<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Unattended on Tokenise</title><link>https://tokenise.rosvetic.com/tags/unattended/</link><description>Recent content in Unattended on Tokenise</description><generator>Hugo</generator><language>en-gb</language><lastBuildDate>Thu, 29 Oct 2026 07:13:00 +0000</lastBuildDate><atom:link href="https://tokenise.rosvetic.com/tags/unattended/index.xml" rel="self" type="application/rss+xml"/><item><title>Your overnight Claude Code run died at 2am. A StopFailure hook and a 30-line wrapper decide whether to wait or wake you</title><link>https://tokenise.rosvetic.com/posts/stopfailure-hook-overnight-retry-wrapper/</link><pubDate>Thu, 29 Oct 2026 07:13:00 +0000</pubDate><guid>https://tokenise.rosvetic.com/posts/stopfailure-hook-overnight-retry-wrapper/</guid><description>&lt;p&gt;A nightly job asks Claude Code to migrate forty files. At 2:14am the API returns a server error, the run ends mid-migration, and the cron job exits non-zero into a log nobody reads. At 8am the working tree is half changed, the log says &lt;code&gt;API Error&lt;/code&gt;, and the only way to find out whether that was a thirty-second blip or an expired key is to read it and guess.&lt;/p&gt;</description></item><item><title>Google let an agent optimise a 205,000-line repo overnight. Two of six changes shipped, and the holdout benchmark is why</title><link>https://tokenise.rosvetic.com/posts/overnight-agent-run-holdout-benchmark/</link><pubDate>Mon, 26 Oct 2026 11:10:00 +0000</pubDate><guid>https://tokenise.rosvetic.com/posts/overnight-agent-run-holdout-benchmark/</guid><description>&lt;p&gt;Google&amp;rsquo;s Agent Development Kit (ADK) team pointed Gemini 4 Argon at the &lt;a href="https://github.com/google/adk-python" target="_blank" rel="noopener noreferrer"&gt;google/adk-python&lt;/a&gt;&#10; repository, about 205,000 lines across nearly 800 source files, and left it running for 11 hours. It proposed six optimisations across 32 files. Maintainers shipped two and rejected four, &lt;a href="https://developers.googleblog.com/optimizing-a-200000-line-python-codebase-overnight-with-gemini-4-argon/" target="_blank" rel="noopener noreferrer"&gt;according to Google&amp;rsquo;s write-up&lt;/a&gt;&#10;.&lt;/p&gt;</description></item><item><title>Microsoft lets an agent work production incidents alone. Four layers decide what it can touch</title><link>https://tokenise.rosvetic.com/posts/four-layers-unattended-agent-microsoft-network-incidents/</link><pubDate>Tue, 20 Oct 2026 16:05:00 +0000</pubDate><guid>https://tokenise.rosvetic.com/posts/four-layers-unattended-agent-microsoft-network-incidents/</guid><description>&lt;p&gt;At 2am a link flaps, alarms correlate, and an agent has already pulled device state, read telemetry and written up the likely cause by the time the on-call engineer logs in. Reading is the easy part. The hard question is whether that agent should be allowed to shut down a BGP session that&amp;rsquo;s carrying live customer traffic.&lt;/p&gt;</description></item></channel></rss>