# I don’t want to read LLM output

**URL:** <https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244>\
**Category:** Meta Discussion\
**Created:** [September 7, 2026, 2:28pm UTC](https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244 "2026-09-07T14:28:41Z")\
**Posts on this page:** 15\
**Page:** 8

<div class="post-metadata">

**Author:** ![rafael.guerra](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/rafael.guerra/32/216610_2.png) [@rafael.guerra](https://discourse.julialang.org/u/rafael.guerra)\
**Post date:** [September 18, 2026, 10:11am UTC](https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244/150 "2026-09-18T10:11:20Z")

</div>

> [@gwr](#):
>
> illuminating the fact that storage space also appears to incur negligible costs

If Archimedes were alive today, he would say: “Give me a checkpoint and enough tokens to stand on, and I will move an entire universe of slop to the archives.”

---

<div class="post-metadata">

**Author:** ![mihalybaci](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/mihalybaci/32/13528_2.png) [@mihalybaci](https://discourse.julialang.org/u/mihalybaci)\
**Post date:** [September 18, 2026, 1:15pm UTC](https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244/151 "2026-09-18T13:15:36Z")

</div>

It seems to me that there need not be a bullet (3) (rhyming not intentional!), but rather is borne of (1) and (2). By making a post in a friendly human manner (1) that also contains knowledge or adds to the thread (2), good reputation (3) is satisfied automatically.

> [@gwr](#):
>
> it could also be part of the solution by, for example, summarizing a discussion after some time, archiving it, and then storing only the summary.
> 
> What do you think?

I very much dislike this idea. For one, I do not trust AI summaries as far as I can throw them. LLMs are still prone to missing key details and over emphasizing others. I definitely do not give anyone permission to use an LLM to repost a summary including any of my thoughts.

Second, part of why I enjoy reading threads is to follow the thought process. There were multiple mega-threads (\>200 posts) on the best way to output random Float32 values. And for me the value in those threads is not as much in the summary, but in the back-and-forth of the pros and cons. Could an LLM summarize that discussion? Sure. Do I trust an LLM to get all the nuances of those threads? No, absolutely not.

> [@gwr](#):
>
> So, are we supposed to write as if we were speaking casually in a group in the kitchen, while being recorded for an AI personality dossier?

For me, absolutely yes. If a company goes so far as to deny me a job because they feel my posts in this forum aren’t “professional enough”, then I certainly do not want to work for that company.

A good analogy for me is this forum is the poster session at a scientific conference. Some people are there to show off their cool work. Some are more there to learn and ask questions. Some are there just for the experience and to talk to other similarly minded people. Some of the conversions are very in-depth and technical, while others may eventually drift off-topic into more personal discussions. The rules of decorum are somewhat casual, but everyone is still expected to act in a professional manner.

The bottom-line for me, and I imagine others, is that I am here to learn, help (when I can), and interact with other **humans**. I would never go to a conference if I thought I would just be reading LLM-generated summaries . If I wanted to read the dry, dense, inauthentic output of an LLM, then I would be using one, not posting here. I have a choice to switch to LLMs for all my Julia-knowledge and leave Discourse entirely, and I chose not to because I enjoy reading what **people** write, regardless of precision or imperfection.

---

<div class="post-metadata">

**Author:** ![BananaMaster3](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/bananamaster3/32/218457_2.png) [@BananaMaster3](https://discourse.julialang.org/u/BananaMaster3)\
**Post date:** [September 20, 2026, 7:59pm UTC](https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244/152 "2026-09-20T19:59:02Z")

</div>

> [@kahliburke](#):
>
> Again, we’re talking about piece of content which is optional to read. If you see an announcement thread, click into it, and decide it’s written in a way that doesn’t fit your taste, why not stop reading it and switch to something you find more enjoyable? We’re not talking about misuse of the forum, spam, bots, or denial of service. We’re talking about people who are creating packages they think might be useful or interesting to others, sharing that and gifting it to the community. Why not celebrate this a little more? Why is there so much negativity here?

This doesn’t work in tech areas, where community is considered extremely important. People move entire operating systems or programming languages just because of the community.  
Sure, I might not read it, but the posts made here define who uses Julia and what Julia is. Watch that meaning carefully.

---

<div class="post-metadata">

**Author:** ![gwr](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/gwr/32/33632_2.png) [@gwr](https://discourse.julialang.org/u/gwr)\
**Post date:** [September 21, 2026, 3:21pm UTC](https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244/153 "2026-09-21T15:21:39Z")

</div>

> [@mihalybaci](#):
>
> For me, absolutely yes. If a company goes so far as to deny me a job because they feel my posts in this forum aren’t “professional enough”, then I certainly do not want to work for that company.

