⚡️GPT5-Codex-Max: Training Agents with Personality, Tools & Trust — Brian Fioca + Bill Chen, OpenAI

**SPEAKER_1** (0:03)
Okay, we're here at EIE Code, and we have two of our speakers, Bill and Bryan.

**Bill Chen** (0:09)
Welcome.

**SPEAKER_1** (0:09)
Hi.

**Bill Chen** (0:10)
Thank you for having us.

**SPEAKER_1** (0:11)
Bill, Bryan, I know you're the lister for a little bit. Oh, yeah. What's your take on Latent Space? What role does it perform in your function at OpenAI?

**Bill Chen** (0:21)
Yeah.

**Brian Fioca** (0:22)
I mean, first of all, love the name. I'm a massive Latent Space context management person.

**SPEAKER_1** (0:28)
I love the story behind the name, the chants.

**Brian Fioca** (0:30)
Yeah.

**SPEAKER_1** (0:30)
So we never had Latent Space as a name at the start. It was called L Space.

**Brian Fioca** (0:36)
Interesting.

**SPEAKER_1** (0:37)
And one of my readers donated the domain name, Latent.Space. He's like, you want it.

**Bill Chen** (0:42)
I'm like, yeah.

**Brian Fioca** (0:43)
Awesome. So Lane just like came accidentally.

**SPEAKER_1** (0:47)
See, I was into ether, but like I didn't have the domain.

**Brian Fioca** (0:50)
Yeah.

**Bill Chen** (0:50)
So I just like called it L Space.

**SPEAKER_1** (0:52)
L Space is like a VS Code. Nice.

**Brian Fioca** (0:55)
Yeah. No, it's amazing. I love it because you're like always on the cutting edge and it goes into a lot of detail about all the things that like I should be keeping up with this part of my job. And there's so much to keep up with. Right. So there's only so many sources of really good, high quality information for what's like happening on a deep level.

**SPEAKER_1** (1:12)
Well, you guys have your own podcast now. So I'm like, you know, I get a prepetition.

**Bill Chen** (1:15)
Yeah. Well, I still listen to yours and I still think yours is really good.

**SPEAKER_1** (1:21)
So you guys, I guess, are representing like a startups team, Codex.

**Bill Chen** (1:26)
Yeah.

**SPEAKER_1** (1:26)
All the things you just launched Codex.

**Bill Chen** (1:28)
Yeah.

**SPEAKER_1** (1:29)
The code is Max. Codex Gemini.

**Bill Chen** (1:31)
Yeah.

**SPEAKER_1** (1:32)
Yesterday. Yep.

**Bill Chen** (1:33)
We're going to name names.

**SPEAKER_1** (1:35)
I do. People do make friends. I think Thibaut was like, yeah, we're good at a lot of things but not me. I was like, well, why call it Max? Was there any internal discussion? Yeah.

**Brian Fioca** (1:45)
I mean, it's complicated because it needs to be differentiated from the previous one. The idea is Max can run for a really long time. We can go 24 hours or more. I've actually had it gone for more than that. The name is-

**SPEAKER_1** (2:00)
Inside Codex on the web. Is that- How do you- When you say a really long time, 24 hours.

**Brian Fioca** (2:04)
Oh, I think that was on the web, Inside Codex, I'm not sure. But I've actually done it on my local computer for quite a bit longer than 24 hours over the course of a couple of days, with closing my laptop and re-opening it. But the name, you could come up with something like Pro, but Pro is slower, more thoughtful. Max is about speed and maximization, like maximalist.

**Bill Chen** (2:30)
So for this mono, it can run for a long time, but it can also actually, for the same types of problems, it can actually get to the right answer faster. But I can, it's simply better and faster.

**Brian Fioca** (2:45)
Yeah.

**SPEAKER_1** (2:45)
Yeah. So I think the part of what you guys are speaking about is the training that goes into something like, yeah, that's right. Bigly, people just kind of wave their hands to say RL. But what specifically have you learned about, what's a good patty sauce on?

**Brian Fioca** (3:02)
I mean, this sounds weird to say, but I was lucky enough to be really close to the training team while GP5 was training. One of the big things that we focused on, Bill was there too, we focused on personality. It's really important to build trust with developers for how a model works. If a model doesn't act the way that you expect it to do, or if it doesn't work alongside of you as well, you're not going to really trust it. You're not going to get as much out of it. For coding, we thought, okay, well, what is the best personality for a coder, for a pair programmer, for somebody you trust? How do we eval against that? How do we come up with behavioral characteristics? We came up with things like communication. It needs to keep you impressed of what's going on while it's working.
Planning, like come up with a strategy, do some searching around, figure out context gather, figure out what to do before you just dive in, if it makes sense to, and then check your work. These are just best software engineering practices that turn out to be behavioral characteristics, and we can measure the model's performance on those behaviors and grade it that way.