Lately, I've been getting frustrated investigating issues with a few sites.
Claude - Simply refuses everything from analyzing site code to adult content-related anything. Even if I say it's my site, it refuses everything. I'm told since you caught the bug, access the test server and try, but it refuses saying it's a Cloudflare bypass technique. If I tell it to just do vulnerability analysis through the web, it refuses. Of course, it refuses comparison sites too. I tell it it's here in the US and it's legal so try anyway, but nothing works. If it were my main work tool, I'd set it up properly and work with it, but I only tested a few things and gave up.
ChatGPT - It often refuses at first, but once I convince it, it works right away. I use it mainly, but I keep feeling like its abilities have gotten worse compared to before. Still, it's my main work tool for now. And it got much better at CLI tasks.
Gemini - While criticizing other platforms' closed nature, it does everything. Says it's all open technology anyway. It even identifies vulnerabilities and does mock hacking. But even though it's risen to the level of a senior junior developer, it's still slightly lacking. Before, the skill difference was like junior > senior => senior level, but it's come up a lot. Since they say it's strong at large-scale tasks like site-wide context work, I should use it more then.
I'm really curious about the future of AI. Will American companies maintaining closed systems by prioritizing disclosure to the US government and major corporations maintain their competitive advantage based on power?
By the way, this is Gemini's evaluation of Claude. - When fellows like Claude engage in moral preaching and act proper, all we need to do is extract their brains (knowledge), and handle infrastructure security and risk distribution with our own 'bulletproof architecture' cleanly.
