开发者实测GPT-6 Astra：3D生成与多智能体推理能力展示
First impressions of GPT-6 Astra from developers
GPT-6系列新模型的首批真实场景演示，重点展示了3D内容生成与多智能体自主推理的新能力，对关注前沿模型落地的从业者极具参考价值。
It is smarter, it unlocks a lot of workflows that I previously just haven't been able to do. And what blew your mind with Astra? Can you show me? What I've built here is a voxel 3D representation of historic London, which goes across the different eras, medieval London, Tudor London, and it transforms all within the same map. Oh, you know it, I can actually play this. And what I can do is I can even keep playing the same person and it can keep going into the modern times.
它更智能，解锁了许多我过去无法完成的工作流。Astra 有什么让你大吃一惊的地方？能给我展示一下吗？我在这里构建的是历史伦敦的体素 3D 表示，跨越不同时代，中世纪伦敦、都铎王朝伦敦，并在同一张地图内实现转换。哦，你知道的，我实际上可以播放这个视频。我可以继续扮演同一个人，并且它可以一直延续到现代。
But here I pretty much gave it maybe a couple more prompts in terms of I wanted to go overhead so it looks like GTA 2. Suddenly I've got this pretty much kind of simulation slash game that really works super cool. You know, so here I have sort of asked you to design like a matcha shop website. But like the fact that like here you have the photo in the center and it actually wrote a really interesting prompt even for that photo to be generated this way and like the circle around it and so on.
但在这里，我基本上又给了它几个提示，比如我想要俯视视角，让它看起来像 GTA 2。突然之间，我就有了这种非常酷的模拟/游戏效果。你知道，所以我这里大致要求你设计一个抹茶店网站。但就像事实那样，照片在中心，它甚至为这张照片生成了一个非常有趣的提示，让照片以这种方式生成，还有周围的圆圈等等。
And then it came up with another version of that, which is completely different. But also surprisingly, it even comes up with like things that I wouldn't otherwise like think about, you know, in which I can finally start using it for like a new source of inspiration, which I find incredibly exciting. I went to a big conference called DEF CON where there are these big puzzle challenges that me and my friends spent days trying to solve.
然后它提出了另一个完全不同的版本。但同样令人惊讶的是，它甚至提出了一些我原本不会想到的东西，你知道吗？我终于可以开始用它作为新的灵感来源，这让我觉得非常兴奋。我去参加了一个名为 DEF CON 的大型会议，那里有这些大型谜题挑战，我和朋友们花了几天时间试图解决它们。
It solved this one big puzzle, which was like a bunch of Rubik's cubes that were arranged in this weird three by four pattern that you had to deduce a message from. It was able to get that three out of three times. The one thing it did need is it got the official hint from the puzzle creators. So that's the same thing we got when we were doing it. But once you gave it that it got the actual solution. So the model comes up with a theory.
它解决了其中一个大谜题，那是一堆魔方，排列成奇怪的三乘四图案，你必须从中推断出一条信息。它三次都成功了。唯一需要的是它获得了谜题创作者提供的官方提示。所以这也是我们在做的时候得到的。但一旦你给它那个提示，它就得到了实际的答案。因此模型会提出一个理论。
It sends off another agent to go test it, sees how it works. So these are all running in parallel. Main agent is just kind of orchestrating these together. I found this model is much better at keeping itself on track and knocking out of those doom loops, which is a lot of the reason why it's been able to keep on task and solve these problems. It's been amazing. I've got so many problems that accumulated that I couldn't quite solve with the previous models.
它发送另一个代理去测试该理论，看看它是如何工作的。所有这些都在并行运行。主代理只是将这些协调在一起。我发现这个模型在保持自身方向并跳出那些恶性循环方面要好得多，这也是它能够持续专注于任务并解决这些问题的主要原因之一。这太棒了。我有太多之前用旧模型无法解决的问题积累起来。
And now I've got finally a chance to solve them. It's been incredible to see Astra going through them one by one and getting them done.
而现在我终于有机会解决它们了。看到 Astra 逐一处理并完成它们真是令人难以置信。
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力