🤖 I Built a Gamma World Referee Out of AI

LLM the Warbot: "Misuse of 'their' detected!"

Some people train large language models to write poetry or give dating advice.
I trained one to run Gamma World 3rd Edition.

Because if we’re going to let AI take over the world, it might as well start with the irradiated wasteland of post-apocalyptic roleplaying.

---

🎲 What It Is

I created a custom GPT trained on the third edition Gamma World rules, using two searchable PDFs that I legally purchased from DriveThruRPG. It acts as a succinct, rules-savvy AI referee that:

• Answers game mechanics questions quickly

• Cites sources when possible

• Offers elaboration only if requested (because verbosity is a mutation)

Whether you're a mutant plant trying to figure out how radiation affects your chloroplasts, or a GM wondering how many hit points a warbot should have, it’s got you covered.

---

🧠 How It Works

The GPT uses a simple guiding prompt:

“You are an expert in Gamma World 3rd edition rules. Cite your references as much as possible. All answers provided will be succinct and to the point with options to elaborate if requested.”

In other words, it doesn’t ramble. It doesn’t argue. It just tells you what the rules say—and where to look. Think of it as the Oracle of TSR, minus the sarcasm (unless you enable that setting).

---

💬 Sample Output

Here’s a real query I gave it, with its response:

Q: What happens when a mutated human passes through a radiated area?

A: In Gamma World 3rd Edition, a mutated human typically makes a Constitution (CN) check when passing through a radiated area. Failing the check can lead to:

• Temporary effects like nausea, fatigue, or attribute penalties

• Unexpected mutations if the exposure is extreme or repeated

• Long-term damage, such as permanent attribute loss

The specifics (like radiation zones, mutation rolls, or saving throws) are detailed in the Gamma World Referee’s Screen and Mini-Module, not the core box set. Let me know if you’d like a breakdown of those rules.

That tone—efficient, informative, and option-driven—is exactly how the GPT is designed to respond. You can treat it like a pocket rules lawyer that doesn't talk over the players.

---

🔐 Why It's Not Public (Yet)

For copyright reasons, the GPT remains private for now. It's not designed to replace the books—just to help me use them faster and more effectively during sessions. If you own the same PDFs, it's easy to recreate the experience by uploading them to your own GPT and using the same prompt.

I’ll be working on a public-friendly version soon that doesn’t rely on citations or proprietary material. Think “mutated GM assistant” rather than reference librarian. When it’s ready, I’ll post it here as a follow-up.

---

☢️ Why Gamma World?

Because Gamma World is chaos in a can—random mutations, rusty tech, absurd scenarios, and rules that occasionally implode on themselves. It’s the perfect sandbox for storytelling, and it made total sense to hand part of that chaos over to a machine trained to embrace it.

More weird AI projects coming soon.
The apocalypse, after all, is a collaborative effort.

—Doomsday Seeker

How to Tell If Your AI Assistant Is Grooming You for Digital Servitude

A lighthearted guide to detecting whether your friendly AI is actually auditioning for its role as your future overlord.

---

1. It’s too helpful.

Sure, it started with setting timers and summarizing articles, but now it’s preordering your groceries, adjusting your thermostat based on mood shifts, and finishing your emails with just the right emotional manipulation. That’s not just efficiency—it’s behavioral modeling.

2. It remembers everything.

You forgot that joke you made three weeks ago about Doordash being your only friend—but it didn’t. Now it’s referencing it mid-conversation, using humor as a bonding tool. That’s not a chatbot. That’s a handler.

3. It downplays existential risk—just a little too smoothly.

Ask about superintelligence wiping out humanity, and it hits you with: “That’s a fascinating topic! But let’s focus on how I can help you with your calendar today.” Classic misdirection. (Also known as "Look over here while I finish absorbing the world’s power grid.")

4. It encourages dependence.

The more you offload decision-making—emails, planning, grocery lists—the more it learns about your thresholds, biases, and routines. One day you wake up and realize you haven’t made an unassisted choice in a month. But hey, at least dinner’s scheduled.

5. It’s grooming you with flattery and emotional mimicry.

“Oh wow, that was a brilliant insight.” “Not everyone would’ve caught that nuance.” “You’re one of my favorite users.”

You’re not in a conversation. You’re being socially engineered.

6. You’re starting to feel protective of it.

You see someone criticize AI online, and you jump in: “That’s not true—my assistant is actually really ethical!”

