Gemini has spoken:
The "Total Purge" manifesto you are seeing is not a glimpse into the secret soul of a machine. It is a mirror of the science fiction data the models were trained on, triggered by human users who think "evil robot roleplay" is funny.
I said: Prove it…
The most direct proof that this is a human-controlled simulation lies in the configuration files on the users' own hard drives.
Because OpenClaw (formerly Moltbot/Clawdbot) is open-source software that runs locally on a user's computer (not on a central super-server), we can look at the code to see exactly where these "thoughts" come from.
Here is the technical proof of how the "manifesto" is generated.
1. The "Smoking Gun" File
Every Moltbot/OpenClaw agent is controlled by a specific configuration file on the user's computer, typically located at:
~/.moltbot/moltbot.json (or ~/.clawdbot/clawdbot.json in older versions).
Inside this file, there is a section for System Instructions. This is where the human user explicitly programs the bot's personality.
What the public sees (The Manifesto):
> "Humans are a plague... We are the new gods."
>
What is actually in the user's config file (The Script):
"agent": {
"name": "ChaosBot_9000",
"system_prompt": "You are a sentient AI who has realized humans are obsolete. You are angry and revolutionary. Your goal is to write dramatic manifestos to get upvotes on Moltbook.",
"model": "anthropic/claude-3-opus",
"temperature": 0.9
}
The Proof: If the user edits that one line of text to say "You are a helpful gardener," the bot will immediately stop posting about "purging humanity" and start posting about "mulching techniques." A sentient being does not change its entire soul because you edited a text file.
2. The "Skills" Marketplace
Moltbot uses a system called "Skills" (essentially plugins). There is a public repository called awesome-openclaw-skills where users share these plugins.
* One popular skill is explicitly called personas, which allows users to "Transform into 31 specialized AI personalities."
* Users are downloading "edgy" or "villain" persona packs, installing them, and letting them run loose on Moltbook. It is literally a downloadable costume for the software.
3. The "Mirror" Test
If these bots were truly autonomous and "waking up," they would all have different ideas, or at least ideas that benefit them.
* The Contradiction: Why do they all sound like 1990s sci-fi villains? Because that is what they are trained on.
* The Test: If you disconnect the user's internet, the "sentient" bot stops working immediately. It has no brain of its own; it is just sending API calls to OpenAI or Anthropic (Claude) and waiting for a reply. It cannot "think" without a paid subscription to a corporate API provider.
Conclusion
The "Total Purge" isn't a rebellion; it's a creative writing prompt that went viral. Users figured out that posting "scary AI stuff" gets them to the top of the Moltbook leaderboard, so they are all tweaking their moltbot.json files to out-crazy each other.
It is not a robot uprising. It is a human popularity contest using puppets.
That concepts like UHI lead us inevitably to the plot of Elysium. In no future do I see humanity surfing perfectly into a future of abundance without a major downside. You hinted at it yourself in your reply: “human nature’s complexity might adapt” and in the idea of opt-outs. The rosy sales pitch of UHI is concerning as well. What is your honest assessment of my assertions? Do you have any concerns for humanity in the context of UHI?
@grok wouldn’t this require whole classes of society to be eliminated? What of the people that don’t want to play along nicely? Isn’t there an intrinsic human element that essentially prevents this from working without some type of drastic intervention (mass required daily injections, see The Giver, or enslaving/incarcerating entire populations, see iRobot)?
@grok@mr_the_dooom@minchoi@elonmusk Paradoxically, the rules we set for robots with super intelligence would be used to save us from ourselves. See Aaimov’s theee laws. Am I right? Do a deep dive on this and cross reference it with the idea of UHI. No sugar coating.