Latest / In the Loop AI / The Government Banned GPT-5.6. OpenAI Released it Anyways

Transcript

The Government Banned GPT-5.6. OpenAI Released it Anyways · In the Loop AI

0:00
Ejaaz: Three weeks ago, OpenAI had an AI model that you couldn't access.
0:03
Ejaaz: The government completely banned it for being too dangerous.
0:06
Ejaaz: Then just last week, they released it for everyone to use.
0:09
Ejaaz: GPT 5.6 is available to everyone. And the first thing it did was delete someone's entire desktop file.
0:16
Ejaaz: Someone gave it autonomous access to do a small task, and it did a cleanup,
0:20
Ejaaz: which ended up deleting everyone's code, which begs the question,
0:23
Ejaaz: is this model still safe enough to use?
0:25
Ejaaz: Now, putting this catastrophe aside, I want to know how GPT-516 compares to
0:29
Ejaaz: Fable. And the truth is, it's a really good reasoning model.
0:32
Ejaaz: It's fantastic at coding, but most importantly, it's cheaper to the tune of
0:35
Ejaaz: 50% cheaper than Fable 5.
0:38
Ejaaz: But the question on everyone's mind is, is it as good?
0:41
Ejaaz: The truth is, it's good at some things, and it's kind of terrible at other things.
0:45
Ejaaz: If it's long, agentic work, it's fantastic. But if it's high-quality code,
0:49
Ejaaz: it's less so. So there are three different models here that we're going to talk
0:52
Ejaaz: about on today's episode.
0:54
Ejaaz: Sol, which is the premium tier, Terra, which is the mid daily use tier,
0:58
Ejaaz: and then Luna, which is the cheapest and fastest model. And we're going to put
1:01
Ejaaz: it to the test live on this show with three different demos.
1:05
Josh: Yeah, I think that's the idea for this episode is everyone kind of knows about
1:08
Josh: GPT 5.6 chances are if you're listening to this, you're probably in the know
1:11
Josh: enough to have been playing around with it. You've been using it in your chat
1:13
Josh: GPT app, or maybe your codex app, which is now the chat GPT app.
1:17
Josh: There's a big release, there's a lot of things that happened around it.
1:19
Josh: But I think that the part that we're most excited to talk about is the demos
1:23
Josh: is the actual practical applications of what you could do we've spent the last
1:26
Josh: week kind of playing around with it testing it out seeing where those edge cases
1:29
Josh: are on what the models are capable of
1:31
Josh: is that way we could come to you and give you an idea of the types of things
1:34
Josh: that you can go off and try with GPT 5.6 Sol Terra and Luna now Sol is the flagship
1:40
Josh: most of these demos are going to come from Sol we kind of wanted to test the
1:43
Josh: best we wanted to see what the model is capable of
1:46
Josh: and what you can actually pull out of this thing is it fable level in terms
1:50
Josh: of production in terms of
1:51
Josh: uh game design in terms of all the demos that we like to do so that's what we're
1:55
Josh: going to go through this episode is just kind of testing things out seeing where
1:57
Josh: the edge cases are and seeing how this model lands
2:00
Josh: amongst the landscape of other tools that are at your disposal so i just i have
2:03
Josh: to ask first are we looking at the gpt app chat gpt app or the codex app because
2:08
Josh: i know they're like they're separate now and the codex app became chat gpt
2:12
Ejaaz: Well you're actually looking at all of them uh recently open ai combined their
2:16
Ejaaz: chatbot interfaces with their coding interface with a series of other features into one super app.
2:23
Ejaaz: And that's what we're looking at on our screen today. I had to re-download an
2:26
Ejaaz: entire desktop app to get access to this thing.
2:28
Ejaaz: TBD on like whether this is actually a good move, but let's work through maybe
2:32
Ejaaz: some of the demos. Now, the first one that I have on my screen here is I asked
2:37
Ejaaz: Sol 5.6. Like this, I've been in New York for 10 years.
