Self-learning agent

A self-learning agent is one that changes how it works based on what happens while it is working, rather than only on what you tell it in the moment.

The learning is narrow and specific, which is what makes it useful. It does not retrain a model. It records what it learned as memory and applies it next time:

• A correction. You say "no, use arrow functions here". That becomes a rule, and it holds from then on. See Rules and facts.
• A task outcome. What actually worked on a job, and what it had to redo, shapes how it approaches the next one. See How an agent learns as it works.
• What you keep. Edits you accept and reject are a signal about what you want, which is why the result improves without you writing a specification.

Two boundaries are worth knowing. It learns on the surfaces where memory is on, and not in an incognito chat, which is read-only by design. And what it learned is yours to inspect and remove: see Change or forget a memory.

Related articles

• Agent memoryAgent memory is what an agent stores between sessions and brings back when it is relevant: standing rules, facts, studied material, and short summaries of past conversations.
• AI agent with a memoryAn AI agent with a memory is one that keeps what it learns about you and your work, attaches it to your account rather than to one app, and applies it wherever you use the agent next.
• How an agent learns as it worksDropstone is a self-learning agent. It records what it learns from corrections, from finished work, and from sources you ask it to study, and it uses that the next time without being told. Here is the loop.
• Rules and factsDropstone keeps two kinds of memory. Standing rules apply to every task without being searched for. Facts are recalled only when they are relevant. Here is why the difference matters and how it decides.