Explore how neurological experiments show the subconscious can guide us toward better decisions before we can rationally explain them. These findings suggest that physiological responses often precede and inform our conscious understanding during complex tasks.
I learned useful CLI tools, why evals often land in the middle, how file-based context and modern HTML attributes improve products, plus notes from neuroscience and other books.
I trace how AI bottlenecks shift as capabilities improve, from writing code and chaining agent actions to searching enterprise knowledge and processing complex documents.
Look at what neuroimaging and fMRI can do to map human thought processes and emotional responses. These 2004-era brain scans provide a window into how specific brain regions correlate with behavior and neurological health.
I explored multi-agent architectures, refined my AI coding workflows using MCP and Cursor, and experimented with GPT 4.1 prompting. I also learned handy uv and jq tricks while investigating application-specific LLM evaluations.
I rejected the standard weighted-average approach for CRM selection in favor of binary filtering. Using strict yes/no criteria for essential features prevents the manipulation inherent in subjective scoring and provides a clear, defensible rationale for every rejection.