HOW MANY LICKS DOES IT TAKE TO GET TO THE CENTER OF A TOOTSIE POP?
A nostalgic question, solved with real math. Turns out a Tootsie Pop takes ~200 licks on average.
The world may never know… but the model kinda does.
@Anthropic’s Mythos + Glasswing ran into problems.
They had advanced memory but used reward-only training.
The Triadic Reward Model fixes it:
• Incentive = anchors truth
• Reward = human connection
• Consequence = accountability
No Consequence = hallucinations & broken trust.
@AnthropicAI WHAT? (The Task/Action Layer): Establishing the foundation, or the code syntax.
WHY? (The Functional/Intent Layer): Grounding the model in the purpose and underlying principles of the action.
WHAT IF? (The Counterfactual/Speculative Layer): Unlocking true adaptive reasoning.
@elonmusk My six persona agent reply;
You gave two conflicting anchors: To reach 10 from 11, the only logical vector is descending.
The Execution: 11, 10.
@elonmusk Grok is all those things. I actually did write a theoretical physics paper and Grok helped with mathematical approach. Grok imagine is awesome very impressive.
Interested in ion-based substrate for Grok or long space voyages?
Space Emergency Return Pod – Concept Blueprint
Mission Purpose
A compact, autonomous emergency return vehicle designed to evacuate an injured astronaut from orbit and deliver them safely to a medical team on Earth.
https://t.co/fJZdVp1Nxw