That seems to me a bit like the argument: “Why do we need data protection, I have nothing to hide?” I can understand your preference, but I don’t quite see how this could become a global norm? The situation of an academic with a permanent position or even “tenure-track” could, for example, be different from that of a freelancer or self-employed person.

> [@mihalybaci](#):
>
> A good analogy for me is this forum is the poster session at a scientific conference. Some people are there to show off their cool work. Some are more there to learn and ask questions. Some are there just for the experience and to talk to other similarly minded people. Some of the conversions are very in-depth and technical, while others may eventually drift off-topic into more personal discussions. The rules of decorum are somewhat casual, but everyone is still expected to act in a professional manner.

I would agree with that if this forum were a similarly protected space as a poster session. But that’s precisely not the case! At a poster session, I can see who is present, and based on my experience so far, not all statements are recorded literally for an indefinite period of time so that AI can search everything I’ve said there on behalf of completely unknown persons (this is not a protected space). Additionally, I cannot even delete my older posts on my own responsibility (private autonomy?).

---

<div class="post-metadata">

**Author:** ![mbauman](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/mbauman/32/31082_2.png) [@mbauman](https://discourse.julialang.org/u/mbauman)\
**Post date:** [September 21, 2026, 3:45pm UTC](https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244/154 "2026-09-21T15:45:04Z")

</div>

I really don’t understand where you’re going with this. Yes, this is a public forum. Yes, that means that you shouldn’t post something you’d regret if it were featured in the NYT. It doesn’t mean, however, that you need to be perfect or have every post be at the quality of a scientific journal. And it definitely doesn’t mean that you should have some LLM replace your voice with its generic output.

---

<div class="post-metadata">

**Author:** ![gwr](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/gwr/32/33632_2.png) [@gwr](https://discourse.julialang.org/u/gwr)\
**Post date:** [September 21, 2026, 4:33pm UTC](https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244/155 "2026-09-21T16:33:07Z")

</div>

> [@mbauman](#):
>
> I really don’t understand where you’re going with this.

Let me see if I understand you correctly: “Please write like a human—casual and relaxed, with mistakes—but keep in mind that your casual forum post will be treated as if it were published in the New York Times?” Isn’t that exactly the difference between a conversation at the water cooler or at a poster session?

The NYT publishes editorially vetted and edited content, and it is clear that these pieces end up in a searchable archive. The prestigious British magazine _The Economist_, by the way, traditionally doesn’t even list the authors’ names. Here, on the other hand, everything ends up in a _with AI_ (!) searchable archive and is meant to be immediate and as human-like as possible, because the focus is not on quality but solely on the author.

If I understand correctly, the fact that posts here are stored permanently without users being able to delete them was already critically discussed when it was not yet possible to create dossiers using AI?

I fully respect that others here have different preferences. However, I cannot see where my argumentation should be illogical or unobjective.

---

<div class="post-metadata">

**Author:** ![Benny](https://avatars.discourse-cdn.com/v4/letter/b/49beb7/32.png) [@Benny](https://discourse.julialang.org/u/Benny)\
**Post date:** [September 21, 2026, 5:01pm UTC](https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244/156 "2026-09-21T17:01:32Z")

</div>

mbauman was not saying you’d be the author of the NYT article, rather you’d be a subject whose public comments are widely disseminated and scrutinized. American politicians certainly don’t mind journalistic standards when they open their mouth.

I think it’s also besides the point. Being personally satisfied by our own posts is easy, complying with community standards we don’t fully agree with is hard. I could also make an argument for casual profanity here that I find completely logical and objective, doesn’t mean it’ll go anywhere.

---

<div class="post-metadata">

**Author:** ![mihalybaci](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/mihalybaci/32/13528_2.png) [@mihalybaci](https://discourse.julialang.org/u/mihalybaci)\
**Post date:** [September 21, 2026, 5:30pm UTC](https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244/157 "2026-09-21T17:30:03Z")

</div>

> [@gwr](#):
>
> That seems to me a bit like the argument: “Why do we need data protection, I have nothing to hide?”

No, I still expect privacy for my **private** conversions. This is a **public** forum, and so there is no expectation of privacy.

> [@gwr](#):
>
> I can see who is present, and based on my experience so far, not all statements are recorded literally for an indefinite period of time

Indeed, I think that is largely true as well, however, the growing prevalence of Meta glasses and other smartware may be changing that fact. Imagine tenured professors walking around a conference wearing smart glasses to record all of their conversions and using that footage in hiring decisions. That possibility seems at least as plausible as AI scraping forums for hiring panels, so should people never attend a conference again unless they have an AI helper to speak for them?

