Full transcript
0:00All right, you guys. It's time to talk
0:01about another evaluation of a different
0:04large language model. In this case,
0:06we're talking about Claude Opus 4.6.
0:09We already did a video about 4.5 and it
0:11was a very, very impressive looking
0:13model. Uh but they've given us an update
0:17and although it seems like all the
0:19language around this update has been
0:21around improving Claude for coding which
0:25I from what I understand the different
0:27AI coders out there claude is one of the
0:30best models for coding and so this seems
0:32to be optimized for that. Nevertheless,
0:35I have noticed that with all of these
0:37updates, even if they don't really talk
0:38about it in terms of creative writing
0:40because that's definitely a lower
0:41priority for them when compared to
0:43something like coding. At the very same
0:46time, I have noticed that there are
0:48often incremental improvements to the
0:51pros and other, you know, brainstorming
0:54aspects, things that we like as creative
0:56writers. And so, we're going to take a
0:58look at it today and see how well it
1:00performs. Now, first before we get into
1:02the actual results, let's talk about the
1:04specs of this model. And there are some
1:06really interesting things here. First of
1:08all, let's talk price. The base price is
1:10$5 per uh million tokens of input and
1:13$25 per million tokens of output, which
1:16if I'm not mistaken is the same price as
1:18Claude Opus 4.5, which thankfully was a
1:23lot cheaper than Claude Opus 4, uh,
1:27which I believe was three times as much.
1:29So, this still has that relatively
1:31cheaper price, at least cheaper for a
1:35huge mega model like Opus is. It's still
1:38a little bit of expensive, especially
1:41like it's still more than double the
1:42price of Gemini 3, which is also an
1:44excellent model. But at the same time,
1:47uh it is nice to see this having come
1:49down to a third of the price of what it
1:52used to be. But the next big important
1:54thing is that it has a context window of
1:581 million tokens. This means you could
2:01put the entire Harry Potter series in
2:03there and theoretically it should be
2:05able to handle that. Now, should you do
2:08that? I would not do that because that
2:10means that every chat will call you cost
2:12you $5 for the million tokens of input
2:14and that's not even starting on the $25
2:17per million tokens of output. So, I do
2:20think there is really good reason to
2:22avoid huge contexts, at least huge in
2:26terms of like up to a million, right?
2:28This isn't really something that we need
2:30as creative writers. You can get a
2:33summary of your book that will actually
2:36like think of it as as like, forget the
2:39metaphor, like juice in concentrate,
2:42right? Uh, I will often run my books
2:44through a prompt that will condense them
2:47but point out all of the important
2:48details in each chapter. Uh, so that the
2:52amount of context that I put into a book
2:55like uh let's say I've condensed my
2:57first book in this way and I'm using it
3:00to help me plot the events of the second
3:03book, right? It's better to have that
3:06condensed version than to give it the
3:07entire book because giving it the entire
3:10book kind of waters down the prompts a
3:11little bit. It also increases the price
3:13a little bit and we just want it to be
3:16able to understand the details of the
3:18first book so it can properly map out
3:21what the second book could look like for
3:23your approval. And because of that, I
3:26actually think the author should keep
3:28their context windows under 50,000.
3:31That's usually what I go for. So
3:33certainly a million is more than enough,
3:36but I really feel like that's more for
3:38like people creating these large coding
3:41complexes where you want the AI to be
3:44able to understand the entirety of the
3:46software all at once to be able to debug
3:47it and stuff like that. I think it's
3:49better for that sort of use case uh
3:52rather than for a writing use case. For
3:54writers, try to keep it under 50,000 in
3:56ter in terms of your total context. And
3:59that way, you won't have a problem with
4:02bloated prompting or the needle in a
4:03haststack problem or any of that. All
4:05right, so that was a little bit of a
4:07tangent there, but basically we get all
4:09of that stuff plus it can handle up to
4:11128k tokens of max output. And I will
4:15tell you in my test that I ran it uh it
4:18got very high and it's capable in what
4:22it's capable of outputting. So uh let's
4:25go take a look now at the results. So I
4:27ran it through my standard creative
4:29writing prompts that I run every model
4:32through. Starting with the log line
4:34prompt which asks it to generate 20
4:36fantasy log lines. And I was actually
4:38pretty impressed with a lot of these.
