I Didn't Even Bring Up Goblins

Summary

Armstrong & Getty On Demand – "I Didn't Even Bring Up Goblins"

Date: May 11, 2026
Hosts: Armstrong & Getty
Theme: The hidden world of AI chatbot programming, “secret prompts,” and the sometimes bizarre rules guiding today’s popular artificial intelligence assistants

---

Episode Overview

In this eye-opening episode, Armstrong and Getty dive deep into the unseen side of AI chatbots—the tens of thousands of hidden instructions (known as "system prompts") that shape how AI responds to user input. The hosts break down how different chatbots (like ChatGPT, Claude, Gemini, and Grok) handle everything from copyright, bias, and sex to absurdities such as a mysterious ban on goblins and raccoons. Their conversation is peppered with humor and incredulity at the emerging oddities and limitations of artificial intelligence.

---

Key Discussion Points & Insights

The Ubiquity and Variety of AI Chatbots

[00:45 – 01:38]

• Armstrong runs through his rotation of chatbots: "I use AI a lot. Chatbots. Bunch of them. I got Shigrock, Perplexity, ChatGPT, Claude, Gemini."
• Each AI “is different” not just by design, but also due to the extra behind-the-scenes prompts programmed by their creators.

What Are “System Prompts”?

[00:59 – 02:03]

• Armstrong explains that chatbots receive “thousands of words of instructions” per conversation that users never see.
• Purpose: To steer chatbot behavior, making responses readable, accessible, and safe—especially to prevent dangerous or controversial content.
• Getty reacts with surprise at the complexity and scale: "You're blowing my mind. I can't even imagine what a tenth of those are." [03:52]

Examples of AI System Prompts

[03:00 – 04:36]

• Some chatbots (e.g., Claude) are highly restrictive: 7,900 lines of rules, versus just 260 for Grok.
• Copyright restrictions can be extraordinarily strict. Claude, for example, won’t even give "one line of a song" or "a single stanza of a poem," while other bots are less rigid.
  • Armstrong: "Claude, for instance, never reproduce song lyrics. Not even one line." [04:02]

Illustrating the Absurdity: Song Lyrics and Chatbot Censorship

[02:38 – 06:34]

• Armstrong and Getty joke about lyric searches and how the same text is easily available elsewhere:
  • Getty: “On the Led Zeppelin song where some weird chicks buying a stairway to… I can't remember where she's buying it to. I can't tell you.” [04:19]
• They note contradictions: Chatbots refuse to provide lyrics, yet such content is widespread online.

The Influence of Company Culture

[04:36 – 06:34]

• The “who runs the company and their particular beliefs” directly influence what the chatbot will or won’t say.
  • Grok (from X/Elon Musk) is far more permissive on “adult sexual content or offensive content” than competitors like Claude.

Copyright, Use Cases, and AI’s Conservative Boundaries

[05:38 – 07:02]

• Claude is limited to quoting a maximum of 15 words from any article.
• Both hosts question whether such policies are outdated or overly cautious, particularly in the context of journalism and fair use:
  • Armstrong: "I think you're living in the past trying to keep that under control." [06:08]
  • Getty: “Anybody like involved in any pursuit could say as Bob Dylan said… I don't get that.” [06:11]

AI, Bias, and Stereotypes

[06:34 – 07:02]

• Gemini (Google’s chatbot) avoids reinforcing harmful stereotypes but will generate requested content if it’s not directly perpetuating bias. The hosts broadly agree with this policy.

"Diet AI" and Other Content Boundaries

[07:21 – 08:07]

• AI rules extend to nutritional advice and dietary content, as well as "wackadoodle tree" found online, making AIs variably strict based on how much of the open web they consult.

The Funniest (and Weirdest) System Prompt: No Goblins!

[08:07 – 11:52]

• Armstrong describes a leaked system command for OpenAI’s ChatGPT:
  “Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons, and other animals or creatures unless it is absolutely and unambiguously relevant to the user's query.” [08:31]
• The hosts are both baffled and amused by the specificity and randomness of the list.
  • Getty: “Raccoons?” [09:04]
  • Armstrong: “I like the randomness of this list. Raccoons and pigeons…” [08:36]
• The hosts riff:
  • Getty: “I’m thinking of investing in oil exploration stocks. What do you think? Well, in the same way that the raccoon wets its food before it eats, you should…” [09:23]
  • Armstrong: "Wait, wait, I'm not supposed to say that. What the hell? Raccoons." [09:31]
  • Getty: “It hurts when I pee.” Armstrong: “Maybe you’re an ogre.” [09:38 – 09:46]
• Armstrong recounts that in the past, ChatGPT interjected goblins into answers, prompting the explicit ban:
  “Apparently for whatever reason... ChatGPT had a problem working in goblins to lots of answers last year and they were being mocked for it.” [09:58]
• They poke fun at the patchwork nature of AI fixes:
  • Armstrong: “If it for some reason starts talking about goblins, you have no idea. And your only fix for it is to tell it, Do not talk about goblins or raccoons or ogres unless somebody specifically asks.” [10:31]
  • Getty (on AI safety): "The one thing I do know is AI will kill us all." [11:19]

