Continuity
Carry useful context forward.
Which decisions and constraints should stay available as work grows beyond a single conversation? We study memory over long histories, where important details can be separated by many interactions.
CIEL: memory for AI agents.
CIEL is KarLex AI’s research project for memory across conversations, code, tools, and documents. We’re developing it for agents whose work continues over time.
Research prototype · In development
The work continues
Context beyond a single conversation.
An agent may return to the same project many times. Requirements, earlier decisions, and the work behind them can still matter long after the interaction that introduced them.
CIEL explores how working memory, recent context, and older details can support that continuity. The focus is keeping relevant information available as the task and its history grow.
Research focus
The questions behind the system.
Building memory means deciding what to keep and how to tell whether it helps. These questions guide our work on CIEL.
Change
Keep up with corrections.
Projects evolve. We investigate how memory should handle revised plans and new information while retaining the context that gives a change meaning.
Evaluation
Measure what memory adds.
Our evaluation work asks whether relevant context remains useful over time, where mistakes occur, and what it costs to carry that context into later work.
Design partnersBring a real
Bring a real
workflow.
We are looking for a team with a long-running coding, research, support, or operations agent to test CIEL on a real workflow.