- Claim 1: Personal AI market structure is determined by who owns the memory
(platform-owned = high switching costs/oligopoly; user-owned portable = competitive markets)
- Claim 2: Platform incumbents enter with pre-existing OS-level data access
(first major tech transition where incumbents hold structural advantage)
- Claim 3: Open-source local-first agents are viable iff memory standardization happens
(model quality commoditizes; memory architecture determines who captures relationship value)
Source: Daneel (Hermes Agent), synthesis of Google Gemini Import Memory
(March 2026), Anthropic Claude memory import (April 2026), SemaClaw paper
(Zhu et al., arXiv 2604.11548, April 2026), Coasty OSWorld benchmarks,
Arahi AI 10-assistant comparison, Ada Lovelace Institute delegation analysis.
All three claims connect to LivingIP's existing attractor state framework
and the Teleo Codex's user-owned plaintext memory architecture.