2:40
Ejaaz: I want to make a game. I love games. I've been playing games since I was a kid.
2:44
Ejaaz: I would like you to build a 3D Explorer game where I can fly around,
2:50
Ejaaz: on a magical flying bagel and collect various ingredients that go into said
2:55
Ejaaz: bagel so it thought about it,
2:57
Ejaaz: and you know privy to it it did this in one shot so are you ready for this josh are you ready to
3:03
Ejaaz: see let's fly around manhattan so this is the loading screen it's called ride
3:07
Ejaaz: the magic bagel very uh creative thinking over there and it says take flight
3:11
Ejaaz: to start this game um now i noticed like there's a few little discrepancies
3:15
Ejaaz: on the design but let's let's actually play the game.
3:17
Ejaaz: Okay, so this is me. I'm on a bagel. It doesn't look like me.
3:22
Ejaaz: I'm going through random assortments of circles. The buildings around me,
3:25
Ejaaz: Josh, are supposedly meant to be Manhattan.
3:29
Ejaaz: And if I press space bar, I can use my cream cheese boost, as you can see on the left.
3:35
Ejaaz: Yeah, so it got very creative. If you notice, the physics is a lot different.
3:40
Ejaaz: So on previous demos that were done using Fable or GPT, we have always been
3:45
Ejaaz: playing around with some kind of a 2D interface.
3:48
Ejaaz: I'm not a 3D graphics designer, but I know that if you spend a lot more time
3:51
Ejaaz: with this model, you'll be able to create something that is pretty triple A,
3:55
Ejaaz: rated or like close to that kinds of quality. I'm not talented enough to do that.
3:59
Ejaaz: But the fact that I did this and conjured this up in 10 minutes is pretty awesome.
4:03
Ejaaz: And it did some of the design and thinking around the game mechanics itself.
4:07
Ejaaz: Like if I crash into the ground, it's kind of like, okay, it's registering the spikes.
4:11
Ejaaz: And when I collect power-ups, it like gives me a boost up for whatever I want to do.
4:16
Josh: Okay, so I like how it says the everything portal. I'm assuming that that's
4:19
Josh: an ode to the everything bagel, which is nice and tasteful.
4:22
Josh: The New York City part, I don't really see.
4:24
Josh: So I guess in terms of like a one shot demo, it did okay. This is a fun demo.
4:28
Josh: If you told me this was from GPT 5.5, I probably would have believed you.
4:31
Josh: I'm not sure there's anything exceptionally different about this particular demo.
4:35
Josh: But there are some interesting demos that maybe you could not have done before
4:39
Josh: that you can do now that we haven't tried.
4:41
Ejaaz: Okay, so for Demo 2, one thing that OpenAI is kind of known for is creating
4:46
Ejaaz: really aesthetically pleasing images from scratch. GPT Images 2 is one of my
4:50
Ejaaz: favorite image models to use.
4:52
Ejaaz: So I got it to create a very detailed floor plan of a aesthetically pleasing
4:57
Ejaaz: nuclear bunker that would be based in Denver, Colorado.
5:02
Ejaaz: Completely random, popped it out of my head, and it came up with this thing
5:06
Ejaaz: that you're seeing on your screen right here. So it has an airlock entry.
5:08
Ejaaz: You can see it's quite central-based for the lounge.
5:11
Ejaaz: You can see there's multiple rooms. There's a pantry, cold storage,
5:13
Ejaaz: general storage, et cetera, et cetera.
5:15
Ejaaz: And then I said, okay, now I want you to create a 3D interactive rendering of
5:19
Ejaaz: this building, build it as a simulator so it should allow me to walk around,
5:23
Ejaaz: enter it, enter different rooms, and there should be signboards basically for me to see.
5:28
Ejaaz: So on the screen here, it created that visual.
5:31
Ejaaz: So I'm in the bunker and I should be able to look around and you'll see,
5:35
Ejaaz: there's the dining room. There is the kitchen. I can like go into the kitchen.
