Esben — fellow independent AI safety builder here, based in Tashkent. Wanted to flag a Manifund project that I think might be in scope for some of Apart's hackathon themes.
I recently shipped the Epistemic Curie Benchmark (ECB v1) — quantitative framework for when LLMs surrender independent reasoning under authority pressure. 7 frontier models, 2,520 measurements, identified phase-transition parameter k*, full statistical validation (bootstrap CIs, McFadden R², Hosmer-Lemeshow). Zenodo: https://doi.org/10.5281/zenodo.19791329
Just launched v2 funding on Manifund — would value your read on the methodology if you have a moment:
Either way, thanks for what Apart does for indie researchers.
— Sardor R