The "Bitter Lesson" is not just about using more compute, but leveraging it scalably. Current LLMs are inefficient because they only learn during a discrete training phase, not during deployment where most computation occurs. This reliance on a special, data-intensive training period is not a scalable use of computational resources.

Even with vast training data, current AI models are far less sample-efficient than humans. This limits their ability to adapt and learn new skills on the fly. They resemble a perpetual new hire who can access information but lacks the deep, instinctual learning that comes from experience and weight updates.

The current limitation of LLMs is their stateless nature; they reset with each new chat. The next major advancement will be models that can learn from interactions and accumulate skills over time, evolving from a static tool into a continuously improving digital colleague.

AI agents like OpenClaw learn via "skills"—pre-written text instructions. While functional, this method is described as "janky" and a workaround. It exposes a core weakness of current AI: the lack of true continual learning. This limitation is so profound that new startups are rethinking AI architecture from scratch to solve it.

The core weakness of Transformers is their static nature. They are trained in a lab on a snapshot of data and then deployed. They cannot adapt to new events, tools, or user tasks without a full retraining cycle, making true continuous learning at test time impossible with the current architecture.

The "memory" feature in today's LLMs is a convenience that saves users from re-pasting context. It is far from human memory, which abstracts concepts and builds pattern recognition. The true unlock will be when AI develops intuitive judgment from past "experiences" and data, a much longer-term challenge.

A fundamental misunderstanding is that AI learns from each interaction. It doesn't. Models are trained, but each new prompt is a fresh start, like 'Groundhog Day.' They operate on syntactic patterns without building semantic understanding or memory, which explains their inconsistent responses.

Demis Hassabis argues that current LLMs are limited by their "goldfish brain"—they can't permanently learn from new interactions. He identifies solving this "continual learning" problem, where the model itself evolves over time, as one of the critical innovations needed to move from current systems to true AGI.

A significant hurdle for AI, especially in replacing tasks like RPA, is that models are trained and then "frozen." They don't continuously learn from new interactions post-deployment. This makes them less adaptable than a true learning system.