danielgaskins.github.io/feed.xml at main · danielgaskins/danielgaskins.github.io · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Daniel Gaskins — Field Notes</title>
    <link>https://danielgaskins.com/blog.html</link>
    <description>Practical notes on agent evaluation, human review, and reliable AI systems.</description>
    <language>en-us</language>
    <lastBuildDate>Tue, 15 Sep 2026 12:00:00 GMT</lastBuildDate>
    <atom:link href="https://danielgaskins.com/feed.xml" rel="self" type="application/rss+xml" />
    <!-- SCHEDULED_FEED_ITEMS_START -->
    <item>
      <title>A response-only eval missed 176 of 263 broken agent runs</title>
      <link>https://danielgaskins.com/response-only-evals-missed-176-faults.html</link>
      <guid isPermaLink="true">https://danielgaskins.com/response-only-evals-missed-176-faults.html</guid>
      <pubDate>Thu, 03 Sep 2026 12:00:00 GMT</pubDate>
      <description>A reproducible comparison of response-only, trace-only, and trace-plus-outcome evaluators against the same 263 controlled agent failures.</description>
    </item>
    <!-- SCHEDULED_FEED_ITEMS_END -->
    <item>
      <title>Human review stops working when everything needs review</title>
      <link>https://danielgaskins.com/human-review-ai-agents.html</link>
      <guid isPermaLink="true">https://danielgaskins.com/human-review-ai-agents.html</guid>
      <pubDate>Wed, 05 Aug 2026 12:00:00 GMT</pubDate>
      <description>A practical way to decide which AI actions need review, what the reviewer needs to see, and how to keep approval queues from becoming rubber stamps.</description>
    </item>
    <item>
      <title>Your agent eval passed. Would it catch a broken tool call?</title>
      <link>https://danielgaskins.com/agent-eval-mutation-testing.html</link>
      <guid isPermaLink="true">https://danielgaskins.com/agent-eval-mutation-testing.html</guid>
      <pubDate>Mon, 03 Aug 2026 12:00:00 GMT</pubDate>
      <description>A practical explanation of mutation testing for agent evals, with examples of broken tool calls that a passing score can miss.</description>
    </item>
  </channel>
</rss>