---
title: "Comment by Kartik N V J K on \"Block-addressed editing: a 10x output-token cost reduction for document LLM agents\" | Hashnode"
description: "Going from 8128 to 775 output tokens by having the model emit stable block IDs instead of re-serializing the whole document is a genuinely clever reframing of t"
url: "https://hashnode.com/posts/block-addressed-editing-a-10x-output-token-cost-reduction-for-document-llm-agents/6a6147a108bd5f72635e388f/comment/6a622343803339683c18043b"
---

[Discussion](https://hashnode.com/posts/block-addressed-editing-a-10x-output-token-cost-reduction-for-document-llm-agents/6a6147a108bd5f72635e388f)

Comment [Block-addressed editing: a 10x output-token cost reduction for document LLM agents](https://hashnode.com/posts/block-addressed-editing-a-10x-output-token-cost-reduction-for-document-llm-agents/6a6147a108bd5f72635e388f)

[Kartik N V J K](https://hashnode.com/@kartik-nvj)

[Jul 23](https://hashnode.com/posts/block-addressed-editing-a-10x-output-token-cost-reduction-for-document-llm-agents/6a6147a108bd5f72635e388f/comment/6a622343803339683c18043b)

Going from 8128 to 775 output tokens by having the model emit stable block IDs instead of re-serializing the whole document is a genuinely clever reframing of the cost problem. Borrowing the dirty-path tracking idea from Slate's WeakMap is the part I didn't expect. How stable do the block IDs stay when the document is restructured heavily, like sections being merged or split?

[Mihai Marinescu](https://hashnode.com/@mmarinescu)

[Jul 24](https://hashnode.com/posts/block-addressed-editing-a-10x-output-token-cost-reduction-for-document-llm-agents/6a6147a108bd5f72635e388f/comment/6a622343803339683c18043b#reply-6a6310073ea1672b12e8fbc9)

Nice!
