spork: LLM Etiquette

back to home LLM Etiquette<br>July 1, 2026<br>ai<br>llms<br>workplace-interaction<br>If you are as much of a judgy queen as I am, you probably avoid scrolling through your LinkedIn feed as a prophylactic against cringe. I’d say 90% (don’t quote me) of the content on this site has always been this bizarre collection of disingenuous, emoji-laden, form-content. I’ve always thought that, barring some pathological disorder in their psyche, writing such drivel must be an exhausting task for the authors of these posts. Since the advent of AI, a vast swath of this disingenuous content has become automated. It makes sense, most folks don’t actually view these posts as high-value so why not offload it to the machine?

I think it is time that LLM users start developing rules around how they use those tools when interacting with each other. Now, I’m not expecting them to start writing their own LinkedIn posts again. Though it would be nice, that place was a lost cause well before AI started writing everything. That being said, I have noticed the use of LLM chatbots more frequently cropping up in my day-to-day interactions. That is something that needs to be nipped in the bud.

Lets walk through a hypothetical situation.

Holcomb and Truman are both data engineers at Middling Millennial Software Company. Eudora, a stakeholder from the marketing department raises a concern to Holcomb that some metrics displayed in the company’s BI tool aren’t aligned with the reporting she gets in the CRM. Holcomb looks at the dashboard and determines it is powered by a DBT model. Holcomb doesn’t like to read very much code, so he spins up Claude and sends it off to review the model.

Claude comes back and reports a critical flaw in the model’s logic that must be remediated immediately. Holcomb finds this very alarming, copies the response given to him by Claude, and pastes into a slack thread with Truman. Truman is very annoyed by the contrastive rhetorical framing and em-dashes littering the message. It is blatantly obvious that Holcomb put no effort into understanding the problem or writing the message.

Truman goes on to investigate the issue and finds, as he suspected, that Claude was wrong. There is no problem in the model logic. The same logic powers the data in the BI tool and the CRM. Claude simply didn’t have the context of where the CRM gets it data. This was actually a timing issue. The BI tool live queries the data warehouse whereas the CRM gets updated by a batch job. Truman reached out to Eudora and let her know about the delay, and she was satisfied.

In this hypothetical, Holcomb broke a few rules that I wish LLM users would follow:

Validate the output of the LLM.

Put effort into relaying the information in your own words.

If you must copy and paste, denote that the output came from an LLM.

Everyone knows that LLMs are frequently wrong. When you don’t take the time to validate the output of the LLM you tell the other person you’re interacting with one of two things: either you don’t actually know what it is talking about or you are too lazy to make sure the information you are sharing is correct.

To show you know what you’re talking about you can put the provided output in your own words. Not only will you prove you went through the first step, but you’re also going to get rid of all the annoying telltale signs of an AI response.

Now, if you don’t really know what it is talking about, you can just say that. Don’t try and pass the content off as your own because, once again, it makes you look like a rube. If you simply say, “Hey, I asked Claude about xyz and this is what it came back with. Do you know what it means?” before you paste in the response that will communicate that you respect the intelligence of the person you’re chatting with.

As an AI-user, these are just a few things you can do to make yourself seem less insufferable. Ideally, you wouldn’t be using AI at all and you would just consult the fucking documentation or use the features of your IDE to navigate the codebase. But hey, I don’t want to ask too much of you, so I’ll leave you with this: you’re a very special boy and nothing you could ever do would make you a bad person.

More Posts<br>&larr; prev next &rarr;

all posts Hello World! 2025-10-07<br>Spooky Season 2025-10-08<br>Style Test 2026-02-26<br>LLM Etiquette 2026-07-02