IBM Research has published a Hugging Face blog post arguing that ACE can be accomplished with fewer tokens. The title, "Thinking of ACE? We Can Do It with Fewer Tokens," frames the post as a direct response to the assumption that agentic context engineering necessarily consumes large amounts of token capacity.
The post appears to target developers and researchers working on agent-based systems, where input length and inference cost are practical concerns. By focusing on token reduction, IBM Research positions its approach as a way to make ACE more economical without abandoning the underlying technique.
Because the analysis is based on a single source, there are no competing findings to weigh here. The claim is presented as a research statement rather than an independent benchmark or survey, so readers should treat it as one team's proposal until broader evidence appears.