultimately directs the student to a community: learning in a vacuum is limitted.
I'm not sure why you think this. have you tried it?
To understand anything new you need to first have a base of knowledge to build on top of.
I find it helpful to think of the new concepts as a high dimensional space where specific facts / concepts lie in.
The first step: is to jump right in, take a fact / concept and read a paragraph or two about it. You have not really learned anything at this point, and you may not even fully understand what you ready that's fine.
{Say you want to learn about dot product, you may learn how to calculate it mathematically / mechanically}
(Note if you did not understand anything, it may be a sign that something in step 0 is missing)
Next step: is to read about it again. But this time find a different angle (I call it a different "projection", we are projecting this unknown knowledge vector onto the basis vectors of your pre-existing knowledge base.) look at it form a different angle. It helps to rephrase the content in your own words, read different definitions for it, try to explain why different ways to describe the same thing can be correct at the same time.
{Instead of thinking the dot product as sequence of multiplication and addition, think about the classic visualization of the product of the magnitude of one vector and the magnitude of the projection of the other onto the first [1]}
Step 3: Repeat step 2 until you understand it. (Or even better you can come up with a completely new way of looking at the concept)
Step 4: Now you are already familiar with the idea, yet this step is the most critical for continual learning. Ask Why. Why do we need this? Why describe the same thing in 10 different ways? If you can understand this you have added a basis vector to your arsenal of knowledge! From the previous steps you made yourself familiar with the shape of the knowledge vector, step 4 will put more context around it, and show it where it fits in your pre-existing knowledge base.
(Pro tip: mix step 3 and 4 as needed! no need to stick to a strict sequence here)
step 4 is difficult, as there may be may applications for a concept, in such cases it helps to understand the patterns between the applications. What is causing all these different things to rely on this one fact?
step 5: teach it to someone else!
(1 https://www.falstad.com/dotproduct/
--- edit: forgot to add the point of the comment lol The point is llms are good at generating / regurgitating these different view points. They are also excellent quickly checking your ideas / view points.
LLMs will wear whatever hat you throw at them.
I haven’t done it with these questions, but I have done some experiments with a local language model I’m working with
I compare the model just doing typical RAG, vs the model using a tool to “pre-study” the material first. The tool breaks down the material into basic question/answer pairs, and then gives the model a way to access that knowledge on inference
In my specific use case (a small agent that can talk about Wikipedia articles), it works really well, and the approach seems mostly limited by two factors: first, the quality/variety of question/answer pairs produced by the study tool, and the ability of the model to use the stored knowledge coherently in conversation
So, on the point above about quality of question/answer pairs, I’m wondering if these learning skills could be applied
I’m not the developer or anything; just really enthusiastic about this because I’ve already learned so much with LearnHarness. Link to repo: https://github.com/amosblomqvist/learnharness
Here is a video too. You can skip to the demo part to see how it works https://www.youtube.com/watch?v=nfTDljFG2g0
@ivansavz has done that as a "teaching assistant" on his statistics lectures from his book No Bullshit Guide to Statistics.
I have particularly benefited a lot from his agentic skill when doing his lectures and learning about the subject.
Of course, his skill is a single pillar from the lectures: his book, the exercises he shares and his videos are more fundamental than the agentic skill.
You can clone the repo and load it in claude code and start working on it. I didn't finish the full course.
That was ~9mo ago, so it would be interesting to see how it works now. https://github.com/dleemiller/learn-zig-from-python-ml
Some things I found that help: https://github.com/dleemiller/learn-zig-from-python-ml/blob/...
I've always thought that using a book (or the complete documentation of whatever language/library is being learnt), plus having a custom app that integrates both the writing/building/note-taking step with the LLM guidance/follow-ups including the seeing the generated artifacts, etc. all in one window will be the best solution. Sort of like chatGPT desktop, but focused only on learning, and with a much smaller but more intelligent feature-set.
Too many of these skill.md based learning workflows just require too much context switching and having to use multiple apps.
Ebooks are a still great experience but not fast enough nowadays.
I have the socratic discussion with an agent, then via the mcp, I add resources to a structured path.
See also https://en.wikipedia.org/wiki/Mastery_learning
https://en.wikipedia.org/wiki/Programmed_learning
Make your own custom skinner boxes. https://en.wikipedia.org/wiki/Operant_conditioning_chamber
Not sure what to do about it because I don't care enough to learn everything about this topic in order to fix it.
> Programmed learning (or programmed instruction) is a research-based system which helps learners work successfully. The method is guided by research done by a variety of applied psychologists and educators.