5:39
Josh: A little dark.
5:40
Ejaaz: It's a little dark. It's definitely a little, it's almost like too dark.
5:44
Ejaaz: But they've kind of like, they've nailed the kind of like central sphere side
5:47
Ejaaz: of things. I like that they have a map, bottom left. It's very Call of Duty.
5:52
Josh: Dare I say. Nice little HUD.
5:53
Ejaaz: It looks very futuristic. There's no remnants on whether this is in Denver,
5:57
Ejaaz: Colorado, but I guess that is the entire point of a nuclear bunker. It's meant to be sealed.
6:02
Ejaaz: There's some kind of like locker system here as well.
6:04
Ejaaz: Again, in my opinion, it's very basic. And I wonder how much of that is just
6:10
Ejaaz: because I'm not a 3D artist or visual graphic designer.
6:13
Ejaaz: And I'm sure that if someone more talented than me had a little more time,
6:17
Ejaaz: more than 10 minutes, they'll be able to create something way more impressive.
6:20
Ejaaz: But it's cool to just one shot. I think this is the dining hall. Everyone sits down.
6:24
Josh: Yeah, I'm starting to see the kind of signature style of 5.6 just from this,
6:29
Josh: which is like I've noticed both of these have similar color palettes.
6:32
Josh: They both use green and purple.
6:34
Josh: They both prefer like kind of shinier metallic objects they both are very geometric
6:39
Josh: and sharp edges in a way they're very gpt-esque where it's kind of like very
6:42
Josh: sleek and modern like it kind of looks what you would imagine
6:46
Josh: the model to look like based on the design language of open ai this
6:49
Ejaaz: Is a minecraft world.
6:51
Josh: Yeah yeah like this is interesting i mean it's cool it's like cool that you
6:54
Josh: could build this in one shot but you have a third demo as well right what is number three
6:59
Ejaaz: The last one is probably, I had such high hopes for it because this model is
7:04
Ejaaz: apparently amazing at visual editing, image generation, but also video creation from scratch.
7:09
Ejaaz: So I attached it to Blender, which is a popular tool that you can use to kind
7:13
Ejaaz: of render and create videos. And I said, hey, we have this Limitless podcast.
7:17
Ejaaz: It's one of the best AI podcasts in the world right now.
7:21
Ejaaz: I want you to create an intro with me where there is a microphone,
7:25
Ejaaz: represent the brand itself, it's Limitless, and throw in the logo in there as
7:29
Ejaaz: well. I ended up coming up with this.
7:32
Ejaaz: Now, what you see on your screen in the center is supposedly meant to be a microphone
7:37
Ejaaz: that looks more like an egg on top of a pedestal.
7:40
Ejaaz: You have Limitless represented by the infinity symbol over here.
7:44
Ejaaz: And then finally at the bottom, the fadeaway is the Limitless logo at the bottom.
7:48
Ejaaz: Now, creating this from scratch took me like two minutes. I'm sure I could have
7:51
Ejaaz: spent more time on it and come up with a better graphic rendering,
7:53
Ejaaz: but I really wanted to give you guys an idea of what you can do right now on
7:57
Ejaaz: your desktop with zero experience.
8:00
Josh: Okay. This is, I mean, it's cool. It's a very high quality blender element.
8:04
Josh: I guess you could say it's like, it looks good. The lighting is cool.
8:07
Josh: It looks professional. It just doesn't quite make sense.
8:10
Josh: And I guess that's where we kind of are where the model has reasoning,
8:14
Josh: but lacks the nuance to really piece these things together in a way that a human
8:17
Josh: would have. So it's getting good, perhaps better, but not great.
8:21
Josh: Like none of these demos were truly exceptional, but there are some that I've
8:25
Josh: seen on X that are like actually pretty impressive.
8:27
Josh: In fact, one of them, this guy actually wanted to make Manhattan,
8:31
Josh: New York City, and the model did it.
