<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>Tapio Vepsäläinen</title><description>Thoughts on software, coding agents, critical thinking, and life in general.</description><link>https://vepsala.dev/</link><language>en</language><item><title>Does using an LLM actually improve your writing?</title><link>https://vepsala.dev/blog/does-ai-make-writing-better/</link><guid isPermaLink="true">https://vepsala.dev/blog/does-ai-make-writing-better/</guid><description>Does using an LLM actually improve your writing? I found no direct answer, but three concepts (anchoring, design fixation, and automation bias) help explain why the question is so hard to pin down, and give some dimensions worth judging AI-assisted writing by.</description><pubDate>Wed, 12 Aug 2026 00:00:00 GMT</pubDate></item><item><title>Sharing secrets with agents</title><link>https://vepsala.dev/blog/sharing-secrets-with-agents/</link><guid isPermaLink="true">https://vepsala.dev/blog/sharing-secrets-with-agents/</guid><description>Coding agents are much less useful without credentials. Two moves — assume the leak, then make the leaked thing worthless — and what that looks like in practice with a sandbox, a local proxy, and expendable tokens.</description><pubDate>Tue, 30 Jun 2026 00:00:00 GMT</pubDate></item></channel></rss>