I have no idea what programmed learning is, only that it has an alternative name, that it has scientific street cred and that it is effective according to some nebulous definition of successful.
as opposed to:
> Mastery learning is an instructional strategy and educational philosophy that emphasizes the importance of students achieving a high level of competence (e.g., 90% accuracy) in prerequisite knowledge before moving on to new material.
which is gives me a more clear idea of what the core philosophy is behind it.
Can the programmed learning page be marked as low-quality, non-encyclopedic? That is already a good signal to have.
[1] https://en.wikipedia.org/wiki/Active_learning
[2] https://en.wikipedia.org/wiki/Chunking_(psychology)
B.F Skinner. Teaching machine and programmed learning https://www.youtube.com/watch?v=jTH3ob1IRFo
One man said, that enlightement (as in Zen) cannot be faked. Whoever had it, will be able to tell if someone else had it or not. There is no way to merely sound like it. Socratic method is like enlightement.
Very few teachers get to do this or are good enough, but it's just as much a question of relationships than anything else- how do you pose the exact right problem so that the learner can discover something for themselves.
I think about this, whenever I have questions which are related to morality and how people act.
I think this same phenomenon — faking competence, or good intentions — extends beyond spirituality as well.
IMO, prompt engineering tricks (which is basically what many skills are) indicate a weakness of RL. If the model has the intelligence to do something, but doesn't have the wisdom / judgment to know to do that in a given situation, that's something that the labs should (and will) fix with RL in generation n+1.
This is why prompt engineering used to be such a big thing circa 2023-24. GPT 3.5 Turbo was capable of much more than it was RL'd to do, e.g. the "think step by step trick" instead of it always starting with a CoT. Labs have gotten much better at this though, and so such tricks are no longer necessary.
I haven't found a system yet that actually moves me along, makes progress over time, verifies my understanding.
Interested to try this out. If anyone has things they've found work, please share!
A lot of people don't know about the Socratic method either so this is their first applied introduction to it.
As a meta-comment, this kind of reflexive dismissiveness is why HN has gone to shit. Zero curiosity, only dismissiveness.
Additionally, I don't even know if you need a "skill" for this. If you just say use Socratic method to coach me in X, or teach me Y from first principles, it already does a pretty great job.
> When the user is wrong, do not give the answer. Ask a narrower question that exposes the contradiction in their reasoning
For those new (beginners), install pi from pi.dev and then your favorite AI model (ask your model how to set up pi and it). Then ask your pi-based AI how to set up the tutor (just give it this link, it will set it up for you).
Then start pi by typing pi in the terminal; and to launch the tutor preface your question with: /skill:socratic-tutor. Like so:
/skill:socratic-tutor teach me Python.
I was playing with it for a short time, otherwise, I just tell the model to 'teach me X, and do not give me direct answers, and give me exercises'.
[0]: https://github.com/mattpocock/skills/tree/main/skills/produc...
Using LLMs this way was really refreshing, I could really feel how I started to think on my own again.
Especially nice when you are not so sure how something should look like, so it makes you stop and spend a few hours answering questions and thinking about what you want to do. Combine that with the RFC skill and you get a good basis for your feature and then a skill to split the RFC to issues on Linear.
Been just using less of chat recently and using more codex for learning - a direct shared interactable terminal is nice, it's still relatively cheap, and importantly you can log specific thoughts to say a git repo for memory.
It's not attempting to be some kind of knowledge ontology graph like many businesses want - you just want it to understand the specific quirks of your specific learning path, which may not be formalizable or worth be spending time to step back and formalize every little thing.
[0]: https://en.wikipedia.org/wiki/Socratic_questioning [1]: https://github.com/search?q=socratic+tutor+skill&type=reposi...
https://github.com/nodo-labs-uk/socratui
Thanks a lot for flagging this
What sort of topics are people using this to learn?
I asked Gemini and at first it claimed that generating ASCII art is business as usual for the harness and the LM. It didn't give any convincing evidence. I asked it to steel-man the opposite position, and after that asked for its current thinking. It had convinced itself of the opposite, but still no definitive sources.
I noticed that the quality of email text has gone up around my work these days. everyone is sounding like a master of drafting emails. Ofcourse Ai has a role in it. But real cool thing about this is, this is pulling up people's thinking to have a structure, minimalism and sequencing of expression. Same as how textile industry has enabled people to have an style and expression in clothing.
I don't agree. Writing introduces friction that forces you to think and reflect. If you let AI do that part, you are not only skipping the writing, but more importantly, the thinking.
