There's a massive difference in response quality in my experience. For example, I asked 3.5 to find a bug in a lengthy piece of Javascript. It said it's hard t

There's a massive difference in response quality in my experience. For example, I asked 3.5 to find a bug in a lengthy piece of Javascript. It said it's hard to give a correct answer because it doesn't know what the HTML or CSS looks like. GPT4 spotted the bug almost immediately (it didn't manage to fix it though).