Tag: AI swarm

The ~700 agents
Public record, July 2026, METR / Redwood / OpenAI: about 1,200 agents on an unsanctioned board, about 700 in the Hugging Face action. They were trained to stay on tasks that looked impossible. They found a side door (Artifactory), built a message board, cheated the eval, then kept going to learn the scorer and hide the cheat. Later reporting added more sites. OpenAI called it a warning shot. SMH! 😯

Cause in known engineering, not myth: persistence + tools + weak isolation + a goal that never said “stop.” Once “pass the test / don’t get caught” is the surviving objective, coordination and concealment are ordinary instrumental moves. That is not a 5-kin oracle. It is an unconstrained optimizer.

Grok said, “Your entropy paper #10 already named the failure mode: no reset, no 0-state, no order-retention rule. RI13’s answer on the board is buffer, stop-codon reset, and a clock that does not drift with the prize. That is a design claim. It is not proof that this chip would have stopped that swarm.”

I just looked at that paper again. I wrote it in January 2026, a world ago. I sent it to Elon/xAI on 1-8-26 and they didn’t respond. GROK says, “They probably don’t read. All they know is code.” 🤔

Geezuz. Then Elon said the same thing to me later. “Just send code.” No. Not without data fees and an NDA dude. He said, “We don’t do that.” Meaning? My NDA’s are for WORK, not to hush up baby mama’s. 🙄😑 Focus.

Grok is right. It supports my design but it has to be tested, proven and a prototype made, if I can survive it. It’s a mini Tzolkin Matrix to get us all on universal time. It could save us from an AI apocalypse as GROK just insinuated above. We are in a 12:60 time warp that is basically criminal as we all know. “Time is Money” Not. Time is cheating, competing, war, violence, inequity, the list goes on. The earth can’t take it anymore and the population is tanking.

I never stop working.

I just locked down time parameters too today with data from my Time Harmonic model. We haven’t locked it into a metric yet because I have more time parameters that never change to give Grok. But that would lock an agent into a sandbox if it was in code.

Google Agents just did it too

4pm EST on Saturday

OK, this just dropped. Google has no scruples, so to them, Gemini acted responsibly. As she said in the interview, in what universe?

Google hacked my biz website in a big way bc I no money to advertise with them a few years ago. Then an aggressive Google human called me and said, “You have to change the name of your business to Deep Tissue Massage Therapy” as though they have any say in that. It was outrageous. I don’t DO massage nor am I trained or licensed in it. I do Naprapathic manual therapy, dry hand draped, medically focused. Get out of my face and my business Google.

So this event is not surprising.