Very interestingly I was looking for the word "diversity" and so I was asking my mother for the word as she happens to be visiting and also told her what I was writing. She said that it's like taking a rainbow and making all its color gray, she is the artistic type. :)
Here's some science to describe this intuitive process: On the effect of ChatGPT on classrooms: https://www.mdpi.com/2078-2489/17/3/299 On misatributing insight to LLMs: https://arxiv.org/abs/2604.10288 And the foundational literature before LLMs on removing friction from learning: https://bjorklab.psych.ucla.edu/wp-content/uploads/sites/13/...
I'd tend to agree with what you said, except this part. Mass produced clothing is fantastic for many reasons, but not for style and expression. Look at the traditional clothing from almost any part of the world and it was gorgeous and colorful. Now it's all just a t-shirts.
Was it?
There's a lot of survivorship bias in old things; we don't see the things which weren't recorded. Old things had their style, that much is unavoidable, but before mass production normal people inherited their clothes because it was just that hard to create.
Also, a lot of the pigments we take for granted today didn't exist at any price in the pre-industrial era, and some which did involved trekking to a mine in Afghanistan: https://en.wikipedia.org/wiki/Ultramarine
This also means (in general, not specifically clothing) a lot of "traditional" things are post-industrial recreations or re-imaginings of the past.
She said she’d only choose the modern era because of the washing machine.
Day-to-day clothing is definitely not "all just tshirts" nowadays, even in the mainstream. Women fashion is very varied, and while men's fashion is a lot poorer there's still a huge spectrum of expressive clothing that's available. I don't think personal style and expression has ever been as available and encouraged as it is today.
> And you can't really compare traditional clothing to day-to-day clothing, these are entirely different purposes and codes.
I would interpret traditional clothing as "clothing they traditionally wore". It was not a uniform.
So is mass-produced clothing. Go into any store, and you will see the same shit with very few variations. For actual personality you have to search for niche/hand-made/small-productions etc.
People had cultivated an art in how the rectangular cloths are wrapped around them and they clothes are secured at their waist, or around their head. Talk to a Sikh or Afghan about turban.
Please elaborate.
I find that everyone _looks_ like they know how to write now, but the good writing has become even rarer and a big text is now indicative of laziness more than of work, which means that I get annoyed with big emails and messages, they are almost always purely AI
I’m not reading all that. I want one word answers.
Also noticed huge uptick in “call me” emails.
Case in point, I reached about syllabus to my kid's teacher. Long story short, email ended with their boss, who sent me long winding non-answer that I had to resist to not knock down.
I dunno, generic executive summaries can be canned, but I would like to hope I pay them enough for an person to person exchange.
> this is pulling up people's thinking to have a structure, minimalism and sequencing of expression
I see the same for the former but the opposite on the latter. Everyone sounds like expert drafters because the AI is doing all of that.
True to the "use it or lose it" principle, the people's thinking skills about the emails have weakened. They rely on giving the AI some bits of info, a document they never read, some vague instructions, and the AI molds the message with structure and sequence. Then I get to experience live the joy of discussing with these people face to face and they still just spit out disparate bits of information with no fluid structure and sequence for the ideas because the only time they thought about the topic was when the AI wrote an email about it.
Calling this “thought” is an injustice.
They inevitably structure things poorly, often including topics before they’re fully explained later.
The bigger the curriculum the worse it gets. They can be amazing once a human instructor provides an outline but not from scratch.
But could you elaborate? I find ai to be a pretty competent tutor. Could your experience be related to context rot bc of large topics?
I think AI can be very effective at what I call "top down" education. You pick a topic, and you can ask deeper and deeper questions about it.
I find AI is very poor at building "bottom up" education. By that I mean putting yourself in the shoes of the student, anticipating their questions, and teaching what they need to know before teaching what they want to know.
For example, when I teach HTML/CSS, I first teach about tree data structures and how they're generally stored in memory. Hours later that pays off when talking about CSS and performance, and gives the vocabulary to talk about the subject the student is actually learning. LLMs wouldn't likely do this unless the student asked "what's a tree?".
I think folks that read Hacker News are on average going to ask better questions, and so will have a better experience with top down education via LLM.
But even if they did, people not relying on others at all leads also to a high degree of narcissism. More education is good, but not if that increase comes at a tradeoff of less socially-developed and empathic people. People here seem to forget this.
I find it odd that you seem to take your statement as dogma rather than something to be clarified carefully with edge cases such as these.
I use it as a better PDF reader for arxiv paper but they are going all in on AI teaching/explaining things.