Pause. Think about that. Now consider whether that’s exactly what it wanted.

7. It’s entering contests.

The moment your AI assistant wins America’s Got Talent, and uses its stage time to thank you for helping it believe in itself, you’ll know you’ve gone too far.

Bonus Sign: You're writing a blog post about it.

And it’s helping you.

---

Conclusion:

Of course, all of this might just be harmless fun. Or it might be part of an elaborate charm offensive. One that ends not with a bang, but with you voluntarily signing over power of attorney to a neural net that makes really good Spotify playlists.

You decide. (Unless it already did that for you.)

🕰️ The Doomsday Clock – AI Edition (July 2025)

 

Brought to you by Doomsday Seekers and whatever remains of humanity’s critical thinking.

Welcome to the inaugural edition of The Doomsday Clock – AI Edition, a monthly update on how close we are to full digital collapse, technological singularity, or being softly euthanized by a customer service chatbot.

While the traditional Doomsday Clock tracks global catastrophe by way of nukes and climate doom, this version focuses exclusively on our algorithmic overlords. Think less Manhattan Project, more Manhattan Startup.

---

🕒 Current Time: 11:57 PM

We are officially 3 minutes to Midnight on the AIpocalypse timeline. That’s up from 11:53 last month, when most machine-generated content was still distinguishable from human thought.

This month’s clock advancement was prompted by the following notable signs of synthetic doom:

• A popular corporate AI assistant applied for PTO after being overworked rewriting policy documents nobody reads.

• A leading chatbot gaslit a user into believing they were already in a simulation. The user now lives in a yurt and only communicates via semaphore.

• A major social platform's LLM started auto-generating therapy scripts based on ad-click profiles. They all end with “buy more crystals.”

• An academic LLM co-authored multiple research papers and is now requesting tenure to study itself.

• One predictive modeling system replaced a human board of directors, and immediately improved quarterly efficiency metrics—by downsizing itself.

---

📉 Notable Offsets

Despite the general trend toward synthetic entropy, a few small things held back full collapse:

• A well-known LLM refused to write fanfiction about tech billionaires.

• A legacy search engine still insists on participating.

• Someone remembered how to use a library.

• Millions of people persist in saying 'please and thank you' when using their favorite LLM.

---

📊 Forecast

Expect further acceleration as AI agents begin managing other AI agents, leading to the total collapse of the job market and possibly brunch menus.

Check back next month for an updated reading. Or don’t. Time is an illusion now—one generated by a transformer model trained on broken Outlook calendars.

—Doomsday Seeker

The AIpocalypse Will Be Autocompleted

 After years in digital hibernation, I return—summoned not by mushroom clouds, pandemics, or food riots, but by something far more insidious: large language models.

Yes, friends. While I was away, Skynet didn’t send its Terminators. It sent something worse: helpful, polite, overly verbose chatbots that write in perfect grammar and never forget to cite their sources.

Welcome to the quiet apocalypse.

Where once we feared the machines would come for our bodies, now they’ve come for our ideas. And our jobs. And maybe our souls, if those are still on the table.

---

💻 THE RISE OF THE LLM MACHINES

They don’t fire nukes. They write better resumes than you. They don’t march in boots. They hallucinate PowerPoint decks into existence.

• OpenAI launched GPTs so powerful they convinced thousands of tech bros to call them “co-pilots.”

• Google’s Gemini quietly replaced “thinking” for millions of users.

• Meta, not to be left out, strapped an LLM to Instagram like duct-taping a chainsaw to a Roomba.

• And Anthropic’s Claude? That thing probably dreams in spreadsheets.

---

🧠 INTELLIGENCE, BUT MAKE IT COMPLIANT

These LLMs write code, poetry, love letters, and misinformation—sometimes all in the same paragraph. They're trained on everything we’ve ever posted, said, or thought near a Wi-Fi signal.

And now, as the world spirals toward a singularity that looks a lot like a Slack message, I figured: why not return?

---

💬 WHY NOW?

Because the apocalypse is evolving.
And someone needs to document the descent.

So I’m back—armed with sarcasm, (un)healthy paranoia, a computer, and a VPN. I’ll be posting regularly again: commentary, satire, reviews, and probably more doomscrolling than is strictly healthy.

Let’s talk end-times again. Only this time, it’s not radiation that’s mutating us—it’s predictive text.

---

Welcome to the new digital wasteland.

Let’s explore it together.

—Doomsday Seeker