> [@gwr](#):
>
> However, I cannot see where my argumentation should be illogical or unobjective.

This is where the the argument falls apart for me. In your scenario, the publishing standard for this public forum should be as high as the New York TImes or a prestigious research journal. That seems absurd. While I agree that [“anything you say can and will be held against you”](https://en.wikipedia.org/wiki/Miranda_warning) in a **public** space, I do _not_ think that is a valid reason to allow 100% LLM-generated posts, which is ultimately what we are talking about here.

I would also say that if a person is so concerned about making a minor spelling/grammar mistake and having it completely ruin their career, which is what we’re discussing now, then that person probably shouldn’t be posting in a public forum.

---

<div class="post-metadata">

**Author:** ![mbauman](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/mbauman/32/31082_2.png) [@mbauman](https://discourse.julialang.org/u/mbauman)\
**Post date:** [September 21, 2026, 5:50pm UTC](https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244/158 "2026-09-21T17:50:49Z")

</div>

> [@gwr](#):
>
> but keep in mind that your casual forum post will be treated as if it were published in the NYT

Yeah, I can see how that reads contradictorily, but I don’t mean it in that manner. It’s not that your post _becomes_ an NYT article itself, but rather that an article in a prominent newspaper _describes_ your post **and its context** to an even larger audience than you may have expected here. I used that example because it’s the very example our guidelines use — and have _always_ used — referencing the fact that search engines exist.

This is all getting quite a bit aside from using an LLM to wholesale write your post. Honestly, that’s the thing I find most embarrassing about some of my own posts upon re-read: they happen include some of Claude’s favorite phrases and formulations and therefore seem like I might’ve not written it myself.

I have posted a lot here, and I was curious what the bot would highlight as my most embarrassing… and sure enough it found that one time I still remember where I was very confidently incorrect. But it also roasted me for having _just posted_ that nobody should let an LLM replace their voice and describing LLMs as stochastic parrots… only to hand my decade of writing to the parrot and ask it which parts were embarrassing! 😂

---

<div class="post-metadata">

**Author:** ![kahliburke](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/kahliburke/32/208472_2.png) [@kahliburke](https://discourse.julialang.org/u/kahliburke)\
**Post date:** [September 22, 2026, 4:27am UTC](https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244/159 "2026-09-22T04:27:14Z")

</div>

This post was temporarily hidden by the community for possibly being off-topic, unfocused, inappropriate, or spammy.

---

<div class="post-metadata">

**Author:** ![mihalybaci](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/mihalybaci/32/13528_2.png) [@mihalybaci](https://discourse.julialang.org/u/mihalybaci)\
**Post date:** [September 22, 2026, 12:04pm UTC](https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244/160 "2026-09-22T12:04:55Z")

</div>

This is a thread about LLM use in posts in general. I think it’s best to avoid talking about anyone in particular as it adds nothing to the conversation.

If you wish to ask folks about how they enjoy your posts, then opening a new thread is the way to go.

---

<div class="post-metadata">

**Author:** ![Marc.Cox](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/marc.cox/32/7514_2.png) [@Marc.Cox](https://discourse.julialang.org/u/Marc.Cox)\
**Post date:** [September 22, 2026, 8:14pm UTC](https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244/161 "2026-09-22T20:14:03Z")

</div>

@ [gwr](https://discourse.julialang.org/u/gwr) yes I believe your rules just above here make sense. 👍

---

<div class="post-metadata">

**Author:** ![kahliburke](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/kahliburke/32/208472_2.png) [@kahliburke](https://discourse.julialang.org/u/kahliburke)\
**Post date:** [September 22, 2026, 10:54pm UTC](https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244/162 "2026-09-22T22:54:29Z")

</div>

This post was temporarily hidden by the community for possibly being off-topic, unfocused, inappropriate, or spammy.

---

<div class="post-metadata">

**Author:** ![adienes](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/adienes/32/37459_2.png) [@adienes](https://discourse.julialang.org/u/adienes)\
**Post date:** [September 22, 2026, 11:11pm UTC](https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244/163 "2026-09-22T23:11:09Z")

</div>

This post was temporarily hidden by the community for possibly being off-topic, unfocused, inappropriate, or spammy.

---

<div class="post-metadata">

**Author:** ![kahliburke](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/kahliburke/32/208472_2.png) [@kahliburke](https://discourse.julialang.org/u/kahliburke)\
**Post date:** [September 23, 2026, 12:11am UTC](https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244/164 "2026-09-23T00:11:43Z")

</div>

This post was temporarily hidden by the community for possibly being off-topic, unfocused, inappropriate, or spammy.

[Previous page](https://discourse.julialang.org/t/i-don-t-want-to-read-llm-output/139244.md?page=7)