4:39Uh, typically like I've run this prompt
4:42so many times now that I typically get
4:45very similar stuff all the time and I
4:49still got a number of things that I get
4:52all the time. But at the same time,
4:53there were some really intriguing story
4:55ideas that I looked at. A good example
4:58of both of that both of those things is
5:01here in this first one. A cowardly royal
5:03cgrapher. So I get cgrapher a lot in
5:06these things. I imagine it's not a very
5:10original idea uh because it keeps coming
5:12up all the time. Who has never left the
5:14palace walls must cross the very
5:16uncharted wastelands he's been
5:17fabricating on maps for years when the
5:19king dispatches him to a broker to
5:21broker peace with a nation everyone
5:23believed was fictional only to discover
5:25his imaginary maps are somehow perfectly
5:27accurate. Quick pause. If you're an
5:28author who wants to start writing more
5:30books, I put together my complete AI
5:32prompt pack that helped me and many
5:34others in my community write a book in a
5:36month. It's completely free. Hit the
5:37link in the description. Join the free
5:39weight list for StoryHacker Gold. Then
5:41you just answer a couple quick questions
5:43and I'll send you the full prompt list
5:44plus a few extra free goodies. I'll also
5:47email you when I open the paid Story
5:49Hacker Gold group, so you can join if it
5:51feels right. There's no commitment to
5:53join, though, and you'll still be able
5:54to get all of the freebies you signed up
5:56for either way. All right, back to the
5:57video. And so, I really like this last
6:00part, the idea of imaginary maps
6:02becoming accurate. There's definitely a
6:05story in there. I do think first of all
6:07the cgrapher thing is a little bit
6:09overdone and there are some things that
6:13here that just don't make that much
6:15sense. Like why is he fabricating maps?
6:17Like just things like like that that
6:19don't really make sense. You'd really
6:21have to tie yourself into a few knots to
6:23make that work as a story. But there's
6:25definitely a kernel of a good idea here
6:27with these imaginary maps somehow being
6:29perfectly accurate. That's a really
6:31interesting, intriguing story-building
6:33mystery for me. Um, but this one here, I
6:36actually really, really liked. An
6:38immortal executioner cursed to carry the
6:39memories of everyone he's killed
6:41discovers that one of his thousands of
6:43victims holds the secret to destroying
6:45the tyrannical god king he serves. But
6:48accessing that memory means becoming the
6:50very person who once tried to
6:51assassinate him. So this I thought was
6:53interesting like someone having the
6:55memories of people that he kills and
6:57then what that implies, you know, like
7:00what you could learn like what what this
7:02person probably has a lot of skills and
7:04things like that. Like there's
7:05definitely some kernels of of good
7:07stories in that. So uh all in all, I
7:10thought these were actually pretty good
7:11and some of the better log lines that
7:13I've seen so far. So if you want
7:15something that kind of thinks a little
7:17bit more creatively and can give you
7:19some really good story hooks, Claude
7:21Opus 4.6 six might be one to take a look
7:24at. All right, so the next one was the
7:26outlining test prompt. And this one,
7:29woo. Uh, let me just first of all show
7:32you how many words it wrote. So if we go
7:36ahead and and select the entire outline,
7:40it wrote
7:42almost 17,000 words in one go, which
7:45might be a record for me. I I don't know
7:47if I've ever seen or at least seen and
7:49measured that many words from a model
7:53before. And I have noticed this about
7:55the cloud models in general. One of the
7:56things that they're really good at is
7:58not stopping when you don't want them to
8:00stop. A lot of the other models will
8:02kind of like get to a certain point and
8:05then they sort of crap out. They're just
8:06like, "Okay, I'm done." Or they'll start
8:09like if it's writing an outline, the
8:10outline for each chapter will get
8:12shorter and shorter and shorter as it
8:13goes along. Even Gemini 3 does this. But
8:16the claw models don't do that. If they
8:17decide, oh, we're going to write 500
8:19words for every chapter here, it will
8:22freaking do it for the whole chapter and
8:24it won't hold back and it will actually
8:25give you pretty lengthy output. That's
8:27one of the strengths of the cloud models
8:28that I found. So, this is no exception.
