<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Eugenio Nerelli — writing</title><description>Eugenio Nerelli, PhD researcher at the University of València. Work on generative models in virtual environments, and open source tools that run on a laptop.</description><link>https://nerln.pages.dev/</link><language>en</language><item><title>One prompt injection out of six got through</title><link>https://nerln.pages.dev/blog/one-prompt-injection-out-of-six/</link><guid isPermaLink="true">https://nerln.pages.dev/blog/one-prompt-injection-out-of-six/</guid><description>Testing an LLM job scheduler against hostile input, what the failing case looked like, and why the number stayed in the README.</description><pubDate>Tue, 04 Aug 2026 00:00:00 GMT</pubDate><category>llm</category><category>testing</category><category>open source</category></item><item><title>Everything runs on the laptop</title><link>https://nerln.pages.dev/blog/everything-runs-on-the-laptop/</link><guid isPermaLink="true">https://nerln.pages.dev/blog/everything-runs-on-the-laptop/</guid><description>Four of these tools do the work on one machine with no account behind them. Here is what that constraint costs, in seconds, in gigabytes, and in things a cloud service would get for free.</description><pubDate>Mon, 03 Aug 2026 00:00:00 GMT</pubDate><category>local-first</category><category>apple silicon</category></item></channel></rss>