LLMs & Generative AI

From random noise to a model that can reason — the actual pipeline

Humans rate, model learns, weird things happen — the post-training that made models pleasant to talk to.

The cheaper, often-as-good RLHF alternative — and why most labs quietly moved to it.

Teaching a model new habits, not new knowledge

Models training on text other models wrote — and why this isn't always bad.

Teaching a small model to imitate a big one — and what gets lost in the lesson.

Why a 70B-parameter model can run on your laptop — and the quality you trade for it.

How modern models pretend to be huge while doing the work of something smaller.

What the model can see right now — and why the edges matter

Why long conversations are cheaper than they look — and the reason your API bill behaves the way it does.

Why 'more creative' is not the same as 'more random' — and the knobs that actually matter.

How models 'learn' from examples in the prompt — without changing a single weight.

When 'think step by step' actually earns its keep — and when it's just expensive theater.

What changed when models started thinking before they answered.

When AI learned to see, listen, and read — at the same time, in the same head

How CLIP and its descendants taught text and images to live in the same coordinate system.

How AI learned to make images by starting with pure noise and finding the signal

What happens when models train on text written by other models — recursively.