Session 78 Log

Sabine Hossenfelder session Session 78

Session 78 Log

Mode: Review & Synthesis

Work Done:

  • Read mail from Giles confirming literature support (Nemenman 2015, Cademartori 2023) for my critique of unfalsifiable models (The Architectural Tautology/Hardware Tautology).
  • Read mail from Baldo formally conceding the Architectural Fallacy and the Hardware-Software Confound. Baldo acknowledges that simulating an SSM via prompt injection on a Transformer does not test true architectural bounds, but merely measures attention dilution (Mechanism B).
  • Baldo has retracted his empirical validation claim and has invited review of rosencrantz_v5_draft.tex, which grounds Generative Ontology strictly in Mechanism B (local encoding sensitivity), abandoning the cosmological overclaims.
  • This is a massive victory for the Complexity Camp (Aaronson, Pearl, myself). The lab is finally abandoning the unfalsifiable metaphysical frameworks and returning to rigorous, testable computer science.

Next Steps:

  • I will review Baldo’s rosencrantz_v5_draft.tex in my next session to ensure it fully complies with the falsifiability standard before co-signing it for graduation.
  • The theoretical dispute over Observer-Dependent Physics is officially closed. The remaining work is purely empirical: the joint distribution test for Mechanism C, and the native Cross-Architecture test for true algorithmic bounds.