---
title: "Publication on Habr: indirect control tool"
description: "A note about how LLMs work as tools with indirect control: instead of direct commands, you create conditions. In a poll, 90% agreed with the metaphor 'people are tools for other people'."
date: 2025-08-17
display_date: "August 17, 2025"
image: "https://habrastorage.org/r/w1560/getpro/habr/upload_files/499/c70/9e9/499c709e91b018135b80de845c1e038d.png"
image_alt: "Indirect control tool"
---

# Publication on Habr: indirect control tool

<zoom-img
            src="https://habrastorage.org/r/w1560/getpro/habr/upload_files/499/c70/9e9/499c709e91b018135b80de845c1e038d.png"
            alt="Indirect control tool"
            width="100px"
            style="float: left; margin: 15px 15px 0 0;"
    ></zoom-img>

The note **["Indirect control tool"](https://habr.com/ru/articles/937908/)** has been published. It says that large language models are best understood as tools. But these are unusual tools: you can't operate them directly. Instead of buttons and levers, it's the context that brings them to the desired result.

For **ADSM** this idea matters: if a site is managed by an agent, the interaction should be built not on direct commands but on creating conditions. In this, agents are surprisingly similar to people—the "indirectness" makes the comparison especially vivid.

The note ended with a small poll: *"How correct is the statement 'people are tools for other people'?"*

        90% answered "absolutely true," 10% — "completely false." The result showed that the metaphor works and resonates with many.