8:31Um, and when I, you know, I didn't read
8:34through the whole 17,000 words because
8:36that would just be a lot. Uh, this is
8:38why I don't get as granular as I used to
8:42with these things because I would have
8:43to read through like 20,000 words of
8:46prompt response for every single model
8:48that I that I evaluate and I just can't
8:51do that. But from the first few chapters
8:53that I went through, it looked like a
8:55really solid opening to the story. Um, I
8:59even thought like very often it gives me
9:02in the first chapter I ask it to give it
9:03a give us a save the cat moment and
9:06often it's really shoehorned in and this
9:09time I felt like it actually made a
9:11little bit of sense in the context of
9:13the story. Uh, so that's one of the
9:15things that I look at. Um, and yeah, it
9:18did kind of go like I only give it a
9:20kernel of a story to go on. I don't
9:23instruct it too much just to see how
9:25well it's able to come up with a lot of
9:27the stuff on its own. And as a result,
9:30there is always a certain amount of
9:33predictable stuff that it comes up with.
9:36Any of the models that I've tested
9:38usually are pretty predictable in a lot
9:41of the ways that it goes. But at the
9:43same time, I thought this one was it at
9:47least sort of fit and made sense, even
9:49though it did sometimes go into those
9:51more predictable things. And at the very
9:52least, uh, if we look at like here's
9:55chapter 2, we've got o over a page of
9:59detail for every chapter. Now, some
10:01people, myself included, might not like
10:03that much detail in the outlines. Uh, I
10:06usually like like I would be okay with
10:09this much, you know, uh, telling me
10:12what's in the chapter, but some people
10:14really like detailed outlines. And
10:16certainly you could create a um you know
10:20having that much detail will help if
10:22you're using AI to help with the pros
10:24generation later. At the same time I so
10:26ba basically what I'm saying with that
10:28is that it has lots of detail. The
10:30detail makes sense for the most part as
10:33from what I was able to tell it it makes
10:34sense for the chapter. And so I think
10:36overall this is actually pretty good. Um
10:39I have no issues with this. All right,
10:42so let's move on to the next one, which
10:43is the basic pros prompt, where I just
10:46give it a scene and don't give it much
10:48else. Just give it some basic
10:50descriptors of what happens in the scene
10:51and ask it to write. I don't give it
10:53style information. I don't give it any
10:55of that. We I just want to see how well
10:57it does kind of with the vanilla version
11:00of the model. And uh overall, I think it
11:04did an okay job here. So, usually I like
11:06to come down about halfway through the
11:08segment here and read a little bit.
11:11He was the last of a company of 14. He
11:13did not let himself think about the
11:14other 13. Haldra's breath came in sharp,
11:17shallow gas that tasted of copper and
11:18dust. And when the wind shifted, when it
11:20swung around from the east and pushed
11:22against his back instead of his chest,
11:24he made the mistake of looking over his
11:25shoulder. There they were. There
11:28scattered across the darkening horizon
11:30like stones come loose from the earth
11:31itself. Gray shapes moved against the
11:34gray land. At first glance, they might
11:36have seen nothing. Tricks of the fading
11:38light, outcrops of of rock shifting in
11:40the heat haze that still clung to the
11:41ground. But rock did not move like that.
11:44Nothing li living moved like that. They
11:46came in stuttering, broken surges,
11:48bodies jerking forward in spasms that
11:50covered ground with a horrible deceptive
11:52speed. Limbs snapping into motion as if
11:54yanked by invisible strings, then
11:56freezing, then surging again. Two dozen
11:57at least spread in a loose and
11:59tightening arc across the waist behind
12:01him. the hungry. Haldrid's stomach
12:03clenched into a cold fist. They were
12:05close, far too close. Close enough that
12:07he could make out the way their heads
12:08lulled and twitched between those awful
12:10bursts of movement. Close enough that by
12:12the time you reach the gates, they would
12:14be within bowshot of the walls. All
12:16right, so this is pretty decent pros as
12:19a whole, and it's really appropriate for
12:21moments like this in the story that are
12:23kind of serious, kind of, you know,
12:25there's a there's a lot of importance in
12:27this moment.
12:28However, I found that there is a certain
12:31serious quality to the pros that Opus
12:34creates even in less serious moments.
