I work inside financial systems and build on top of them. Two sides of the same thing — the industry access, the code, the questions I can't stop asking.
No code written. No fine-tuning. Just questions — and somewhere in those months, a reasoning system started evolving on its own.
Three versions. Three different failures. Memory creates bias, fallback creates blindness, and adversarial architecture creates honesty.
Neural networks don't think. They roll downhill. The same equation behind spin glass physics is identical to the math training GPT-4.
I went looking for a template for life. Found a return mechanism instead. Then realized I'd built the exact same structure into my AI systems — without ever consciously making the connection.
I work inside financial systems and build on what I see from the inside. Computer engineer exploring what happens when finance meets code at the edge.