<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: the truth</title>
    <description>The latest articles on DEV Community by the truth (@th0truth).</description>
    <link>https://dev.to/th0truth</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3798648%2Ff501bbd5-240e-4cea-8791-0c48e68b8697.jpeg</url>
      <title>DEV Community: the truth</title>
      <link>https://dev.to/th0truth</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/th0truth"/>
    <language>en</language>
    <item>
      <title>Next gen hardware monitor with AI</title>
      <dc:creator>the truth</dc:creator>
      <pubDate>Sat, 28 Feb 2026 18:01:38 +0000</pubDate>
      <link>https://dev.to/th0truth/i-wrote-a-hardware-monitor-in-c-that-talks-directly-to-the-linux-kernel-34k2</link>
      <guid>https://dev.to/th0truth/i-wrote-a-hardware-monitor-in-c-that-talks-directly-to-the-linux-kernel-34k2</guid>
      <description>&lt;p&gt;
  &lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fraw.githubusercontent.com%2Fth0truth%2Fhwmonitor%2Fmaster%2F.github%2Fassets%2Flogo.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fraw.githubusercontent.com%2Fth0truth%2Fhwmonitor%2Fmaster%2F.github%2Fassets%2Flogo.png" alt="hwmonitor"&gt;&lt;/a&gt;
&lt;/p&gt;

&lt;p&gt;Every hardware info tool on Linux does the same thing: spawn &lt;code&gt;lspci&lt;/code&gt;, pray the output format hasn't changed, regex the hell out of it, and call it a day.&lt;/p&gt;

&lt;p&gt;I wanted something that just asks the kernel directly. No middleman. So I wrote hwmonitor in pure C — it reads &lt;code&gt;/sys&lt;/code&gt; and &lt;code&gt;/proc&lt;/code&gt; raw, spits out structured JSON, and runs in under a millisecond.&lt;br&gt;
Then I gave it a brain.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight console"&gt;&lt;code&gt;&lt;span class="go"&gt;hwmonitor --gpu -A "Can I run a 70B model locally on this hardware?"
╭─ AI Hardware Analysis (Groq)
| With 8GB VRAM on your RTX 5060 a 70B model won't fit even with
| INT4 quantization. You'd need ~35GB VRAM. Stick to 7B-13B models,
| or offload layers to your 32GB RAM at the cost of ~4x slower
| inference.
╰─
&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;You ask your hardware a question. It answers. From the terminal.&lt;br&gt;
CPU, GPU, RAM, Storage, Mainboard, Battery — all discovered, all JSON-serializable, all without a single fork().&lt;/p&gt;

&lt;p&gt;~2000 lines of C. One dependency.&lt;/p&gt;

&lt;p&gt;MIT licensed.&lt;/p&gt;

&lt;p&gt;hwmonitor: &lt;a href="https://github.com/th0truth/hwmonitor" rel="noopener noreferrer"&gt;https://github.com/th0truth/hwmonitor&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;What hardware questions would you ask?&lt;/p&gt;

</description>
      <category>ai</category>
      <category>performance</category>
      <category>cli</category>
      <category>discuss</category>
    </item>
  </channel>
</rss>