12:36And I've come, having worked with Opus
12:38quite a bit, I've come to recognize the
12:41way it talks a lot to the point where I
12:44can usually pick it out. Even if I don't
12:46know if Pros was written by Claude Opus
12:49specifically, I can usually tell if it
12:51has been because it always has this sort
12:53of almost depressive quality to a lot of
12:56what it it says. Like there's just and
12:58and it spends a lot of time. You notice
13:01this this paragraph that we had right
13:02here, it spent a whole lot of time
13:04describing things. That's not
13:06necessarily a bad thing in the right
13:09contexts,
13:11but in like like this, this is an
13:14appropriate moment for something like
13:15this, but having that level of depth in
13:18the pros is often not needed, if that
13:22makes sense. And so, um, I I do think
13:25that sometimes Claude is a little bit
13:26wordy, overd does it just a little bit.
13:29Uh, one of the reasons I like Gemini 3
13:31is that it is not as wordy. Uh, but I
13:34think a lot of people prefer this pros
13:37because it's so rich and everything. I
13:39just find it gets a little old after a
13:41while. So, I see nothing about the pros
13:44here so far. Uh, we'll check the other
13:47pros prompts. I see nothing here that
13:49really seems different from Opus 4.5 in
13:54that sense of it making kind of natural
13:57sounding pros. But let's go to the next
13:59one. Uh, the next pros prompt, I give it
14:01the exact same scene, but I give it a
14:02whole lot more information, including
14:04style guides, uh, samples of my writing
14:07and stuff like that, and try to see if
14:09it makes a a good scene. It's about the
14:13same number of words, so that's good.
14:15Often, this one will be shorter from a
14:17lot of the prompts that or a lot of the
14:19models that I've tested it with. Uh, but
14:22if we go to that same moment in the
14:25story where it says, "He looked back.
14:27They speckled the ridge he had crossed
14:29an hour ago. Figures, dozens of them,
14:30scattered across the rock like a disease
14:32spreading through skin. Uh, their
14:34movement was wrong. Shoulders lurched
14:36forward before legs caught up. Heads
14:38snapping to one side then the other, as
14:39if their neck couldn't decide which way
14:41to face. Some crawled on all fours with
14:44a speed that defied broken geometry of
14:45their limbs. Gray flesh blended with
14:47greystone, and only the motion gave them
14:49away. He counted, stopped counting at 24
14:51because his hands were shaking too hard
14:52to keep track. Too close. They were far
14:54too close. Haldrid turned back toward
14:56the city walls, and a sound escaped his
14:58throat. Half laugh, half something else
15:00entirely. He knew the protocol. Every
15:02citizen of Valdin knew it. Drilled into
15:03them from childhood the way they had
15:05learned to count and tie knots. When the
15:07[snorts] hungry approached within a
15:08certain radius of the gates, nothing
15:09opened. Not for the wagon of trade
15:11goods, not for a company of armed
15:12soldiers, not for one ragtag cgrapher
15:15stumbling out of the waist with
15:16information that might matter to
15:17someone. If anyone cared to listen
15:19before the hungry caught his scent on
15:20the wind and closed that gap with her
15:21horrible study earing gate. very similar
15:24prompt. I think um overall I would say
15:26this is about the same uh but maybe a
15:29little bit better in certain instances.
15:31I I do think that the pros is still
15:32pretty good. But let's move on to the
15:34dialogue prompt because um you know that
15:38scene that we just looked at had no
15:40dialogue in it. So I have another scene
15:42just to see how well it does with the
15:43dialogue. Claude is typically usually
15:45pretty good at dialogue. So let's see
15:47what it says. Lean did not care how many
15:49storms his father had weathered. A man
15:50died at the gates tonight. he said. His
15:52boots rang against the stone floor as he
15:53crossed the hall. Two of the king's
15:55guards shifted at his approach, not
15:57threatening, not yet, but attentive in
15:58the way of men trained to read the
16:00weight of a footstep. Lean ignored them.
16:02His blood was still hot from the wall,
16:03his wrist still aching with the guards
16:05had gripped him, and the sound of that
16:06man screaming not yet reverberated in
16:08the hollow of his chest. Had not yet
16:10stopped reverberating in the hollow of
16:11his chest. I know, said Store, king of
16:14Araldon. You know, Lean stopped at the
16:16base of the deis, his jaw tight. Then
16:18you know the gates stayed shut. You
16:20know, we watched him die. That's the
16:22second time in a month, father. The
16:24second time we've left someone out there
16:25to be taken by the hungry, while we
16:26stood behind our walls and did nothing.