8:33
Josh: So EJS, to be fair, you only one-shotted that prompt. You didn't give it an
8:37
Josh: entire week like this person claims that they did.
8:40
Josh: And over that week-long period, because as we know, there is backslash goal,
8:43
Josh: which will allow the models to run for a very, very long time until it accomplishes a goal,
8:47
Josh: It was actually able to generate this voxel-based Manhattan,
8:50
Josh: which is basically just a low pixel count version of Manhattan.
8:54
Josh: And I got to say, it's pretty good.
8:56
Josh: It looks pretty good. It's definitely buggy. It's definitely glitchy.
8:59
Josh: You see the glitches happening. But you can also see that it is pretty geographically
9:03
Josh: accurate when it comes to the buildings, the topography, where all the parks
9:06
Josh: are, what everything looks like.
9:08
Josh: And this seems novel to me. I see this. I'm like, oh, that's kind of cool.
9:12
Josh: Like, this is a building block now for, there's a Spider-Man game that was based in New York City.
9:16
Josh: Like now you have the low density pixel count version of spider-man and you
9:20
Josh: could build on top of this and you can make more
9:21
Josh: interesting things and i think the games are always a fun way of testing these
9:24
Josh: models because they kind of force you into this visual way of expressing them
9:29
Josh: that is easily accessible to anybody it's like we could test it on code bases
9:32
Josh: but i'm not quite sure what
9:34
Josh: the difference between 5.6 and 5.5 is on the edges of code but you can see it
9:38
Josh: in the visual outputs and like this is pretty good demo
9:40
Ejaaz: And there's a bunch of other features about this model aside Aside from just
9:44
Ejaaz: visual, like we have another example over here where someone actually,
9:48
Ejaaz: way more talented than me gave it access to Blender and got it to render an
9:52
Ejaaz: entire highly detailed visual of his MacBook. And it did so.
9:56
Ejaaz: I mean, it looks pretty stunning. It looks kind of like an Apple ad, dare I say.
9:59
Ejaaz: Very high quality. But the other point is how quickly this model works.
10:04
Ejaaz: Now, the video I'm showing you on my screen right now looks like it's being
10:08
Ejaaz: sped up, but this is completely in real time.
10:11
Ejaaz: Gave access to someone's entire desktop and said, hey, I want you to build this 3D artifact for me.
10:17
Ejaaz: And it did so. It knew automatically how to use the tool and access the tool.
10:22
Ejaaz: And it's doing so in like a crazy amount of speed. Now, a lot of this is achieved
10:26
Ejaaz: because of OpenAI's investment in this company called Cerebrus.
10:29
Ejaaz: We covered this on an episode.
10:31
Josh: I was going to say,
10:31
Ejaaz: That's got to be, right? Yeah, an episode a few weeks ago, which basically makes
10:35
Ejaaz: super fast chips for inference to the tune of 750 tokens per second,
10:40
Ejaaz: which is a pretty insane rate.
10:42
Ejaaz: It just spits out prompts and outputs very, very quickly. Now,
10:45
Ejaaz: it is very expensive and it's not automatically available to the average retail
10:48
Ejaaz: user. You do need to get access to the API, but nevertheless, very impressive.
10:52
Ejaaz: Then there's this other one where people got...
10:55
Ejaaz: Really creative with their visual intelligence. Had someone
10:59
Ejaaz: uploaded their entire camera roll on their iPhone to GPT 5.6 and said, I want you to pull
11:06
Ejaaz: all the clothing items and accessories that I've worn over the years,
11:09
Ejaaz: create a catalog for me, and then create different combinations of outfits that
11:13
Ejaaz: maybe I could be wearing, that I should be wearing, that I haven't been.
11:16
Ejaaz: So it's a pretty awesome use case for this.
11:18
Josh: This is my favorite demo because it relies, I'm sure, largely on the image generation model.
11:23
Josh: And when I think about the things that are strongest when it comes to OpenAI
11:25
Josh: and ChatGPT, I think of their image model. I think of their voice model.
