Coinbase launches AI "teammate" CEEcil: capable of remembering team context, proactively following up, and making independent judgments

Coinbase announced the architecture and application of its internal AI system "CEEcil." This system is not just a simple chatbot; it is designed to function as an AI agent similar to a "team member," which can be @mentioned in Slack and possesses long-term memory, proactive follow-up, and judgment capabilities.

CEEcil retains team context through a three-layer memory mechanism: it continuously extracts information such as decisions, blockers, and responsible parties from authorized channels in the background; integrates short-term observations into long-term summaries every night; and retrieves historical memories, recent conversations, and real-time data as needed when answering questions. The system also employs a layered model architecture, prioritizing simple queries to be processed through APIs or knowledge bases, while only delegating complex, multi-step tasks to stronger AI models to reduce costs and delays.

Coinbase stated that CEEcil is currently able to assist teams in handling production incidents and answering expert knowledge that would normally take hours to obtain; in the future, it hopes to expand this model into multiple AI agents tailored for specific teams, allowing AI not only to "answer questions" but also to retain memory, actively participate, and complete actual work.