Reinforcement learning achieves superhuman results not by inventing alien concepts, but by surfacing and combining rare behaviors that are already possible within a model's vast pre-trained distribution. The goal of pre-training is to make this search for novel solutions more efficient and less random.

RL fine-tuning is less likely to cause catastrophic forgetting than SFT because it works within the model's existing pre-trained pathways, or "grooves." SFT, by contrast, makes much larger weight updates that can aggressively overwrite and destroy latent knowledge.

The argument that LLMs are just "stochastic parrots" is outdated. Current frontier models are trained via Reinforcement Learning, where the signal is not "did you predict the right token?" but "did you get the right answer?" This is based on complex, often qualitative criteria, pushing models beyond simple statistical correlation.

While RL is compute-intensive for the amount of signal it extracts, this is its core economic advantage. It allows labs to trade cheap, abundant compute for expensive, scarce human expertise. RL effectively amplifies the value of small, high-quality human-generated datasets, which is crucial when expertise is the bottleneck.

Reinforcement Learning (RL) is ideal for fine-tuning AI agents because it allows them to self-learn and align their behavior by exploring vast, non-deterministic environments. This is a more scalable approach than supervised fine-tuning, which would require an impossibly large, pre-curated dataset to cover all potential pathways.

Pre-trained models ingest knowledge from both experts and novices. A key function of RL, especially in its early stages, is to "sharpen the distribution" by tuning the model to consistently adopt the persona of an expert who provides correct answers, not a student who is still learning.

Basic supervised fine-tuning (SFT) only adjusts a model's style. The real unlock for enterprises is reinforcement fine-tuning (RFT), which leverages proprietary datasets to create state-of-the-art models for specific, high-value tasks, moving beyond mere 'tone improvements.'

Goodfire's research operates on the premise that post-training processes like RL don't teach models fundamentally new capabilities. Instead, they primarily make low-likelihood events and behaviors already present from pre-training more probable, essentially shaping the model's existing knowledge.

The transition from supervised learning (copying internet text) to reinforcement learning (rewarding a model for achieving a goal) marks a fundamental breakthrough. This method, used in Anthropic's Opus 3 model, allows AI to develop novel problem-solving capabilities beyond simple data emulation.