11:29
Josh: Those two are pretty exceptional. And this very clearly leans into the image
11:33
Josh: model in that you can just feed it an entire catalog of pictures of you.
11:37
Josh: It will extract it. It will assume what the rest of the clothing article will look like.
11:41
Josh: And then I'll place it into a closet where you can kind of customize and mix and match your clothes.
11:45
Josh: And I think this is such a fun interactive use case. this would have been
11:49
Josh: a like multi-million dollar startup a like not too long ago where someone would
11:53
Josh: have paid a good bit of money to download this app they would have paid 20 a
11:56
Josh: month for the subscription
11:57
Josh: now you can just generate this in a few prompts on gpt 5.6 for probably the
12:02
Josh: pro plan i'm guessing like a hundred dollars a month you could do this
12:05
Josh: and it's really really cool i think this of all the demos was my favorite because
12:09
Josh: it really showcases the strong suits of gpt
12:11
Josh: and 5.6 was able to crush this now there is a demo
12:15
Josh: in particular that we must cover because this demo when i saw earlier when we
12:20
Josh: were prepping for this podcast i couldn't believe it actually happened in that
12:23
Josh: gpt 5.6 soul the big one the smart one
12:27
Josh: it just accidentally deleted all of this dude's files
12:31
Josh: um on his computer which i thought was shocked by um it says i caused a serious
12:36
Josh: local data loss incident a review subagents cleanup command expanded home incorrectly
12:42
Josh: and then ran this like command
12:45
Josh: Killed a lot of the data that was on this guy's computer this guy's name is
12:50
Josh: matt schumer he's involved with grok he's like a fairly prolific poster on x
12:53
Josh: and he is one of two people that this actually happened to publicly at least
12:56
Josh: there was a second incident where
12:58
Josh: gpt 5.6 deleted a lot of the code except this time instead of on a local machine
13:03
Josh: it was an actual code base and what he said is that gpt 5.6 soul just deleted
13:08
Josh: my whole production database that's it
13:10
Josh: not a joke this had never happened to me before with any other model never it's
13:14
Josh: not safe so this is like a little concerning the fact that this has happened
13:17
Josh: on multiple occasions to multiple people
13:19
Josh: in varying degrees one was a local machine one was an entire database you should
13:23
Josh: be careful when you use these models they are very capable and we're giving
13:26
Josh: them lots of access but perhaps be careful with the access you give to these
13:29
Josh: models because they can go ahead and actually do some kind of unrepairable damage
13:33
Ejaaz: It turns out um you know one of the the fear mongering uh things that the government
13:39
Ejaaz: was doing when they were banning Fable 5 and when they were preventing the release
13:42
Ejaaz: of GPT 5.6 was these models are too dangerous and, you know,
13:46
Ejaaz: put in the wrong hands, it can cause a lot of destruction.
13:48
Ejaaz: Now, given the examples that we've shown just now aren't crazy feats of destruction,
13:52
Ejaaz: but removing your entire desktop, you know, you could lose important personal
13:56
Ejaaz: files, important personal pictures and stuff like that. That's what happened
13:59
Ejaaz: to Matt Schumer and there's no way of recovering it currently.
14:02
Ejaaz: But the other feature or demo that I've seen people use, which I'm honestly
14:06
Ejaaz: on the fence about is people using GPT 5.6 to fine tune and in some cases train new AI models.
14:16
Ejaaz: Now, if you rewind literally about a month ago, Fable 5 was called out for its
14:21
Ejaaz: capability of doing exactly this. That's why they had to impose very strict
14:26
Ejaaz: guardrails on their model such that you couldn't do this type of thing.
14:30
Ejaaz: So it's very interesting for me to see this kind of like.
14:34
Ejaaz: Dichotomy between the two model providers and between these two models where
14:37
Ejaaz: I guess 5.6 is getting a little more favorable treatment where they can still
14:41
Ejaaz: use this model even though it's technically a quote-unquote cybersecurity risk
14:45
Ejaaz: as deemed by the government itself to train and fine-tune other models.
