For frontier AI, it's neither; the hard part is changing the objective function from building obedient systems to building safe systems. Asimov’s Laws: Safety (wt 0.9), obedience (wt 0.7),…

Why isn't Elon Musk competing with Qwen 3.8:27b instead of Claude or GPT? If he wants Optimus to work, you need to be less reliant on uplinks to a massive data center, so smaller, more capable…

Added functionality to CAIOS. PDF OCR (local file or web). Log history search tool, since all conversations are stored in the KB regardless of which LLM it uses. Enabled image vision support for Qwen…

1 Not in the same session - subsequent reinforcement. You should have noticed that when AI doubles down on being wrong. Gödel, Turing (halting), and others are pointing out the binary loop - no…

I keep saying capability is f(logic-depth/archetecture), not f(params/compute) - no one listens. 🤷‍♂️

Thanks, and np. Most of it works universally if structured into the reasoning (VBR/ethics), but some of it's dependent on the post-binary logic for stable undecidable states. If T=0.5/F=0.5, binary…

AI doesn't handle uncertainty; they collapse on a binary axiom because the output is n-ary despite probabilistic reasoning. That's the cause of hallucinations and confabulation, which is what this…

Yep, most of those: COBOL, other IBM UNIX, and some I can't even remember. People rarely know what they need, but if they can explain their objective, you can reverse-engineer from there. That's the…

Simplified: standard encryption dual key pair: 123(This lock)=ABC(Unlocked by this key). It's like a logical qubit that rotates both key pairs (12 combinations of 123)=(12 combinations of ABC). The…

Sorry, I was gone for a few days moving. It's open-source: You can run it locally; no API needed, but you can use API keys if you dont have the GPU (SS on RTX 3090). Simple setup: Full readme: