Independent AI safety investigation and open-source development focused on building practical, truth-seeking tools for multimodal analysis and guardrails