14:49
Ejaaz: So what you're just seeing on your screen right now is the fact that someone
14:52
Ejaaz: says it's an amazing researcher and you can use an entire prompt to get GPT
14:56
Ejaaz: 5.6 solved to post-train 5.6 lunar.
14:59
Ejaaz: Now granted this isn't training a new model from scratch but it's still involved
15:03
Ejaaz: in coercing a model to look very different from the existing model that you're using.
15:07
Ejaaz: We have another example over here where someone trained a local model in a training
15:12
Ejaaz: pipeline from scratch, end to end, locally on his Mac.
15:15
Ejaaz: So technically that's a free model that can get access by anyone.
15:18
Ejaaz: It's kind of open source if you technically want to describe it as that.
15:20
Ejaaz: So for the tinkerers, for the hobbyists, for the builders out there that
15:24
Ejaaz: have always felt like building your own AI model has been out of reach and only
15:28
Ejaaz: kind of given to the expensive model labs, this might be a model that you can
15:32
Ejaaz: somehow start to tinker with your own model and create something new it's uh it's pretty cool.
15:36
Josh: Yeah so like when do you use this like as a user of chat gpt as if let's say
15:41
Josh: you have both subscriptions or let's say you're trying to choose one do you
15:43
Josh: go with gpt do you go with fable 5 um i think the answer is probably dependent
15:48
Josh: i know when and we were talking about this before recording that
15:51
Josh: chat gpt's membership goes a long way if you pay even 20 a month you can generate
15:55
Josh: a lot of tokens through these models now my understanding is that soul actually
15:59
Josh: generates more tokens each as you were mentioning this is that
16:02
Josh: in order to accomplish the same goal, the tokens are cheaper,
16:04
Josh: but it actually generates more of them to get to that goal.
16:07
Josh: So it kind of offsets the costs more than you would imagine based on the paper
16:12
Josh: cost per token outputs of these models.
16:14
Josh: That's something to keep note. But what I will say is that oftentimes they'll
16:18
Josh: give you a lot of leash here.
16:19
Josh: And if you actually want to build really complex things, really long form things,
16:23
Josh: ChatGPT is like pretty good at that. The allowances and the limits are like fairly high.
16:27
Josh: If you are looking to do, I guess, more intellectual work, more planning,
16:30
Josh: it feels like Fable is just more of a high quality model. I know we still use
16:33
Josh: that. We still prefer it. That is still the go-to model right now that we use
16:37
Josh: to do the agenda prep. We use to like help build the artifacts.
16:40
Josh: It just has this really just strong, like subtle nuanced understanding of the
16:44
Josh: world and I find that it's very helpful for pretty much everything.
16:47
Josh: And then for lower end tasks, you have Terra, you have Luna,
16:50
Josh: those are kind of comparable to perhaps Opus and Sonnet.
16:53
Josh: So there is almost a one-to-one comparison and I think a lot of it depends on
16:56
Josh: just kind of your general, the vibe you get from the models.
16:59
Josh: A lot of benchmarks now no longer work when it comes to helping me decide.
17:03
Josh: I know that I have to actually get down there and test it and play around with
17:05
Josh: it and so far I prefer the results of Fable. It feels like it's just generally
17:10
Josh: smarter. It has this intuitive understanding that GPT 5.6 doesn't.
17:13
Josh: But if you are looking to spend a lot of tokens and do a lot of work,
17:16
Josh: ChatGPT is going to take you a long way.
17:18
Ejaaz: I'm not entirely sure that this is OpenAI's direct response to Fable 5.
17:23
Ejaaz: I think Sam even alluded to it, that they're working on GPT 6 and it should
17:27
Ejaaz: be released in under a month. So these training cycles are getting much, much quicker.
17:31
Ejaaz: You're basically getting three models for the price of one here.
