<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Ai on Sergio Comerón Blog</title>
    <link>https://sergiocomeron.com/blog/en/tags/ai/</link>
    <description>Recent content in Ai on Sergio Comerón Blog</description>
    <generator>Hugo</generator>
    <language>en</language>
    <lastBuildDate>Thu, 28 May 2026 00:00:00 +0200</lastBuildDate>
    <atom:link href="https://sergiocomeron.com/blog/en/tags/ai/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Claude Opus 4.8 lands: better at code, more honest and with sub-agents in parallel</title>
      <link>https://sergiocomeron.com/blog/en/posts/claude-opus-4-8/</link>
      <pubDate>Thu, 28 May 2026 00:00:00 +0200</pubDate>
      <guid>https://sergiocomeron.com/blog/en/posts/claude-opus-4-8/</guid>
      <description>&lt;p&gt;Anthropic unveiled Claude Opus 4.8 this afternoon, its most capable production model. It arrives less than two months after Opus 4.7 and, most strikingly, without a price bump: it costs the same as the previous version.&lt;/p&gt;&#xA;&lt;h2 id=&#34;the-improvements-in-numbers&#34;&gt;The improvements in numbers&lt;/h2&gt;&#xA;&lt;p&gt;The benchmarks Anthropic published compare directly against Opus 4.7:&lt;/p&gt;&#xA;&lt;ul&gt;&#xA;&lt;li&gt;&lt;strong&gt;Agentic coding&lt;/strong&gt;: 64.3% → &lt;strong&gt;69.2%&lt;/strong&gt; — Long-running tasks where the model chains tool calls, reads files, runs tests and self-corrects.&lt;/li&gt;&#xA;&lt;li&gt;&lt;strong&gt;Multidisciplinary reasoning with tools&lt;/strong&gt;: 54.7% → &lt;strong&gt;57.9%&lt;/strong&gt; — Problems requiring jumps between domains and the use of external tool context.&lt;/li&gt;&#xA;&lt;li&gt;&lt;strong&gt;Knowledge work&lt;/strong&gt;: 1753 → &lt;strong&gt;1890&lt;/strong&gt; — Anthropic&amp;rsquo;s internal metric for analysis, writing and synthesis tasks.&lt;/li&gt;&#xA;&lt;/ul&gt;&#xA;&lt;p&gt;According to the data Anthropic shared, Opus 4.8 beats GPT-5.5 and Gemini 3.1 Pro on several of these benchmarks. On the internal Super-Agent benchmark, it&amp;rsquo;s the only model that completes every case end to end.&lt;/p&gt;</description>
    </item>
    <item>
      <title>How I code in 2026: my stack in the age of agents</title>
      <link>https://sergiocomeron.com/blog/en/posts/como-programo-en-2026/</link>
      <pubDate>Wed, 27 May 2026 00:00:00 +0200</pubDate>
      <guid>https://sergiocomeron.com/blog/en/posts/como-programo-en-2026/</guid>
      <description>&lt;p&gt;I&amp;rsquo;ve been turning over an uncomfortable thought for weeks: I open an editor to write code less and less. I direct it, I review it, I approve it, but I type less every day. And yet I&amp;rsquo;m publishing more than ever: two podcasts, a personal website with nine tools, server monitoring, my own Jitsi instance, an Apple fan site, more Moodle plugins than ever, this blog. Something has shifted in how I program, and I think it&amp;rsquo;s worth writing down.&lt;/p&gt;</description>
    </item>
    <item>
      <title>Gemini 3.5 Flash: speed over intelligence for AI agents</title>
      <link>https://sergiocomeron.com/blog/en/posts/gemini-35-flash/</link>
      <pubDate>Fri, 22 May 2026 00:00:00 +0200</pubDate>
      <guid>https://sergiocomeron.com/blog/en/posts/gemini-35-flash/</guid>
      <description>&lt;p&gt;Google presented Gemini 3.5 Flash at I/O on May 19th. Fast, cheap, optimised for agents. Good news for developers. But the most interesting thing is not what it does for Google — it&amp;rsquo;s who else is going to use it.&lt;/p&gt;&#xA;&lt;p&gt;Apple has a multi-year agreement with Google to integrate Gemini models into Apple Intelligence. WWDC 2026 is on June 8th. All signs point to the new Siri — the one they&amp;rsquo;ve been promising for years and never quite delivering — running, at least in part, on this very model.&lt;/p&gt;</description>
    </item>
    <item>
      <title>An AI-generated podcast: how we built it and what we&#39;ve learned</title>
      <link>https://sergiocomeron.com/blog/en/posts/podcast-ia/</link>
      <pubDate>Thu, 21 May 2026 00:00:00 +0200</pubDate>
      <guid>https://sergiocomeron.com/blog/en/posts/podcast-ia/</guid>