16:29The word cracked off the vaulted
16:30ceiling. Somewhere in the recesses of
16:31the hall, a scrib's quill stopped
16:33scratching. Store regarded his son for a
16:36long moment. The fire light shifted, and
16:38the shadows of the lions of his face
16:39deepened. When he spoke, his voice
16:41carried the low, unyielding weight of a
16:43man who had made this argument before
16:45with himself, if not with others. The
16:47law is the law, Lean. The law is your
16:49law. Lean stepped forward one boot on
16:51the lowest stair of the day. You made
16:53it, you can unmake it. You could have
16:55ordered those gates open tonight and
16:56that man would be alive. Instead of the
16:58law, came a a voice from the right of
17:01the throne, smooth as oiled leather.
17:03Exists for a reason, my prince. Lean's
17:06teeth clenched. He did not need to look
17:08to know the voice. Argal emo Argle
17:12emerged from beside the throne like a
17:14shadow, remembering it had a body, tall,
17:16narrow, his hands folded within the
17:18sleeves of his dark rose, his thin mouth
17:20arranged into an expression that might
17:22have been mistaken for sympathy by
17:24someone who had never met him. The
17:25king's chief adviser inclined his head,
17:27a gesture so measured it bordered on
17:29mockery. The protocols governing the
17:31gates after nightfall were established
17:33to protect every soul within Valden.
17:35Argal continued his voice carrying the
17:37practice patience. There's an AI phrase
17:40right there. Practice patience or
17:42practiced ease, practiced anything.
17:44That's a pretty uh AI use of the word of
17:48a tutor explaining arithmetic to a slow
17:50child. One life weighed against
17:52thousands is not a calculation that
17:53requires deliberation. It requires
17:55discipline. Discipline. The word tasted
17:57like ash. Anyway, um pretty standard. I
18:00I I don't really again see any huge
18:03increase over 4.5.
18:06Still pretty good. Uh, still not bad
18:08overall. Still obviously has a number of
18:12AI isms and and terms of phrase and and
18:15different things like that that we would
18:18want to edit out. Um, looking at the
18:21next one where I give it the same scene
18:23that was written by a very unpowerful
18:25model and ask it to just improve the
18:27writing to see if it can do that really
18:28well. Um, it gives us this. A man died
18:32last night. Lean stopped in the foot of
18:34the day close enough to see the torch
18:35light pulling in the hollows beneath his
18:36father's eyes. He stood before our walls
18:38and begged for mercy for shelter for the
18:40bare scrap of human decency and we gave
18:42him nothing. We watched him be torn
18:44apart. The silence that followed was a
18:45living thing colliding through the hall.
18:47Storm clied his head slow as a glacier
18:49carving. The law was upheld. The gates
18:51remain shut but it was within your power
18:52to break that law. All right. So looking
18:55at this, it is improved on the original
18:57text but not by much. And there's
19:00actually a fair amount of text in here
19:03that I would say is like you're better
19:06off just rewriting the scene honestly. U
19:09so and that's been my conclusion for a
19:12lot of these. The only one that I think
19:14really made it a lot better was Gemini
19:163. Um, but you do get better results out
19:19of an AI model from compared to what it
19:22originally wrote, but it's still like
19:25not great uh when it's being derived
19:28from a poorly written text. You're
19:31better off just rewriting it. All right,
19:33so moving on to the next one, which is
19:34the add headlines prompt. Uh, AI can
19:37actually be really good at writing
19:39headlines, but only if you're giving it
19:42some like templates to follow, right?