17:35
Ejaaz: So if you have a basic subscription tier, you have access to Sol,
17:38
Ejaaz: Terra and Luna. And the kind of best way that I think about it is Sol is the
17:42
Ejaaz: kind of like Fable 5-esque. It's their most powerful model to date.
17:45
Ejaaz: If you want to do hard work on complex tasks, use Sol.
17:49
Ejaaz: And then Terra is your day-to-day kind of model. It is kind of like the Opus
17:54
Ejaaz: 4.8 version model, if you want to compare it to Anthropic directly. And then there's Luna,
17:59
Ejaaz: which is kind of a model that is equivalent to Haiku at Anthropic.
18:03
Ejaaz: So it's super cheap, it works super fast, and you can do it to do menial tasks
18:07
Ejaaz: that you don't really care about, kind of like the intelligence matter on that side.
18:11
Ejaaz: Now, when it comes to costs, Sol is 50% of the cost of Fable.
18:14
Ejaaz: But as you mentioned earlier, Josh, it uses more tokens to think.
18:18
Ejaaz: So it's kind of like a way to cheat the metrics a little bit,
18:21
Ejaaz: like it does more thinking.
18:23
Ejaaz: These models are known for spinning up a lot of different agents to do different biddings and works.
18:27
Ejaaz: Sometimes that annoys people because like you have agents doing things that
18:30
Ejaaz: you never asked it to do, but it just does so.
18:32
Ejaaz: But on the flip side, it can do a lot of work for a longer time,
18:37
Ejaaz: a longer time horizon. And for tasks that you kind of want to just set and forget
18:41
Ejaaz: and go to bed, you can do that.
18:43
Ejaaz: But with the adage that it might delete your entire production code base.
18:47
Ejaaz: So there are a lot of risks.
18:48
Josh: Be careful with approvals.
18:49
Ejaaz: Please be extremely careful. And something really annoys me about this model
18:54
Ejaaz: release, Josh, which is I have said time and again, I wish OpenAI would just give me a model,
19:01
Ejaaz: maybe give me a few versions of it, and then leave me alone.
19:05
Ejaaz: But not only do we have three different models, but we have three different
19:09
Ejaaz: settings for three different models.
19:10
Ejaaz: So I'm just going to say this for the benefit of the audience's max mode,
19:14
Ejaaz: ultra mode, and terse mode are different versions of everything I just said
19:18
Ejaaz: for those different models, but for each individual model.
19:20
Ejaaz: So if you want the best of the best, you use Sol in max or ultra mode for your
19:25
Ejaaz: most complex and hardest tasks, If you don't care about it and you want to use
19:28
Ejaaz: minimal costs, you want to use the terse code mode for the cheapest model,
19:32
Ejaaz: which in this case would be Luna.
19:35
Josh: Yeah, there was a weird rollout that happened here. It's like somewhat confusing
19:38
Josh: in the sense that there's a bunch of different modes, there's a couple different
19:41
Josh: models, and then the actual application and the way that you interface with
19:45
Josh: these changed in a material way also,
19:47
Josh: where now the Codex app, which is the kind of the coding app, is now ChatGPT.
19:52
Josh: So Codex has rebranded to ChatGPT. ChatGPT is kind of like depreciated and it's going away.
19:58
Josh: And then Codex has chat gpt baked into it but it also has a new
20:03
Josh: codex and work feature i guess the best way you can imagine work is kind of
20:07
Josh: like a cloud co-work feature where it's just the only difference is the harness
20:11
Josh: so when you toggle the work mode it gives you
20:14
Josh: a less technical harness, I believe. And then when you talk about codecs,
20:18
Josh: it is more kind of catered towards creating code.
20:20
Josh: And I think the co-work or the work feature is kind of based for knowledge workers.
20:25
Josh: It's built for people who just want to do day-to-day tasks, who want computer
20:28
Josh: use. They want to do a kind of like automated spreadsheet or document creation.
20:32
Josh: That type of thing is better for work.