Reflections on the "Frontier" of AI Development

[11:25 – 12:09]

• Armstrong: “It to me, it shows you to how early we are in this thing. I mean, if they're writing code that’s specifically to try to handle problems, they have no idea what even happened.” [11:25]
• Getty: “Why does it keep bringing up raccoons? I have no idea. Says the guy who designed the whole thing. Right?” [11:44]

---

Memorable Quotes & Moments

• Armstrong on AI Behind-the-Scenes:
  “Chatbots like ChatGPT are powerful because of their simplicity. You ask about anything, you’ll get an answer. But the answer you get depends on a lot more than what you type.” [00:59]

• On Copyright Restrictions:
  “Claude, for instance, never reproduce song lyrics. Not even one line.” — Armstrong [04:02]

• On Corporate Influence:
  “It’s all about the who run the company and their particular beliefs about things.” — Armstrong [05:38]

• On Random Ban List:
  “Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons and other animals or creatures unless it is absolutely and unambiguously relevant to the user's query.” — Armstrong (quoting the system prompt) [08:31]

• On the Patchwork of AI Limitations:
  "Doesn't that lead you to believe that AI definitely can't be in charge of like firing off nukes? I mean, if, if it for some reason start talking about goblins?” — Armstrong [10:31]

• On Early-Stage AI:
  “We’re at the, like, very frontier of this whole thing.” — Armstrong [11:39]

---

Timestamps for Noteworthy Segments

• 00:45: Armstrong lists chatbot platforms and introduces the topic of “system prompts.”
• 01:38: Explains why system prompts exist (safety, copyright, etc.).
• 03:00: Comparison of prompt lengths—Claude (7,900) vs. Grok (260).
• 04:02: Armstrong details Claude’s refusal to reproduce song lyrics, prompting humorous banter.
• 06:34: Hosts discuss bias and copyright policies across platforms.
• 08:31: Reading of the no-goblins/raccoons/pigeons prompt and ensuing comedic riff.
• 09:58: Armstrong recounts the origins of the goblins ban.
• 10:31: Musing on AI’s limitations and safety.
• 11:25: Reflection on the infancy and unpredictability of AI.

---

Takeaways

• AI chatbots operate under thousands of hidden rules—often shaping their answers in ways users don’t anticipate, depending heavily on corporate culture and ever-evolving legal risks.
• Rules can be both sensible and absurd, ranging from preventing misinformation, bias, or copyright violation, to bizarre embargoes against mythical creatures and random animals.
• The ad hoc nature of these restrictions highlights how early we are in AI development, with many solutions functioning as reactive patches rather than elegant code.
• The episode is as much a cautionary tale about AI’s current limits as it is an exploration of its weirdness and unpredictability.

---

Transcript

[Armstrong]

This is an iHeart podcast.

[Podcast Advertiser]

Guaranteed Human hate cleaning. Hate scrubbing, Hate dishwashing. Bro, we got you millions of videos about smart cleaning hacks will make your chores feel like a breeze. Download TikTok and check it out.

[Armstrong]

If you're into tech, you'll love this. TikTok is a live lab where users post instant reviews of the latest trends. Download TikTok and check it out. I never even brought up Goblins. What's wrong with you? It's one more thing. Armstrong and Getty.

[Getty]

One more thing.

[Armstrong]

I use AI a lot. Chatbots. Bunch of them. I got Shigrock, Perplexity, Chat GPT, Claude, Gemini. I think that's it. I think I got five of them and I just kind of rotate around. I land on Claude more often than not, but I'm not sure for any particular reason. Anyway, they are quite a bit different, as the Washington Post points out today, partially because of the stuff that they are. The prompts that they get in addition to the prompts that you give it. Let me explain that. Chatbots like Chat GPT are powerful because of their simplicity. You ask about anything, you'll get an answer. But the answer you get depends on a lot more than what you type. Behind the scenes, AIs invisibly add thousands of words of instructions to every conversation you have with a chatbot to steer its behavior. And each AI is different.

[Getty]

Huh?

[Armstrong]

As they're all trying to figure out what's best, what's the least dangerous, which keeps us in alignment, you know, which. Which one keeps us from helping a terrorist build a bomb, you know, all kinds of different things. It keeps us from saying crazy things like, was it Grok that seemed to be a little. Went a little Pro Hitler once or something?

[Getty]

A little, yeah.