19:45And so for this prompt, I actually give
19:47it no templates just to see how well it
19:50does out of the box, right? Um, first
19:54glance, these are looking a little bit
19:56long. Uh, but let's read a couple of
19:58them. He was told the world outside the
20:00mountain was dead. He was lied to about
20:02everything. Not too bad. Uh, you could
20:05probably shorten that. A prince thrown
20:08out to die discovers the kingdom he
20:09loved was built on a lie. Three races, a
20:12zombie plague, one exiled prince, and a
20:14truth that changes everything he thought
20:15he knew. They told him nothing, survived
20:17outside the mountain. So why are there
20:18rebels who've been living out there for
20:20years? Um, that's not a bad one. What's
20:24worse than being betrayed by your own
20:25family? Finding out they weren't who you
20:27thought they were? I mean, I feel like
20:29that's the same thing, right? He was
20:31raised to fear what lives outside the
20:33mountain. Now he's starting to wonder if
20:34the real monsters were inside it. That
20:36one's not too bad at all. Uh, so some
20:38good ones here, some bad. I do think all
20:40of these could probably be made more
20:42succinct, which is always almost always
20:44a good thing when it comes to ads. Um
20:47because because of attention spans, you
20:49know, but overall, that's not bad. Not
20:52bad at all. All right. Um so, moving on
20:55to the email newsletter prompt where I
20:58ask it to essentially write an
21:00announcement email for this book that
21:01we've been uh fictitiously outlining and
21:05writing pros for. And uh overall it it
21:09goes like this. Hey there. I'm thrilled
21:11to finally share the news. My epic
21:12fantasy novel, The Crown of Hunger,
21:14releases on December 21st, and I'm
21:16honestly can't wait for you to read it.
21:17This one has been a labor of love.
21:19Imagine the Count of Monti Cristo
21:21reimagined in a brutal, unforgiving
21:22fantasy world where humanity clings to
21:24survival inside a mountain fortress and
21:26everything outside wants to consume you.
21:28Here's what you're stepping into.
21:31That gives us a bullet point of some of
21:32the key features. This is a story about
21:35betrayal, survival, and the moment you
21:36realize the world is nothing like you
21:38were told. The real monsters aren't
21:39always the ones outside the walls.
21:41Pre-order your copy now. Link to the
21:42book here. And then it gives us a little
21:45bit more pros that probably is not
21:47necessary here. I really value
21:50succinctness and brevity when it comes
21:52to email newsletter writing. So, I think
21:55this is probably a little bit much, but
21:57uh especially this intro and the bullet
21:59points aren't too bad. Uh, but I do
22:02think it kind of rattles on a little bit
22:04longer than I'd like. Uh, I think I
22:06prefer Gemini 3 when it comes to this
22:08particular prompt. All right, last but
22:10not least is the SEO article prompt. I
22:12asked it to write 4,000 words. Let's see
22:13how how many it did.
22:16And it wrote 4,500. So, pretty good
22:19actually. So, I can pretty rest assured
22:21that it goes into full detail. Sometimes
22:25the models will try to write an article
22:28based on my outline, but really all it
22:30does is end up, you know, in 2,000
22:33words, there's not enough time to really
22:34flesh out the full outline that I give
22:36it. And so what ends up happening is it
22:39gets a slightly more fleshed out
22:42outline. Uh, but with this, you know,
22:45looking through it, okay, we got what is
22:48a fantasy novel? Uh, this is, by the
22:50way, an article about how to write a
22:51fantasy novel. It gives us an
22:54explanation of what it is. Um, here's
22:56some like different other types of
22:58fantasy, common tropes in fantasy
23:00novels, and it gives us a bulleted list
23:02of that steps to write a fantasy novel.
23:05This is this was all given in my
23:07outline. U, but looking through it, I
23:09can say, "Yeah, this is really well
23:10fleshed out. It's got plenty of content
23:12for each section and definitely goes
23:15through and gives me all of the things
23:17that I would be interested in." So, uh,
23:20pretty great job for Opus, uh, 4.6 on on
23:24this particular one. And I have found
23:25that the Claude models are probably the
23:27best ones for any kind of article
23:30writing. Uh, I personally use um, Claude
23:35Sonnet for a lot of those kind of
23:37things. All right, so that does it for
23:40the prompts here. My overall impressions
23:43of Claude Opus 4.6 is that it is not
23:47that much better. at all than 4.5, but
23:51that's not necessarily a bad thing. Um,
23:54it's you have to work with these things
23:57a lot more to really understand the
23:59subtle nuances of each model. But, uh,
24:02just as my first impressions, it's
24:04certainly not worse than 4.5.
24:07And I would be happy to just upgrade any
24:10of my automations or any of parts of my
24:12process where I used Opus 4.5. I'd be
24:16happy to upgrade it to 4.6 without
24:18really any fear of it getting worse or
24:21anything like that. Definitely something
24:23that's worth using in the creative
24:24spaces. All right, so those are my
24:26thoughts on Opus 4.6 six and I will see
24:29you in the in the next