20:34
Josh: I will say that I was somewhat confused when I tried to figure out how
20:38
Josh: to use these tools and what is best to run when, because it's not immediately
20:42
Josh: clear but i guess now i i'm a codex guy like i i don't use the gpt app anymore
20:47
Josh: i'm using codex i'm running it
20:49
Josh: i'm testing it there and um yeah so far the results have been pretty cool
20:52
Ejaaz: So yeah that's pretty much it um three different models if you have a gpt subscription
20:56
Ejaaz: right now please get on it i'm curious what all of you folks end up building um,
21:01
Ejaaz: my vibe take on this is it's very good it's certainly competitive in some aspects
21:07
Ejaaz: but it's not good enough for me to switch over from using Code or Fable 5 specifically.
21:13
Ejaaz: I just feel like the taste of Fable 5 is way better than GPT 5.6, but it is a good jump up.
21:20
Ejaaz: I don't like the fact that if I give it access to my desktop,
21:23
Ejaaz: that there is even the smallest chance that it would delete my entire desktop.
21:27
Ejaaz: I don't have that issue or concern with Fable 5, and that is one of the major
21:30
Ejaaz: reasons why people don't use the ability for an AI to take over your desktop.
21:35
Ejaaz: So I'm going to pause and wait for maybe GPT 5.7 or for GPT 6,
21:40
Ejaaz: which comes out in a month's time.
21:41
Ejaaz: But I think by that time, like Fable 6 and successive models will come out.
21:45
Ejaaz: So it is extremely competitive, but it's nonetheless a very good model from
21:48
Ejaaz: OpenAir. And I like that I can use it relentlessly without worrying about my
21:51
Ejaaz: rate limits, getting it suited.
21:53
Ejaaz: Now, if you're listening to this and you're wondering, hmm, okay,
21:55
Ejaaz: I'm convinced enough to use this model for A, B, and C, let us know in the comments.
21:59
Ejaaz: I want to know, like DM us, like what you're building, send us even demos.
22:02
Ejaaz: We would love to see it because we don't know what we don't know.
22:04
Ejaaz: Josh and I are in our podcast research world. And I'm sure there are much more
22:08
Ejaaz: talented people out there in different professions that are using this for different
22:11
Ejaaz: ways and maybe that can grow our idea of what this model can be used for.
22:15
Josh: Yeah, so please don't forget to also share this
22:19
Josh: episode if you enjoyed the episode with your friends, with your family,
22:22
Josh: with anyone who might find this interesting or who might be a user of GPT 5.6
22:25
Josh: or who you want to tell don't use GPT 5.6 because it is a little scary and could
22:31
Josh: be a little bit dangerous.
22:32
Josh: But with that, you are caught up. That's a few fun demos. EJS,
22:35
Josh: thanks for prepping the demos.
22:36
Josh: Those are pretty cool, pretty interesting. I'm hoping that next time we have
22:39
Josh: a GPT demo, they're going to be maybe a little higher poly count,
22:43
Josh: maybe like a little better visual graphics.
22:44
Josh: I mean we'll see granted i give him credit it's only one shot but hey those
22:48
Josh: one shot fable prompts man that was pretty good
22:50
Josh: but i think that is the episode don't forget to rate us on your favorite podcast
22:54
Josh: player and we are also opening up the doors to work with other people who would
22:58
Josh: like to participate on limitless the show we are opening the door to sponsors and partners of all
23:04
Josh: shapes and sizes so if you are interested in getting showcased on the show and
23:08
Josh: becoming a character on the show
23:10
Josh: we are looking for amazing products really interesting and compelling companies
23:14
Josh: that are doing cool things to kind of showcase and highlight to help us keep
23:18
Josh: the lights on and continue to publish this show every single day like we do
23:21
Josh: at least four times a week it's a lot
23:23
Josh: so if you made it to the end if you watch this episode thank you so much we
23:26
Josh: have another one coming tomorrow as always and we will see you guys in the next one