https://jamiecraane.dev/tags/security/ Recent content in Security on Jamie Craane personal blog Hugo en-us Thu, 30 Apr 2026 00:00:00 +0000 https://jamiecraane.dev/2026/30/04/reliability-engineering-in-ai-era/ Thu, 30 Apr 2026 00:00:00 +0000 https://jamiecraane.dev/2026/30/04/reliability-engineering-in-ai-era/ <p>We&rsquo;ve all seen the viral stories circulating on X recently: an AI coding agent runs amok and deletes a production database. Immediately, the blame game begins: users point fingers at the agent framework, at coding assistants like Cursor, or at cloud platforms like Railway.</p> <p>My take? The operator carries the responsibility. Tool vendors absolutely should ship safer defaults — scoped tokens, confirmation prompts on destructive operations, dry-run modes — and I&rsquo;d happily criticize them for not doing so. But even with perfect vendor defaults, the buck stops with whoever wired the agent into production. Cursor and Railway are stand-ins here for <em>any</em> autonomous agent with tool access; the same arguments apply to whatever framework you&rsquo;re running.</p>