---
title: "Discussion on \"A Tool Result Is Evidence, Not Ground Truth\" | Hashnode"
description: "Discussion on \"A Tool Result Is Evidence, Not Ground Truth\". MemToC shows how easily small instruction models replace a correct answer with a plausible wrong tool payload."
url: "https://hashnode.com/posts/a-tool-result-is-evidence-not-ground-truth/6a94bd1b37a23da4b7a5a41e"
---

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

Discussion

# [A Tool Result Is Evidence, Not Ground Truth](https://letinix.hashnode.dev/a-tool-result-is-evidence-not-ground-truth?utm_source=hashnode&utm_medium=feed)

MemToC shows how easily small instruction models replace a correct answer with a plausible wrong tool payload.

Aug 30 · 7 min read · letinix.hashnode.dev  
Updated Oct 17  
Read the article

-   [#ai](https://hashnode.com/tag/ai)
-   [#software-engineering](https://hashnode.com/tag/software-engineering)
-   [#agentic-ai](https://hashnode.com/tag/agentic-ai)
-   [#llm-evaluation](https://hashnode.com/tag/llm-evaluation)
-   [#tool-use](https://hashnode.com/tag/tool-use)
-   [#reliability](https://hashnode.com/tag/reliability)
-   [#applied-research](https://hashnode.com/tag/applied-research)

## Written by

[Muhammad Azlaan Zubair](https://hashnode.com/@mdazlaanzubair)

@mdazlaanzubair · Karachi, Pakistan

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

Software Engineer | Applied AI Researcher | AI Workflows & Automation

Engineer. Researcher. Builder. I'm a Software Engineer and Applied AI Researcher interested in building reliable software, experimenting with AI systems, and understanding the engineering decisions behind modern technology. I write about software engineering, AI, developer tools, automation, and technical research. My work combines practical engineering experience with curiosity-driven experimentation and evidence-based analysis. Publication: https://letinix.com Portfolio: https://mdazlaanzubair.com

## Responses 1

Markdown supported · ⌘↵ to post

[Adam Lewis](https://hashnode.com/@nark3d)

[Sep 2](https://hashnode.com/posts/a-tool-result-is-evidence-not-ground-truth/6a94bd1b37a23da4b7a5a41e/comment/6a97f18ba237dafb1fa12440)

The valid-JSON-but-wrong case is the one that worries me with agents, because everything downstream treats parsed as true. I keep a verification step in the workflow itself, so the agent has to run the check or quote the source before a result gets used, whichever tool produced it. The MemToC numbers are better evidence for that habit than I had before.
