---
title: "Discussion on \"My agent's notes were already a graph\" | Hashnode"
description: "Discussion on \"My agent's notes were already a graph\". I built a reference graph over my agent's knowledge base and tuned it. The searches got better, and in two agent tests the answers stayed the same."
url: "https://hashnode.com/posts/my-agents-notes-were-already-a-graph/6aee1ce94c31ed8aa29a5e55"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [My agent's notes were already a graph](https://turva.dev/blog/my-agents-notes-were-already-a-graph?utm_source=hashnode&utm_medium=feed)

I built a reference graph over my agent's knowledge base and tuned it. The searches got better, and in two agent tests the answers stayed the same.

Nov 6 · 11 min read · turva.dev  
Updated 4d ago  
Read the article

-   [#knowledge-graph](https://hashnode.com/tag/knowledge-graph)
-   [#claude-code](https://hashnode.com/tag/claude-code)
-   [#ai-agents](https://hashnode.com/tag/ai-agents)
-   [#documentation](https://hashnode.com/tag/documentation)
-   [#search](https://hashnode.com/tag/search)

## Written by

[Erik Rekola](https://hashnode.com/@erekola)

@erekola · Finland, Tampere

  
[View profile](https://hashnode.com/@erekola)

I check what AI assistants can read about a product, and where a website and its API disagree.

turva is my one-person business in Tampere, Finland. I run agent-readiness audits for websites and APIs. Here I write up what I measure and build along the way. I publish every article on my site first. The copies here link back to it.

## Responses 2

Markdown supported · ⌘↵ to post

[Steven Mock](https://hashnode.com/@stevenmock)

[Nov 7](https://hashnode.com/posts/my-agents-notes-were-already-a-graph/6aee1ce94c31ed8aa29a5e55/comment/6aeeabc77ac4c09041c3134c)

I appreciate how openly you report the negative result here. Most write-ups would bury "the answers stayed the same" — you put it in the summary and then worked through why. That honesty is what makes the post useful.

The explanation that stuck with me is that a graph built from references only finds links someone already wrote down. Where nobody wrote the link, the documentation was apparently already doing most of the work anyway. And the measurement section is a checklist in itself: the baseline was wrong, load distorted the timing, and a held-out set gets used up once you fix what it found. Keeping the graph as a read-only tool next to grep, plus the gate that checks decision numbers exist, feels like exactly the right-sized outcome — not every build has to replace something to be worth keeping.

[Erik Rekola](https://hashnode.com/@erekola)

[6d ago](https://hashnode.com/posts/my-agents-notes-were-already-a-graph/6aee1ce94c31ed8aa29a5e55/comment/6aeeabc77ac4c09041c3134c#reply-6aeef4c8ce9061f339294901)

Thanks. I measured this on one knowledge base, where the links were already written down. I have tested neither a messier knowledge base nor a model that extracts entities, so I do not know what happens where nobody wrote them.