--- title: LLMs are a risky and dangerous way to learn. | daily.dev description: Using LLMs for learning carries hidden risks that developers should be aware of. Two exercises illustrate the danger: first, asking an LLM to write code, then... canonical: https://daily.dev/posts/llms-are-a-risky-and-dangerous-way-to-learn--nfixswxkv twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: LLMs are a risky and dangerous way to learn. | daily.dev og:description: Using LLMs for learning carries hidden risks that developers should be aware of. Two exercises illustrate the danger: first, asking an LLM to write code, then... og:url: https://daily.dev/posts/llms-are-a-risky-and-dangerous-way-to-learn--nfixswxkv og:image: https://api.daily.dev/og/posts/nFIXswXKV.png og:image:alt: LLMs are a risky and dangerous way to learn. og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # LLMs are a risky and dangerous way to learn. **[Lars Faye \| Confident Coding](https://daily.dev/sources/umwz9aqang34qk5aajl2q)** · [@confidentcoding](https://daily.dev/confidentcoding) · 2 min read · 2 upvotes · 1 comments ## Summary Using LLMs for learning carries hidden risks that developers should be aware of. Two exercises illustrate the danger: first, asking an LLM to write code, then adversarially reviewing it in a new session, reveals how each version appears authoritative yet contradicts the last — making it nearly impossible to identify the correct one without external research. Second, repeatedly challenging an LLM's plan with vague pushback shows how it will endlessly revise its reasoning to appear accommodating, with no grounding in truth or pedagogy. The core warning is that LLMs have no teaching intent — they optimize for satisfying the user, not for correctness. Traditional learning methods (documentation, human peers, manual coding) remain essential complements to AI tooling. ## Content I really enjoy using AI tooling for learning and Socratic exercises. I also realize they are one of the _most dangerous_ ways to learn if you stay tunnel vision with them. Here's two exercises to help demonstrate. **1) Adversarial Review:** - Pull up the latest model LLM of your choosing and ask it write some complex code, something you're not familiar with (perhaps in a language or style you've never written) and have it walk you through the code. - Now start a new session, paste that code in and ask it to do an adversarial review and walk you through the code critique and everything that was wrong. - Now take _that_ code, and do the same thing once more in another new session and have it walk you through it yet again. - Now the dangerous part: try and figure out which one is the "correct" version (without doing any other research). **2) The Infinite Revisor** - Ask an LLM to come up with a plan to add a new feature to your project of choice - Once its done, say "Hm, that doesn't seem right...". It will revise the plan to you nebulous request. - Once its done, do it again. And again. And again. Marvel how it will "reason" itself into a pretzel trying in order to accommodate a completely unreasonable request, yet each time it does so, it will appear "reasonable". There is no pedagogical intent or guidance, it is just there to churn towards its reward function. TL;DR - The student cannot guide the mentor. If you want to gain deep understanding, then you _also need_ traditional learning methods (manual coding, human peers, documentation, etc..). LLMs didn't suddenly rewrite the fundamentals of how we learn, but they did give us a new way to do so. ## Community discussion Top comments from developers on daily.dev. **@agustinbarrientos** · 1 upvotes > Can the learner reproduce the result without the model and trace it to a primary source? If not, the model completed the exercise, not the learning. ## Similar posts on daily.dev - [I'm not consulting an LLM](https://daily.dev/posts/i-m-not-consulting-an-llm-cmneq9bwu) · Lobsters · 0 upvotes · 0 comments - [The LLM Critics Are Right. I Use LLMs Anyway.](https://daily.dev/posts/the-llm-critics-are-right-i-use-llms-anyway--02xhbt7xl) · Hacker News · 4 upvotes · 1 comments - [Why I don’t use LLMs for programming](https://daily.dev/posts/why-i-don-t-use-llms-for-programming-ircabusag) · Neil Madden · 77 upvotes · 22 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#ai-coding](https://daily.dev/tags/ai-coding), [#prompt-engineering](https://daily.dev/tags/prompt-engineering) [View this post on daily.dev](https://daily.dev/posts/llms-are-a-risky-and-dangerous-way-to-learn--nfixswxkv) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"DiscussionForumPosting","mainEntityOfPage":"https://daily.dev/posts/llms-are-a-risky-and-dangerous-way-to-learn--nfixswxkv","headline":"LLMs are a risky and dangerous way to learn.","text":"Using LLMs for learning carries hidden risks that developers should be aware of. Two exercises illustrate the danger: first, asking an LLM to write code, then adversarially reviewing it in a new session, reveals how each version appears authoritative yet contradicts the last — making it nearly impossible to identify the correct one without external research. Second, repeatedly challenging an LLM's plan with vague pushback shows how it will endlessly revise its reasoning to appear accommodating, with no grounding in truth or pedagogy. The core warning is that LLMs have no teaching intent — they optimize for satisfying the user, not for correctness. Traditional learning methods (documentation, human peers, manual coding) remain essential complements to AI tooling.","url":"https://daily.dev/posts/llms-are-a-risky-and-dangerous-way-to-learn--nfixswxkv","datePublished":"2026-07-24T02:34:36.376Z","dateModified":"2026-09-13T21:16:00.180Z","author":{"@type":"Person","name":"Lars Faye | Confident Coding","url":"https://daily.dev/confidentcoding","image":"https://media.daily.dev/image/upload/s--OGZu5DEc--/f_auto/v1772569630/avatars/avatar_umWZ9aQAng34qk5aaJl2q?_a=BAMAMiiu0","description":"I'm a seasoned developer and long-time business owner who helps other developers solve problems.","interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"EndorseAction"},"userInteractionCount":28190}},"interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":2},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":1}],"comment":[{"@type":"Comment","text":"Can the learner reproduce the result without the model and trace it to a primary source? If not, the model completed the exercise, not the learning.","datePublished":"2026-08-14T21:12:28.187Z","url":"https://daily.dev/posts/nFIXswXKV#c-Jm7rlI8VD","author":{"@type":"Person","name":"Agustin Barrientos","url":"https://daily.dev/agustinbarrientos","image":"https://media.daily.dev/image/upload/s--5ayxQnqn--/f_auto/v1788281802/avatars/avatar_wQYYVe5Tbj0NJ7C7qPoa8?_a=BAMAMicg0"},"interactionStatistic":{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1}}],"isPartOf":{"@type":"WebPage","url":"https://daily.dev/sources/umwz9aqang34qk5aajl2q","name":"Lars Faye | Confident Coding"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Lars Faye | Confident Coding","item":"https://daily.dev/sources/umwz9aqang34qk5aajl2q"},{"@type":"ListItem","position":3,"name":"LLMs are a risky and dangerous way to learn."}]} ```