XNews

Evaluation: FactoryAI’s structured summarization preserves agent context better than OpenAI/Anthropic

Evaluation: FactoryAI’s structured summarization preserves agent context better than OpenAI/Anthropic

The post reports that FactoryAI’s anchored iterative summarization retains intent, file changes, decisions and next steps better than OpenAI’s high-compression /responses/compact and Anthropic’s full-regeneration summaries. It introduces a probe-based evaluation (Recall, Artifact, Continuation, Decision), uses GPT-5.2 as scorer across six dimensions, and shows Factory leads in accuracy and context-awareness on 36,611 real messages.

Source posts