This is the story of Project Cyber-Gothic: an end-to-end journey of fine-tuning and deploying a stylistically specialized Large Language Model using modern, memory-efficient techniques.

From Foundation Models to Personality Engineering

A Large Language Model (LLM) is fundamentally a prediction engine. Given a sequence of tokens, it predicts what comes next. What makes it powerful isn’t consciousness or intent — it’s scale, data, and alignment. But raw capability isn’t enough.


That’s where fine-tuning comes in: taking a GPM and specializing it with a small, high-quality dataset. In this case, the specialization was a distinct Cyber-Gothic aesthetic — dark, futuristic, architectural, and surreal.