<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>AI Frontier India engineering blog</title>
    <link>https://aifrontierindia.com/blog</link>
    <description>Technical write-ups from building the ENT Clinical OS: real API calls, real diffs, and what we have not measured yet.</description>
    <language>en</language>
    <atom:link href="https://aifrontierindia.com/blog/feed.xml" rel="self" type="application/rss+xml"/>
    <lastBuildDate>Sun, 04 Oct 2026 00:00:00 GMT</lastBuildDate>
    <item>
      <title>sarvam-105b returned null. Tripling the token budget didn't help.</title>
      <link>https://aifrontierindia.com/blog/sarvam-reasoning-effort-null</link>
      <guid>https://aifrontierindia.com/blog/sarvam-reasoning-effort-null</guid>
      <pubDate>Sun, 04 Oct 2026 00:00:00 GMT</pubDate>
      <description>Our first live extraction call spent its whole token budget on reasoning. The fix, then a fault-injection run of all three versions of the provider against 14 failure modes, and the open bug it found.</description>
    </item>
  </channel>
</rss>
