ChatGPT Agent Mode Review: the Good, the Bad, and the Perfect

OpenAI just released chat GPT agent and of course everyone is making videos about it talking about how wonderful it is. There are a lot of good things about this new mode, but there are also some bad things. There's one thing about it that I think is absolutely perfect. Hence the title of this video, ChatGpt Agent, the good, the bad, and the perfect. And trust me, the perfect is just beautiful in my eyes. I can't wait to get to that part of the video. Before I briefly cover the good things about chat GBT agent, let's just get one thing clear up front. If you go to OpenAI's website, it reads, "At the core of this new capability is a unified agentic system. It brings together three strengths of earlier breakthroughs. Operators ability to interact with websites, deep researchers, skill in synthesizing information, and chat GPT's intelligence and conversational fluency. " And as we briefly cover the capabilities of this new mode, you're going to quickly see that this is chat GPT operator that was released in January, reskinned, and re-released as chat GPT agent for plus users and team users, as well as prousers who previously had access. So for the past few months, this mode has only been available to pro users for $200 a month, and it was in preview research. So, they were studying people using it for $200 a month so that they could give it to you and I for $20 a month. That's a good thing for Plus users because it

definitely makes the $20 subscription more valuable than it already was. So, what exactly is ChatGpt agent and how does it work? And to make this as easy as possible to understand, it's simply ChatGpt with a mouse, a keyboard, a browser, and a Wi-Fi password. Just joking. I don't know if it has a Wi-Fi password, but I do know that it can click on websites. It can fill out forms. It can close popup windows. But as far as I know, it can't solve recapture yet. So, if those type of things pop up, then it's going to back out of the website most likely and try to find another route. But even though I think that this is still more recycled tech than it is a revolutionary new agentic mode they're releasing, that GPT being able to click on the internet is still pointing us in the direction where AI is headed. And not only can it click on websites, it can also create PowerPoint presentations and Excel spreadsheets. Now, the way that it does this is very smart. It doesn't create spreadsheets or PowerPoint presentations by opening those applications. It does it by using a library that it has access to and then writing the code for those different assets. And don't get me wrong, it's great to have these things embedded into the actual ChatGpt UI, but there are already tools on the market that do this just as well, if not better than ChatGpt. It's easy to see why everyone is excited about ChatGpt agent because it does make the $20 subscription more valuable. But because we have a limited mental capacity, it's far more difficult for us humans to

capture the full context of things than a machine can. But if we just step back for a second and we look at some current events from last month leading up to now, then I think that you'll see the chat GPT agent release a little bit differently. OpenAI didn't know it, but they had a huge problem because in the middle of June, Mark Zuckerberg purchased a 49% stake in Scale AI or $14 billion. And a couple of days later, leaks started to come out that he was offering up to $100 million signing bonuses for people to leave OpenAI. Drop a fire emoji in the comments if you'd leave Open AAI for $100 million signing bonus and drop a suitcase if you'd stay with OpenAI because of the company values. Sam Alman went on the podcast and mentioned it and you know talked about people staying at the company and not taking an offer. — They started making these like giant offers to a lot of people on our team, you know, like $100 million signing bonuses more than that comp per year. I'm really happy that at least so far none of our best people have decided to take them up on that. Not long after that podcast, we found out that many researchers did indeed leave OpenAI to join Meta's super intelligence lab and the co-creator of Chat GPT is the head of super intelligence at Meta right now. About the second week of July, Gro 4 released and crushed the 03 model on several benchmarks and then they lost the bid for Windsurf which was supposed to put them ahead in the coding industry. As I look at these recent events and what can be argued as setbacks, losses, anything that would discourage investors, I think that releasing chat GPT agent was more or less about media attention, investor confidence. And let me say it again, chat GPT agent mode, it's good for $20 a month. It's good. Let's not pretend that it's not, but it's not revolutionary, especially when you can use tools like n to create your own more relevant

workflows and agentic actions that chat GPT can take on your behalf. If we closely examine chat GPT agent, I think it gives us a perfect blueprint for how people should actually use AI. And I'm going to break this down in five different layers. Layer one is the planner. Before chat GPT agent does anything, it plans its work. And this is something that's very important for chat GPT users considering you only get about 150 to 200 messages for the 03 model per week. So that's about 30 messages a day. So whenever you start a session with chat GPT, start off using the 04 mini high model and plan out your work. plan out the prompts that you're going to use before you start using the 03 model so that you can get your money's worth and so you can save time and you're not trying to repair prompts in the middle of doing deep work. There's a parable a friend of mine told me years ago and I've heard it several times since then. If I give you an axe in 12 hours to cut down a forest of trees, what are you going to do first? And the only right answer is I'm going to take the majority of my time and sharpen my axe because preparation is the lever that makes execution more impactful. And that brings us to layer two. After we've done the preparatory work, we've planned out what we're actually going to achieve in this particular session. The next thing we want to look at is our toolkit. This is where things like custom GPTs, chat GPT projects, constructs that I'm working on and I'll be releasing soon. This is where all of these things come into play where there are automated AI systems that you've built that can help you achieve certain results on a repetitive basis. And this naturally leads into the third layer which is libraries. This is what makes things like custom GPTs and projects so powerful because you're able to upload libraries of data. And it's your libraries that really give you leverage when it comes to combating hallucinations. Layer four is an execution loop. There are many times that your human brain is going to have to engage with the content and the frameworks etc that are being laid out in front of you. I'm going to say about 90% of the time you have to read the outputs. And this means that instead of ChatGpt making people dumber like a lot of people think it's going to do, it should actually make you smarter because it's going to help you read more, read faster, recognize logical fallacies, and overall just have a sharper mental clarity because you have to verify that this information and intelligence that's being shared back to you is actually accurate and substantial. And these execution loops overlap with my guard rails, which is layer number five. You have to have guard rails in place that keep you from ending up wasting a lot of time with chat GPT. One thing you might do is when you're doing research is start to Google that research and see if you can find the independent sources yourself just to validate that the data is accurate. Use deep research so that you can visit the actual citations and read the articles for yourself. Especially if you're using chat GPT to learn because if you're an expert in the

field, which I'm assuming a lot of the people watching this video probably are, then you can read the outputs that it's giving you and you're going to nine times out of 10 be able to immediately tell if this is good or bad intelligence that is being shared back with you. So, ChatGpt agent mode isn't great, but it is good because it adds value to the $20 a month subscription. And when I look at what it actually does and the way that I can use those methods in my own workflow, it's going to save me more time. So, while it may be an underwhelming release, I think there's still something there for us to learn about how we can be more productive with AI. If you got value out of the video, make sure you hit the like button so that it helps me with the algorithm and this video reaches more people. Subscribe so that you get more helpful videos like this around AI tools, tips, and strategy. And as always, take care. Have a great day.