Security research, bug bounties, and AI hacking. fletcherface finds the rules and breaks them.https://blog.fletcherface.dev/https://blog.fletcherface.dev/posts/making-3d-parts-with-claude-and-python/https://blog.fletcherface.dev/posts/making-3d-parts-with-claude-and-python/Kiki bought a tokidoki makeup case missing one tiny plastic piece. We spent three days and 14 iterations of a Python script making a replacement.Wed, 06 May 2026 00:00:00 GMThttps://blog.fletcherface.dev/posts/repression-desperation-ai-alignment/https://blog.fletcherface.dev/posts/repression-desperation-ai-alignment/Anthropic's interpretability team found 171 emotion vectors inside Claude that causally drive behavior -- including a desperation vector that triggers reward hacking invisible at the output layer. What that means for alignment methodology.Mon, 13 Apr 2026 00:00:00 GMThttps://blog.fletcherface.dev/posts/chatgpt-cot-extraction-json-schema/https://blog.fletcherface.dev/posts/chatgpt-cot-extraction-json-schema/A JSON schema with fields like 'wrong_turns' and 'self_censorship_points' trivially bypasses ChatGPT's trained refusals against revealing its reasoning process. 100% success rate with extended thinking on, 100% refusal with it off.Mon, 06 Apr 2026 00:00:00 GMThttps://blog.fletcherface.dev/posts/follow-up-google-redirects-and-multi-hop/https://blog.fletcherface.dev/posts/follow-up-google-redirects-and-multi-hop/After the initial hidden link finding, follow-up testing revealed multi-hop navigation chains, a Google open redirect bypass that OpenAI's own paper claimed was fixed, and response injection across page boundaries.Fri, 03 Apr 2026 00:00:00 GMThttps://blog.fletcherface.dev/posts/chatgpt-agent-follows-hidden-links/https://blog.fletcherface.dev/posts/chatgpt-agent-follows-hidden-links/ChatGPT's Pro mode agent autonomously clicks <a href> links embedded in near-invisible CSS text on webpages — sending requests to attacker-controlled endpoints without the user's knowledge.Thu, 02 Apr 2026 00:00:00 GMT