GPT-5.6 is ABSURD! artwork

GPT-5.6 is ABSURD!

AI News Today | Julian Goldie Podcast

July 10, 2026

GPT 5.6 Just Dropped: Sol vs Terra vs Luna + 50 Real-World Builds in Agent OSThe script reviews the release of GPT 5.
Speakers: Julian Goldie
**Julian Goldie** (0:00)
Gpt 5.6 just dropped today, and it is absolutely awesome, especially Sol. So these are some of the things that we've created with it. We actually built out 50 different tasks, tested out with 50 different real world tasks like you can see right here, all sorts of crazy stuff. And I will say it is mind blowing in terms of the stuff it can create and how powerful it actually is. So this is open world game, for example, we create like a Doom style game, as you can see right here. The other thing that I would say here is like the games it creates are really smooth, really nice to use, really nice graphics. And on terms of our leaderboards, I'll show you how it performs in a second. So these are some of the builds, we can come on to more of them in a second. Here's another one. And basically what we've got here is a 5.6 solar system. So there are three new models, the sun, the moon, the earth from Gpt 5.6. You got Sol, which is the sun, that is a flagship model, the most powerful. It's basically like Fable 5 equivalent. Then you have Terra, the workhorse. And then you have Luna, the moon. And these are three different models that you can pick and choose from to make sure you get the best out of them. Now, if you're wondering, okay, which one should you actually choose? I would go with if you want like something super cheap. And then Terra, if you've got like those sort of day to day tasks. And then Sol, if you need something flagship level. So for example, for our 50 builds that we created over here, all of these were built with Sol. Sol is by far the most powerful example so far that I've built with. And it's the best model from Gpt 5.6. The other thing I would say here is that it is the best model has come from ChatGpt for a long time. I mean, for example, if you look at this, this is created with a video. This is a full video created with Gpt 5.6 using our Agent OS system. And you can see how beautiful it is, right? It literally we just said to it, go off and create a video. And then it went from there. If you're wondering, OK, what are the best ways to use it? How can you get the most out of it, et cetera? Here's what I recommend.
So, for example, what I recommend here is that you go over to Codex. We've actually built it already into the Agent OS. And the great thing about that is we can just use OAuth, so we don't have to pay for the API. We can just use our chat Gpt subscription. And then from here, we can make sure that we have Gpt 5.6 selected and we can ask it to build things directly inside here. Now, if you want something autonomous and powerful, then I would actually recommend the forward slash goal mode. So with forward slash goal, you can say, for example, build me a 3D racing game, single HTML, and then build out a website to promote it. Here's exactly what we did before. It runs in its own sandbox. So this is like fully siloed away from everything else. And then you can see the actual creation over here. So actually created the website and the game autonomously by itself. This is the website they created. Looks pretty nice. You can create nice designs and that sort of thing. And then if we click on this, you can see an example of the actual game created as well, which is pretty nice. This was just a one shot forward slash goal. If you wanted to make it even better, you can say, okay, don't actually complete the task until the UI is like next level, whatever. But the whole point here is like you can build autonomously using goal mode or you can go back and forth inside the chat mode. Bear in mind, it can also generate images for you. And then the great thing about having an agent OS is like when you're building out with this. So for example, like that game I just showed you, you have everything to preview over here. You can open up in a new tab, you can save it, copy it for later, F and L. So Gpt 5.6 is really, really good. Here's some other examples of what we've created. And the cool thing about this is everything is saved inside one workspace. So it's easy to organize, easy to come back to. And it is all ready to go in one place. I don't know about you, but for me, there's no way I'm going to be going off to chat Gpt, and then I'm going to be clicking over to Claude like that's just too much work. And it would make everything way too fragmented. Better to have it inside one system, like you see with this. And also the cool thing with this is every time we use it, it's learning from our memory galaxy, updating our memory galaxy, and then making a whole obsidian system way better to use, adding new memories in all the time, which makes it exponentially more powerful the more we use it. Now, if you're wondering how to get the outputs from the video that I just showed you a minute ago, what you can do is use the video agent. So what we have here is a video agent trained on Remotion as a skill, open source skill, then we have Gpt 5.6, and Gpt 5.6 can create and edit the video, put it all together, generate the script, generate the voiceover, and then also generate the avatar and plug it into one beautiful video, like you can see right here. The other thing that I'd recommend is that you try using it with Hermes. So Hermes is a really powerful system. We've got Gpt 5.6 working over here and you can go back and forth with it. And again, like unlike, for example, Fable 5 where you have to pay for credits unless you have it on the subscription, which you're going to use 50% of your subscription on.

8 more minutes of transcript below

Feed this to your agent

Try it now — copy, paste, done:

curl -H "x-api-key: pt_demo" \
  https://spoken.md/transcripts/1000651996090

Works with Claude, ChatGPT, Cursor, and any agent that makes HTTP calls.

From $0.10 per transcript. No subscription. Credits never expire.

Using your own key:

curl -H "x-api-key: YOUR_KEY" \
  https://spoken.md/transcripts/1000776324784