After Half a Day with GPT 6, What I Like Most Is Being Asked Less

After using GPT 6 for the better part of today, the most obvious feeling is that it's faster. Fast enough that I don't need a careful comparison to notice it.

When I used 5.6 in Codex, I set the reasoning level to “medium.” Today I switched to 6 and also chose “medium.” For roughly similar questions, the wait from starting to receiving a reply felt much shorter. That's a personal impression; I can't give a number for how much faster it was.

What made the experience most comfortable, though, was that it stopped to ask me for confirmation less often.

I've been using Codex heavily for two months. 5.6 helped me do quite a lot: a mini program, a website, and the various projects I've been advancing recently. But I used to run into this situation often: I'd already told it what to do, it would prepare some materials and list the steps, then stop and ask whether I wanted it to continue.

Often, after reading it, I had nothing to add and nothing to correct. All I could say was: go ahead.

Once in a while is fine, but frequently it becomes a waste of time. I might want to hand it a project and go paint. If it stops midway, I still have to remember to check on it, or the work sits there without moving.

After switching to 6 today, that feeling was noticeably reduced. When I gave it a deliverable, it more often carried on with the work and brought me the result at the end.

This was particularly noticeable with the website. Previously, when I sent a WeChat article link and asked it to add the article to the site, it would still ask midway whether to deploy. Today, doing similar work, I asked it to add the article. Once it was done, it gave me a website preview that I could open straight away.

It's a little like a boss and an employee.

The boss assigns a task. The employee gets the materials ready, then comes over and says, “Boss, I'm ready. Shall I start?”

The preparation has already involved a pile of technical questions that the boss doesn't understand. Nothing has been made yet, so what can the boss judge? In the end, they probably just stumble out, well, go ahead and start.

In another version, the employee makes the thing and brings it over: “Boss, here's where it is. Have a look.” Now there's something to inspect. Is it what they wanted? What needs adjusting? Is the project moving along properly? All of that becomes easier to judge.

Using 6 today felt roughly like the second situation. For someone like me, who doesn't understand the technology but can judge the finished effect, it really takes a load off my mind.

Of course, the number of confirmations also depends on collaboration rules and permission settings. I can't attribute all of it to changed model capabilities after just half a day. What I can say is that, in today's actual tasks, I was interrupted less.

Today, I also gradually handed over more than a dozen projects I'd recently been using and advancing frequently to the new model, and asked it to reassess some older work.

It proposed quite a few improvements. I haven't examined some of them closely, and I'm not sure why it found so much to pick apart. Whether the suggestions are useful still depends on the results after revision. My first reaction wasn't to feel pressured, though. After all, the earlier work was also made with AI. It wasn't only finding fault with me; some of the faults belonged to the previous AI too.

Watching the mini program and website move forward version by version, I've begun to feel a little like a product manager.

That's different from how I feel when painting. Once a painting is finished, however it has turned out, I often accept it. It represents my ability, fluency, and what I could see at that particular time.

These projects made with AI, though, can be picked up again after a while and improved. Their earlier form may have reflected limits in my understanding as well as in the tools. Once I've thought things through more clearly, or the model has been updated, there's an opportunity to push a little further.

What once could only be made this way can now be taken further. Watching it gradually approach what I wanted is quite satisfying.

And once a website exists, people come and look at it. Over the last few days, I've seen visits rising. I feel people may still be interested in seeing what I'm doing. Something leaves my own computer and begins being opened by others. That adds another layer of enjoyment.

As for the improvements in aesthetics mentioned in some reviews, I haven't specifically used it to generate images yet, so I'll leave that for now. I'll try it when a real need comes up as I move my projects forward.

I don't have experience of earlier Codex models. 5.6 was the first model I used in it. When I had just changed laptops and installed Codex, the models I encountered were GPT 5.6 and Kimi K3, and I used Seedance 2.5 for video generation. I feel I arrived at a pretty good moment.

Over these two months, I've really enjoyed the creative process. Many results have gone beyond what I could previously have accomplished alone, while still meeting my own standards of taste and delivery.

Generating something is only one step. I still have to be willing to show it to other people.

So now that it's updated to 6, I naturally want to keep using it heavily. To be honest, I'm not that concerned about how it will affect the entire AI ecosystem. I care more about whether the things in my head can be expressed through the tools in my hands.

Speaking of heavy use, quota is interesting too.

I've seen Plus users in groups and comments say it gets used up too quickly and they simply can't afford to keep playing with it. My concern is a little different. Since upgrading to the higher-quota Pro20x plan about twenty days ago, I've kept wondering how to make full use of it.

And just then, there were several resets. Sometimes I'd only used ten-something or twenty percent before it reset again. Every time, I'd look at it and think: what a waste, I haven't even used much yet.

Today with 6, I finally felt that the drop in the number was noticeable. It felt as though I could at last use the quota I'd bought by doing actual work.

I also received two quota vouchers. At least this month, I have the excitement of feeling well supplied. The projects are there, and there's plenty I still want to try. I can work freely for a while.

Gradually, what I need to consider more is my own energy and imagination. I don't have to force the ideas. Some seeds were planted long ago; I just didn't know how to make them at the time. Others have grown from things I've encountered recently.

Now that I have a partner who can help me get started quickly, I really want to bring them out and try.

That's probably what felt so good today.

The views, experiences, and final judgments in this article are the author's. AI assisted with organizing the text.