      <description>&lt;p&gt;A few days ago I published the first episodes of El podcast de Sergio and El informativo on Apple Podcasts. Neither was recorded by a human.&lt;/p&gt;&#xA;&lt;p&gt;The entire process —finding the topic, writing the script, synthesising the voice, assembling the audio and publishing it— is handled by an AI agent. Here&amp;rsquo;s how it works and what we&amp;rsquo;ve learned along the way.&lt;/p&gt;&#xA;&lt;h2 id=&#34;why&#34;&gt;Why&lt;/h2&gt;&#xA;&lt;p&gt;Not because it&amp;rsquo;s the most comfortable way to make a podcast. I did it because I wanted to explore how far synthetic voice quality can go in Spanish, and because building an automated production pipeline seemed like an interesting technical problem. You choose the topic and review the script — the agent handles the rest.&lt;/p&gt;</description>
    </item>
    <item>
      <title>From Recorded Class to Moodle Quiz in One Click</title>
      <link>https://sergiocomeron.com/blog/en/posts/ia-grabaciones-cuestionario-moodle/</link>
      <pubDate>Sat, 16 May 2026 00:00:00 +0200</pubDate>
      <guid>https://sergiocomeron.com/blog/en/posts/ia-grabaciones-cuestionario-moodle/</guid>
      <description>&lt;p&gt;Every time a teacher records a class, a learning opportunity is created that usually goes to waste. The video sits in storage, students watch it (or don&amp;rsquo;t), and the teacher moves on. No assessment, no summary, no way to navigate the content.&lt;/p&gt;&#xA;&lt;p&gt;&lt;a href=&#34;https://moodle.org/plugins/mod_jitsi&#34;&gt;mod_jitsi&lt;/a&gt; changes that with three AI-powered features that activate directly from the Moodle activity.&lt;/p&gt;&#xA;&lt;h2 id=&#34;automatic-quiz&#34;&gt;Automatic Quiz&lt;/h2&gt;&#xA;&lt;p&gt;With one click, the plugin generates a complete quiz based on the actual content of the recording. The questions aren&amp;rsquo;t generic: they&amp;rsquo;re extracted from what the teacher explained in that specific session. The quiz is created directly as a Quiz activity in the same course — no copy-pasting required.&lt;/p&gt;</description>
    </item>
    <item>
      <title>What&#39;s new in the Jitsi plugin v5.0</title>
      <link>https://sergiocomeron.com/blog/en/posts/jitsi-50/</link>
      <pubDate>Sun, 03 May 2026 00:00:00 +0200</pubDate>
      <guid>https://sergiocomeron.com/blog/en/posts/jitsi-50/</guid>
      <description>&lt;p&gt;With version &lt;strong&gt;5.0 of &lt;a href=&#34;https://github.com/SergioComeron/moodle-mod_jitsi&#34;&gt;mod_jitsi&lt;/a&gt;&lt;/strong&gt; the plugin takes a significant step forward in four directions: &lt;strong&gt;installation management&lt;/strong&gt; through a developer portal, &lt;strong&gt;recording analytics&lt;/strong&gt; with a viewing heatmap and attendance report, &lt;strong&gt;real-time presence&lt;/strong&gt; in the room, and several &lt;strong&gt;security and performance improvements&lt;/strong&gt;.&lt;/p&gt;&#xA;&lt;h2 id=&#34;1-developer-portal&#34;&gt;1. Developer Portal&lt;/h2&gt;&#xA;&lt;p&gt;Until now there was no mechanism to know how many installations of the plugin existed or which features were being used. v5.0 adds an opt-in registration and telemetry system managed from &lt;a href=&#34;https://portal.sergiocomeron.com&#34;&gt;portal.sergiocomeron.com&lt;/a&gt;.&lt;/p&gt;</description>
    </item>
    <item>
      <title>What&#39;s new in the Jitsi plugin v4.6</title>
      <link>https://sergiocomeron.com/blog/en/posts/jitsi-46/</link>
      <pubDate>Sun, 19 Apr 2026 00:00:00 +0200</pubDate>
      <guid>https://sergiocomeron.com/blog/en/posts/jitsi-46/</guid>
      <description>&lt;p&gt;Version &lt;strong&gt;4.6 of the &lt;a href=&#34;https://github.com/udima-university/moodle-mod_jitsi&#34;&gt;mod_jitsi&lt;/a&gt; plugin&lt;/strong&gt; is probably the biggest update since we introduced automatic provisioning on GCP. It touches four very different but complementary areas: &lt;strong&gt;recordings&lt;/strong&gt;, &lt;strong&gt;AI on top of the recordings&lt;/strong&gt;, &lt;strong&gt;private sessions&lt;/strong&gt; between users, and a new &lt;strong&gt;tutoring-hours system&lt;/strong&gt;. Let&amp;rsquo;s go through them one by one.&lt;/p&gt;&#xA;&lt;h2 id=&#34;1-recordings-on-gcp-the-jibri-pool&#34;&gt;1. Recordings on GCP: the Jibri pool&lt;/h2&gt;&#xA;&lt;h3 id=&#34;what-is-jibri&#34;&gt;What is Jibri&lt;/h3&gt;&#xA;&lt;p&gt;&lt;a href=&#34;https://github.com/jitsi/jibri&#34;&gt;Jibri&lt;/a&gt; is the official recording component of Jitsi Meet. It is a separate virtual machine that joins the room as an invisible participant, captures video and audio with Chrome and saves the result as an MP4 file.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
