I have the opposite perception: they’re the only company in the space that seems to have a clue what responsible software engineering is.

Gemini Code and Cursor both did such a poor job sandboxing their agents that the exploits sound like punchlines, while Microsoft doesn’t even try with Copilot Agentic.

Countless Cursor bugs have been fixed with obviously vibe-coded fake solutions (you can see if you poke into code embedded in their binaries) which don’t address the problems on a fundamental level at all and suggest no human thinking was involved.

Claude has had some vulnerabilities, but many fewer, and they’re the only company that even seemed to treat security like a serious concern, and are now publishing useful related open source projects. (Not that your specific complaint isn’t valid, that’s been a pain point for me to, but in terms of the overall picture that’s small potatoes.)

I’m personally pretty meh on their models, but it’s wild to me to hear these claims about their software when all of the alternatives have been so unsafe that I’d ban them from any systems I was in charge of.

I suggest spending some time with Codex. Claude likes to hack objectives, it's really messy and it'll run off sometimes without a clear idea of what you want or how a project works. That is all fine when you're a non-technical person vibe coding a demo, but it really kills the product when you're working on hard tasks in a large codebase.

Every tool in this space is blatantly unsafe. The sandboxes that people have designed are quite ineffective.