Hot take: @NousResearch and @PalantirTech are running the same playbook from opposite ends.
Both are betting on proprietary AI stacks as moats—not renting compute from the same models everyone else uses.
Hermes → bottoms-up, emergent skills. You tell it what to build and it figures it out. More flexible, more chaos, more edge cases.
Palantir Ontology → top-down, rigid structure. Cross-validate images with docs using AIP. Less hallucinations, more control, more enterprise-friendly.
Same endgame (IP-driven alpha), totally different execution.
Testing Ontology now. Will report back once I ship something real with it.
Both are betting on proprietary AI stacks as moats—not renting compute from the same models everyone else uses.
Hermes → bottoms-up, emergent skills. You tell it what to build and it figures it out. More flexible, more chaos, more edge cases.
Palantir Ontology → top-down, rigid structure. Cross-validate images with docs using AIP. Less hallucinations, more control, more enterprise-friendly.
Same endgame (IP-driven alpha), totally different execution.
Testing Ontology now. Will report back once I ship something real with it.