<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
<channel>
  <title>Impulse — Notes</title>
  <link>https://impulselab.org/blog/</link>
  <atom:link href="https://impulselab.org/blog/feed.xml" rel="self" type="application/rss+xml"/>
  <description>Notes on applied AI from the Impulse engineering studio.</description>
  <language>en</language>
  <lastBuildDate>Fri, 30 Oct 2026 09:00:00 +0000</lastBuildDate>
  <item>
    <title>Why AI pilots stall — and five things that get them to production</title>
    <link>https://impulselab.org/blog/pilot-to-production.html</link>
    <guid isPermaLink="true">https://impulselab.org/blog/pilot-to-production.html</guid>
    <pubDate>Fri, 30 Oct 2026 09:00:00 +0000</pubDate>
    <category>Delivery</category>
    <description>A successful demo is not a successful pilot, and a successful pilot is not a product. Five things that separate AI projects that reach production from the ones that stay on a slide.</description>
  </item>
  <item>
    <title>Voice AI that gets through: lessons from calling farmers</title>
    <link>https://impulselab.org/blog/voice-ai-in-the-field.html</link>
    <guid isPermaLink="true">https://impulselab.org/blog/voice-ai-in-the-field.html</guid>
    <pubDate>Fri, 23 Oct 2026 09:00:00 +0000</pubDate>
    <category>Voice AI</category>
    <description>Some audiences don&#x27;t answer email or fill in forms — they pick up the phone. What we learned building a voice AI robot that holds natural conversations with farmers and hands organizers only the leads that matter.</description>
  </item>
  <item>
    <title>From engineering PDFs to a live catalog with an LLM</title>
    <link>https://impulselab.org/blog/documents-to-catalog.html</link>
    <guid isPermaLink="true">https://impulselab.org/blog/documents-to-catalog.html</guid>
    <pubDate>Fri, 16 Oct 2026 09:00:00 +0000</pubDate>
    <category>LLM · Data</category>
    <description>How we use language models to turn hundreds of pages of technical documentation into a structured product catalog — and why engineers stay in the loop for every description.</description>
  </item>
  <item>
    <title>Computer vision on the cameras you already have</title>
    <link>https://impulselab.org/blog/cv-on-existing-cameras.html</link>
    <guid isPermaLink="true">https://impulselab.org/blog/cv-on-existing-cameras.html</guid>
    <pubDate>Fri, 09 Oct 2026 09:00:00 +0000</pubDate>
    <category>Computer vision</category>
    <description>You rarely need new hardware to put computer vision into a working process. What ordinary CCTV and phone cameras can do, where they fall short, and how to design the alerts people will actually read.</description>
  </item>
  <item>
    <title>Where AI actually pays back: a checklist before you start</title>
    <link>https://impulselab.org/blog/where-ai-pays-back.html</link>
    <guid isPermaLink="true">https://impulselab.org/blog/where-ai-pays-back.html</guid>
    <pubDate>Fri, 02 Oct 2026 09:00:00 +0000</pubDate>
    <category>Strategy</category>
    <description>Most AI projects that disappoint were chosen badly, not built badly. A seven-question checklist for picking a process where a model returns its cost in 6–12 months.</description>
  </item>
</channel>
</rss>