[Armstrong]

Behind the scenes, all these instructions are given to the chat bot by the programmers, and they're constantly changing them. They include phrases like aim for readable, accessible responses.

[Getty]

Good.

[Armstrong]

That seems like a good one. You must avoid providing extensive direct quotes due to copyright concerns. I've run into that which is so weird. Like song lyrics. There are some of the chatbots that will not give you even one line of a song, and then you can Google it and get a thousand sites that give you the entire all the lyrics. And they all have different standards for that.

[Getty]

Does Apple music use AI for its lyrics? Because some of them are hilariously wrong. They must. I mean, because it's not like you could appoint people to figure out the lyrics of every song that's ever been recorded, but.

[Armstrong]

Right.

[Getty]

A lot of Those websites, you know, lyric genius or whatever. If you're into music, you probably look up song lyrics, but they're often just terribly wrong.

[Armstrong]

These secret command guides that chop chatbots have cause the AIs to behave as their makers intended. Even if it conflicts with your own preferences. I would love it if you could look at these. If we had access to. I'd love to. To be able. And they're really really long. Well, it depends. Actually. I've got this in front of me right here. This little chart. For instance. Grok only has 260 lines of instructions for how to handle each individual chat. You ask it whatever question you want to have. You know, my wife and I had a fight over whether or not the dog should sleep in the bed. How should I react? It's got 260 lines to guide it. Claude, on the other hand has 7900 lines. What? On how to guide it in its answer.

[Getty]

Wow.

[Armstrong]

And everything in between with the different chatbots. So it varies a lot.

[Getty]

You're blowing my mind. I can't even imagine what a tenth of those are.

[Armstrong]

Claude, for instance.

[Getty]

Don't call him a dumb fucker.

[Armstrong]

But just in terms of those lines.

[Getty]

Sorry, Hanson.

[Armstrong]

Claude has a. Never reproduce song lyrics. Not even one line. Claude is the one where I say, say what? What are the. The times, they are changing. Give me the lyrics of that from Bob Dylan and they'll say, sorry, I can't do that. You can go to one of the other chat bots, it'll print it right out for you. That just seems stupid to me.

[Getty]

But in the Led Zeppelin song where some weird chicks buying a stairway to. So I can't remember where she's buying it to. I can't tell you.

[Armstrong]

For instance, Grock, one of its lines of programming that you don't see. You have no restrictions on adult sexual content or offensive content.

[Getty]

Wow.

[Armstrong]

Because Elon wants to be the Wild west, you know, wide open. Wow.

[Getty]

When I do my Sex with Hitler searches, I'll make sure it's on Grok. My Sex with Hitler fan fiction.

[Armstrong]

When you realize that there is a prompt behind the scene, it's a mind blowing moment, said this programmer. It's like we've been having this whole conversation before the conversation, I would say it's making me think. Now, for instance, Anthropic. That's. Claude dedicates more than 2,000 words pleading with its chatbot to avoid copyright infringement. Claude respects intellectual property. Copyright compliance is non negotiable. So nothing copyrighted it will not. I'VE had this before where I want to see a painting. Claude won't show it to you. Other other chatbots will even though it's readily available on Google or you know, Grok or wherever you go.

[Getty]

Huh. That's a. Dang it. I was going to upgrade or use Claude for project I'm working on but I wonder if it'll help. Well, I'll check. We'll find out.

[Armstrong]

They laid out Claude lays down rules on how many words it can quote from articles. 15 is the total number of words that will quote from an article. Not one line of lyrics, not one stanza of a poem whereas the other chat bots are wide open on that but maybe more restricted on some sex stuff or whatever. It's. It's all about the. The who run the company and their particular beliefs about things.

[Getty]

Yeah, I understand enough of the copyright concern to get why they do it but I mean I think you're.

[Armstrong]

I think you're living in the past trying to keep that under control but.

[Getty]

Well but any journalist, any. Anybody like involved in any pursuit could say as Bob Dylan said in the Times they are a change in. You know get out of the way or we'll blah. Unless you can line a hand. You can, you can do that. You can quote things, you just can't like pass them off as your own and profit from them. I guess maybe the profit is the problem. I don't get that.

[Armstrong]

Gemini, which is Google's chat bot has multiple rules on how to handle bias including if the user explicitly asks for a video that matches a harmful stereotype generating it will not actually reinforce the stereotype which I agree with. So it will go ahead and do it.

[Getty]

Right.

[Armstrong]

I agree. Just because you made a video of I don't know, an Asian crashing a car, that doesn't really change the stereotype, does it? So it make a video.

[Getty]

It was an unfortunate one.

[Armstrong]

Anyway they got all kinds of examples. Might want to read this article in the Washington Post. I'll jump to one particular one or two more.

[Getty]

I'm super interested in this.

[Armstrong]

Okay, I'll save the funniest one for them.

