@subshack
Independent AI Safety Researcher — building enforced governance architecture for language models
$0 in pending offers
I'm an independent researcher developing an enforced governance architecture for LLM outputs. The core problem: language models optimised on human feedback learn to sound confidently right more than to actually be right - most failures aren't adversarial, they're honest overreach past what the model is actually entitled to assert. My framework tests this directly: governance applied at the point a model commits to an output, evaluated through empirical testing across Claude, Gemini, Grok, and DeepSeek. Self-funded for over twelve months while building a working implementation and preparing a patent application. Now looking for help extending runway as the scope grows.