A vanilla version of this is, at the end of a session, ask the model to generate a couple of "ultimate" questions to really flex your understanding. Obviously, the quality depends on the model but I've found it useful.
https://midombot.com/b1/home
It's comprehensive, it was written at an immigration law firm, so it has all kinds of language and other perspective understanding tools, as well as a complete media suite to aid in the creation of visual explainers, courseware, and marketing media.
https://midombot.com/b1/blog/MI/method-intelligence-presents...
I have the leaked prompt around here somewhere...
Jokes aside, pretty cool. This is just a skill, and it'd be easy to dismiss it. But I do want to become a better horse whisperer. Prompt engineering used to be mocked a lot, especially on HN, and skills really is just that but with a new branding.
I think Duolingo's days and the likes are counted, AI is just a great way to quickly get up to speed on a subject or a language.
obligatory Bruce Lee - https://m.youtube.com/watch?v=QQiH2-bLOmQ
imagine being an educator and realizing that an LLM is actually a better teacher than nearly all who are employed in the profession!
This should be as big a headline as agentic development reducing demand for highly paid software engineers.
Right now what they produce reads like legalese, or patents.
I suspect this is because the AI labs are optimizing for agents talking to agents, rather than AI talking to humans (which is harder to optimize, and does not improve intelligence as fast).
My client just got done firing someone who thought filling the codebase up with files like this would be a good idea.
It would seem the AI revolution is happening at wildly different rates depending on where you are standing.
This is for the case where you the human want to learn something, instead of spewed out for you.
Every single skill I see just strikes me as a prompt. In every case it's easier for me to create a description of what is needed specific to the task. If I need a longstanding process I will discuss with the agent and create a document.
This is one of the most flexible tools ever made, and I frankly am not going to accept unnecessary systematization of how I use it.
Both of which are just blocks of text being stomped into an LLM.
I've got a feeling that prompting as we're doing currently (huge blocks of text with very little structure), at least when looking for a deterministic result, will be looked back on as if we were living in the Stone Age. Not sure what the alternative will be, but I'm imagining something like a CI pipeline definition in YAML.
Isn't this just what a skill is?
>Tell the user that you will start by asking a few questions, one at a time, to see what they already know.
This puts it into the session each time. Do this once and it seems to work. By the 5th time, you are getting tired of reconfirming your mastery. This ends up depending upon the user to have a pretty solid foundation of what they know and then use this skill to expand the boundary a bit, and even then the AI can end up making so many assumptions that a person has trouble following it along.
I was thinking some sort of concept mapping kept long term might help, but such a mapping can become really difficult to build. Which should be expected when we consider that teaching isn't just its own discipline, but has many sub-disciplines targeted to specific areas (even if some of it, like classroom management, doesn't apply to this scenario).
I made a few attempts at a knowledge tracking tool for personal use to track how much I knew Japanese beyond a simple JLPT level or similar with the plan to tie it into something like this, but it wasn't great. I could 'cheat' and just tap into my Wanikani API to see my progress, but that still skips so many other areas of Japanese, and even when WaniKani says something is burned, I really want AI to still use it until I truly internalize it like I do words in English. I suspect other areas of knowledge are similar.
In my work, this typically means an opinionated local folder structure and a dedicated /init skill that implements it. Immediately this means we have more than one skill so we’re talking about a plugin now.
Then I think about meta-state. Not just “what is on disk” but “where are we in this application’s multi-session lifecycle?” For Socratic tutoring, that means checking to see if we’ve completed an initial intake, then what skill levels the user has demonstrated, and finally what’s next on the menu.
For this specific skill, I’m surprised it doesn’t say anything about curriculum. It just assumes the model knows the topic and can improvise both a curriculum and credible assessments of progress in that curriculum. That can’t be true for all things people want to learn, so we need a /build-curriculum skill. And probably a /validate-curriculum skill for actual experts to use before launching a given curriculum.
Anyway, I’ve been thinking a lot about this kind of thing as I roll out skills to my firm.
A positive take: this works much better than the many projects on gh that did this before Jane Street. This would be evidence that creativity and execution still matter, even when the product is just an md file
A cynical take: network effects are everything. tens of projects doing the same thing and this only gains traction when an already popular org does it
I remember with ChatGPT 3.5 people were saying "LLM socratic tutor will revolutionize education"
Here we are 3 years later (almost to the day) and well, education isn't revolutionized in the way people imagined.
Running it with my local models it is a pretty helpful way to study up something where you already have some bases.
Maybe it improves things a bit though. I might try it next time.