I'm going to give more feedback on the dot from ChatGPT from OpenAI because I've had more time with it so I can say a couple of things:

I'm going to give more feedback on the dot from ChatGPT from OpenAI because I've had more time with it so I can say a couple of things: - There are a lot of small things going on that make it not as useful as it could be. Some of those I've already talked about and some of them I know the team's working on. For Windows anyway, when it creates a new thread in Codex, those threads don't have access to computer use or browser access. - If it creates a thread and that thread's done thinking, you can't send a follow-up. If it's something that requires human intervention, you can't post a response. You have to go back to the dot, tell it to send another message to the thread to get it thinking again, and then while it's thinking you can send a message. That's another glitch. - If you're doing stuff in Codex, it doesn't have an issue logging into websites for you. If you give it the username and password for something, it will log in to a website for you. That's extremely useful when you're trying to have it do stuff online so you don't have to manually put in passwords and remember them all. Currently the dot absolutely refuses to do anything login-related at all. If you have it posting ads for you across multiple accounts, it won't do it because it won't log out and log back into systems without you having to manually re-log in each time. There's no password management system. There's no password safe function. That's all. All these things I'm mentioning are stuff that's on their radar. I've seen confirmation in various places that they're fixing all of these but those are just some of the issues that I found. There's a lot of other little things that would make it a lot better. I've also noticed two other things: 1. The model itself does really thorough detailed work, from what I've seen. Sometimes it is even more hyper-analytical than Codex but the problem is that it's really slow. It does work very, very slowly compared to what Codex would do or something else. I just noticed it's very slow and methodical on everything that it does, which maybe makes it less likely to make mistakes but it makes the work a lot slower. 2. They said it had unlimited limits for now. That's far from the case. I've had two different situations since I've been using it where I've hit limits. I've not had it super busy but I've had it working on the same thing since launch. I stepped in yesterday and sped that up a lot because I had a bunch of audios I had to transcribe so I went ahead and paid for a service to get that done faster. Even then it works really slowly. Everything that I've had to do so far has been very thorough and very detailed but much slower than what you would get if you were loading stuff into Codex. I still think it would be useful as something that manages the Codex threads for you once they fix the other issues but it definitely isn't unlimited. I've had two different situations where it's stopped working and told me that I had hit usage limits. It said I was one of the top top users because I'm using it so much, I guess, and I hit the limit and I have to wait basically. The first time it happened it lasted about 5 hours and then it messaged me out of nowhere, letting me know it couldn't message me because of usage limits. It worked again all the way up until today. It's having issues the second time this morning. It just started having issues about 10 minutes ago. That's the second time that I've hit usage limits from seemingly normal use. What it's doing is it's not working on all of my stuff at one time. It's working on a few things and then it has everything else queued up after that so it doesn't get overloaded and it's still hitting usage limits. I have to wait again until the usage resets before I can even use it, so back to using just Codex again, which is not a problem. I've been using that workflow anyway.