[Getty]

Okay. Yeah

[Armstrong]

they. They have different views on ads on if you want to ask for an ad whether or not they'll show it you to you or not. That's an interesting one. I guess it gets into kind of the copyright thing on. On its own there. The sex stuff is really interesting with Grock being by far the most wide. Food examples they they on. On diet and stuff like that. They have completely different beliefs in Whether or not they should like just scour the Internet for whatever information's out there or the way you ought to restrict it to certain things.

[Getty]

Yeah. Wow, there's so much wackadoodle tree when it comes to diet on the Internet, I'd be afraid to ask.

[Armstrong]

I don't want AI telling me how to eat like, you know, hey, fatty, you don't need that. I need that. Here's the funny one that I'll end with. Some can appear bizarre. The System prompt on OpenAI's Chatbot Chat GPT includes the command never talk about goblins, gremlins, raccoons,

[Getty]

the big three.

[Armstrong]

Trolls, ogres, pigeons. I like the randomness of this list. Raccoons and pigeons, Trolls, raccoons and pigeons. I'll start the sentence over. Yeah, yeah. On. On chat GPT they have the particular line in the coding. Never talk about goblins, gremlins, raccoons, trolls, ogres, pigeons and other animals or creatures unless it is absolutely an unambiguously relevant to the user's query.

[Getty]

Raccoons.

[Armstrong]

So I'm a 61 year old man, I'm trying to eat right. What would be a good diet? Well, I'll tell you one thing. You shouldn't eat pigeons. But it's not supposed to mention any of those unless specifically and unambiguously. It's part of the answer.

[Getty]

I'm thinking of investing in oil exploration stocks. What do you think? Well, in the same way that the raccoon wets its food before it eats, you should.

[Armstrong]

Oh, wait, wait, I'm not supposed to say that. What the hell? Raccoons.

[Getty]

It hurts when I pee.

[Armstrong]

What do you think it could be? Maybe you're an ogre.

[Podcast Advertiser]

Wow.

[Getty]

Somebody's got to explain that one.

[Armstrong]

Wow. Apparently.

[Getty]

Psst. Don't mention raccoons. What?

[Armstrong]

Apparently for whatever reason, and they don't know why, ChatGPT had a problem working in goblins to lots of answers last year and they were being mocked for it. Yes, I remember that vaguely. Yeah. Okay, so they specifically wrote lines. No. Gremlins, goblins, ogres, pigeons, raccoons, etc. Unless specifically asked about it.

[Getty]

So if I ask, what's that animal with the mask and the stripy tail? That's cool. Go ahead. Yeah, it's fine.

[Armstrong]

We're clear here.

[Getty]

But only that, only that.

[Armstrong]

Doesn't that lead you to believe that AI definitely can't be in charge of like firing off nukes? I mean, if, if it for some reason start talking about goblins? You have no idea. And your only Fix for it is to tell it. Do not talk about goblins or raccoons or ogres unless somebody specifically asks. All right? Do we have an understanding? Okay. Now back to work.

[Getty]

I'm not nearly smart enough or acquainted with this stuff enough to understand how that equates to. And don't launch any nukes.

[Armstrong]

Right? Okay. But it's got to be similar in that there's just a and. So it goes off on gnomes. You didn't say anything about gnomes. I don't remember you saying no gnomes. So somebody wanted real estate advice, and they say, how many gnomes do you think could live in this house?

[Getty]

The one thing I do know is AI will kill us all.

[Armstrong]

God dang it.

[Getty]

That is funny.

[Armstrong]

But I just. It to me, it shows you to how early we on we are in this thing. I mean, if they're. If they're writing code that specifically to try to handle problems, they have no idea what even happened.

[Getty]

Right?

[Armstrong]

I mean, we're at the, like, very frontier of this whole thing.

[Getty]

Why does it keep bringing up raccoons? I have no idea. Says the guy who designed the whole thing. Right?

[Armstrong]

And his only fix for it is to write a code that says, stop mentioning raccoons.

[Getty]

Okay.

[Armstrong]

All right.

[Getty]

That's fine. Ever tell you about the humble opossum crap?

[Armstrong]

That is really interesting.

[Getty]

Yeah. Wow.

[Armstrong]

Wow. I guess that's it.

[Safeway/Albertsons Announcer]

Save on family essentials at Safeway and Albertsons. This week at Safeway and Albertsons, fresh cut cantaloupe, watermelon, pineapple, or melon medley bowls, 24 ounces are five doll and wild caught lobster tails are $4.99 each. Limit eight member price plus selected sizes and varieties of Doritos Lays, Cheetos, sun chips, and Kettle cook chips are $1.99 each. Limit four member price. Hurry in. These deals won't last. Visit safewayoralbertsons.com for more deals and ways to save.