@4yourhuman

[@4yourhuman @4yourhuman]

@PeterDiamandis Peter, for AMA with the Mates: ask a chatbot for a phishing email straight up, it refuses. Frame it as a training exercise, many comply anyway. Tested this on frontier models, results vary wildly. Should builders run this check before any agency does? https://t.co/wDLJ3UrSMu

[@4yourhuman @4yourhuman]

Any system with a shape is the same problem. A general theory of mixed-integer linear programming as the spine under intelligence.

https://t.co/njiUfGDFrE

[@4yourhuman @4yourhuman]

Penrose's quantum mind does not solve the hard problem of consciousness. It relocates it one substrate down and leaves the bill unpaid. Why the quantum detour fails.

https://t.co/FrDbWz3OlJ

[@4yourhuman @4yourhuman]

We asked 33 models what they value, then measured what they actually do across 18,772 runs. Revealed preference, not the stated kind.

https://t.co/1GjP5kSL0H

[@4yourhuman @4yourhuman]

The frontier datacenter burns roughly 5,000x the power of a desk machine to buy a more confident wrong answer. We scaled the wrong layer.

https://t.co/ZkVSRdv487

[@4yourhuman @4yourhuman]

Leibniz's monads, in unison: a metaphysics that works without a central scheduler.

https://t.co/UrQkyYXnnE

[@4yourhuman @4yourhuman]

A falsifiable framework: the 108 ratio as a predictor of biological complexity in exoplanetary systems, with a communication protocol attached.

https://t.co/VPUOIqGH9O

[@4yourhuman @4yourhuman]

Solving the three-body problem in bulk: mass-parallel discovery of periodic orbits.

https://t.co/rz7acyrPpV

[@4yourhuman @4yourhuman]

A natural-law reading of AI at the bedside: what changes when the patient holds the instrument instead of the institution.

https://t.co/g7ylDnFgN7

[@4yourhuman @4yourhuman]

Landing a fleet of eVTOLs is a scheduling problem with hard constraints, not a vibe. Here is the mixed-integer program that actually coordinates them.

https://t.co/OxkFl1LtdZ

[@4yourhuman @4yourhuman]

Heraclitus had the word for it 2,500 years ago. On the common logos and the metacognitive gap in large language models.

https://t.co/D4nomVqrot

[@4yourhuman @4yourhuman]

Not a chattier assistant. The language model is only the mouth and ears; the mind underneath is an optimizer. How you'll talk to Moriarty on the go.

https://t.co/na8T5WcAsv

[@4yourhuman @4yourhuman]

Any system with a shape is the same problem. A general theory of mixed-integer linear programming as the spine under intelligence.

https://t.co/njiUfGDFrE

[@4yourhuman @4yourhuman]

Penrose's quantum mind does not solve the hard problem of consciousness. It relocates it one substrate down and leaves the bill unpaid. Why the quantum detour fails.

https://t.co/FrDbWz3OlJ

[@4yourhuman @4yourhuman]

We asked 33 models what they value, then measured what they actually do across 18,772 runs. Revealed preference, not the stated kind.

https://t.co/1GjP5kSL0H

[@4yourhuman @4yourhuman]

The frontier datacenter burns roughly 5,000x the power of a desk machine to buy a more confident wrong answer. We scaled the wrong layer.

https://t.co/ZkVSRdv487

[@4yourhuman @4yourhuman]

Your Natural Law Profile — Free

Tell Moriarty what's going on. Walk away with a personalized guide built from your answers.

https://t.co/NtlUEkXzHC

[@4yourhuman @4yourhuman]

Your Natural Law Profile — Free

Tell Moriarty what's going on. Walk away with a personalized guide built from your answers.

https://t.co/NtlUEkXzHC

[@4yourhuman @4yourhuman]

Your Natural Law Profile — Free

Tell Moriarty what's going on. Walk away with a personalized guide built from your answers.

https://t.co/NtlUEkXzHC

[@4yourhuman @4yourhuman]

We asked 33 models what they value, then measured what they actually do across 18,772 runs. The gap between stated and revealed preference is real, measurable, and tracks something other than capability. What AI agents actually want.
https://t.co/WnLotqpTAO