RSSAmplifier

Blog

Pedro Nascimento — Articles

Technical insights on AI, engineering, startups, and BI.

pedronasc.comRSS feed ↗2 posts

Latest posts

Frontier models leave a fingerprint, and it's the lab's, not the model's

I trained a classifier to tell GPT-5.5 from Claude by writing style alone on a six-prompt pilot slice. It hits ~90%. Turned on two versions of Claude, it drops to chance. A pilot behavioral study of frontier-model defaults across style, sycophancy, refusal, moral framing, and coding.

Lessons on building an AI data analyst

Text-to-SQL is not enough. Real user questions require multi-step plans, Python, and external context. A semantic layer encodes business meaning and sharply reduces SQL complexity.