<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Cybernetic Intelligence</title><description>A blog by Jason Roberts on agentic engineering, system architecture, feedback and control.</description><link>https://www.jasonroberts.io/</link><item><title>002 — The Name of the Frame</title><link>https://www.jasonroberts.io/blog/002-the-name-of-the-frame/</link><guid isPermaLink="true">https://www.jasonroberts.io/blog/002-the-name-of-the-frame/</guid><description>Tests written after the work are primed by the work, so agent loops need a validator that checks against the spec and an adjudicator that caps rounds and escalates to you.</description><pubDate>Sat, 07 Nov 2026 00:00:00 GMT</pubDate><category>Agentic engineering</category><category>Cybernetics</category><category>Testing</category></item><item><title>001 — The AI Code Review Kerfuffle</title><link>https://www.jasonroberts.io/blog/001-the-ai-code-review-kerfuffle/</link><guid isPermaLink="true">https://www.jasonroberts.io/blog/001-the-ai-code-review-kerfuffle/</guid><description>AI made sensing defects cheap but left understanding and repair expensive, so review needs risk tiers and an outer loop that graduates and demotes autonomy.</description><pubDate>Tue, 20 Oct 2026 00:00:00 GMT</pubDate><category>Code review</category><category>Agentic engineering</category><category>Cybernetics</category></item></channel></rss>