How to Import Sessions from Claude to Hermes (Even in Telegram)

1
00:00:00,600 --> 00:00:05,200
[แปลไทย 1] So, I asked two models to create a mini

2
00:00:03,000 --> 00:00:07,800
[แปลไทย 2] game where I fly around in a jet

3
00:00:05,200 --> 00:00:09,400
[แปลไทย 3] collecting orbs and dodging enemies. And

4
00:00:07,800 --> 00:00:12,500
[แปลไทย 4] you can see the two options that we got

5
00:00:09,400 --> 00:00:14,200
[แปลไทย 5] here. On the top, we have a very kind of

6
00:00:12,500 --> 00:00:17,100
[แปลไทย 6] dynamic

7
00:00:14,200 --> 00:00:19,400
[แปลไทย 7] game. You have a dash effect. Uh, lots

8
00:00:17,100 --> 00:00:22,000
[แปลไทย 8] of different lighting elements and very

9
00:00:19,400 --> 00:00:23,400
[แปลไทย 9] 3D. The camera also moves quite a bit.

10
00:00:22,000 --> 00:00:25,800
[แปลไทย 10] The bottom is a bit more of a simple

11
00:00:23,400 --> 00:00:27,800
[แปลไทย 11] take on it. It did achieve the prompt.

12
00:00:25,800 --> 00:00:30,700
[แปลไทย 12] It is a jet that I'm flying around

13
00:00:27,800 --> 00:00:33,200
[แปลไทย 13] getting energy elements. There is an orb

14
00:00:30,700 --> 00:00:36,900
[แปลไทย 14] in the middle, but pretty clear that the

15
00:00:33,200 --> 00:00:40,000
[แปลไทย 15] top one uh did a better visual job. So,

16
00:00:36,900 --> 00:00:44,400
[แปลไทย 16] as you can tell from the title, the top

17
00:00:40,000 --> 00:00:47,900
[แปลไทย 17] model is Opus 5.5 and the bottom is GBT6

18
00:00:44,400 --> 00:00:50,200
[แปลไทย 18] Soul. So, this morning I woke up to find

19
00:00:47,900 --> 00:00:53,200
[แปลไทย 19] this new model from Claude. Claude Opus

20
00:00:50,200 --> 00:00:56,800
[แปลไทย 20] 5.5 dropped today. So they skipped

21
00:00:53,200 --> 00:00:58,400
[แปลไทย 21] everything from five to 5.5

22
00:00:56,800 --> 00:01:01,000
[แปลไทย 22] and they say it performs to the levels

23
00:00:58,400 --> 00:01:03,900
[แปลไทย 23] of Claude Fable 5.1 for most tasks and

24
00:01:01,000 --> 00:01:06,200
[แปลไทย 24] costs 40% less to run than Opus 5, which

25
00:01:03,900 --> 00:01:08,600
[แปลไทย 25] would be great news if it's true. Um I

26
00:01:06,200 --> 00:01:11,200
[แปลไทย 26] have a Claw subscription, only the $100

27
00:01:08,600 --> 00:01:13,400
[แปลไทย 27] plan and I usually use it for Fable. I

28
00:01:11,200 --> 00:01:15,500
[แปลไทย 28] still feel like Fable is the best model,

29
00:01:13,400 --> 00:01:17,300
[แปลไทย 29] but I run out of it, right? I run out of

30
00:01:15,500 --> 00:01:20,200
[แปลไทย 30] my Fable usage and I still have over

31
00:01:17,300 --> 00:01:24,400
[แปลไทย 31] half of my weekly usage to go. So I have

32
00:01:20,200 --> 00:01:27,000
[แปลไทย 32] to use Opus and I have been using Opus 5

33
00:01:24,400 --> 00:01:29,300
[แปลไทย 33] and it is as bad as everyone has said.

34
00:01:27,000 --> 00:01:33,000
[แปลไทย 34] It is really step down obviously from

35
00:01:29,300 --> 00:01:35,800
[แปลไทย 35] Fable and using it in some of my more

36
00:01:33,000 --> 00:01:38,200
[แปลไทย 36] complex code bases, more advanced

37
00:01:35,800 --> 00:01:40,900
[แปลไทย 37] projects is really really difficult

38
00:01:38,200 --> 00:01:42,600
[แปลไทย 38] because it will often introduce mistakes

39
00:01:40,900 --> 00:01:45,400
[แปลไทย 39] that were not there previously and then

40
00:01:42,600 --> 00:01:47,900
[แปลไทย 40] I need to get another model to fix them.

41
00:01:45,400 --> 00:01:49,800
[แปลไทย 41] So I'm very hopeful that this is

42
00:01:47,900 --> 00:01:52,200
[แปลไทย 42] actually going to work.

43
00:01:49,800 --> 00:01:54,100
[แปลไทย 43] Now, I saw this and I was going to test

44
00:01:52,200 --> 00:01:56,300
[แปลไทย 44] it out myself, maybe do a first look

45
00:01:54,100 --> 00:01:59,000
[แปลไทย 45] video, but then just a couple hours

46
00:01:56,300 --> 00:02:02,100
[แปลไทย 46] later, I saw this announcement. Uh,

47
00:01:59,000 --> 00:02:05,100
[แปลไทย 47] OpenAI just announced on the same day, a

48
00:02:02,100 --> 00:02:07,700
[แปลไทย 48] couple hours later, that GPT6 Soul and

49
00:02:05,100 --> 00:02:10,800
[แปลไทย 49] Luna are now going to be added alongside

50
00:02:07,700 --> 00:02:13,200
[แปลไทย 50] Astra. and same issue that I faced using

51
00:02:10,800 --> 00:02:14,600
[แปลไทย 51] my uh codec subscription even more

52
00:02:13,200 --> 00:02:17,000
[แปลไทย 52] really because I feel like Astro takes

53
00:02:14,600 --> 00:02:19,800
[แปลไทย 53] up more of my usage and then I end up

54
00:02:17,000 --> 00:02:22,800
[แปลไทย 54] having to use GBT soul anyway. So if

55
00:02:19,800 --> 00:02:25,300
[แปลไทย 55] they have a GBT6 soul that is better

56
00:02:22,800 --> 00:02:27,300
[แปลไทย 56] than the 5.6 version uh that would be

57
00:02:25,300 --> 00:02:29,300
[แปลไทย 57] great. So, in this video, I figured

58
00:02:27,300 --> 00:02:32,600
[แปลไทย 58] instead of doing two separate uh first

59
00:02:29,300 --> 00:02:37,300
[แปลไทย 59] look videos, I would do one head-to-head

60
00:02:32,600 --> 00:02:40,200
[แปลไทย 60] with Clawude Opus 5.5 versus Open AAI's

61
00:02:37,300 --> 00:02:42,200
[แปลไทย 61] GBT6 Soul

62
00:02:40,200 --> 00:02:44,300
[แปลไทย 62] because if these two models are really

63
00:02:42,200 --> 00:02:46,600
[แปลไทย 63] good, that'll really actually be kind of

64
00:02:44,300 --> 00:02:48,900
[แปลไทย 64] an upgrade beyond just having like if

65
00:02:46,600 --> 00:02:51,500
[แปลไทย 65] they introduce another better version of

66
00:02:48,900 --> 00:02:53,800
[แปลไทย 66] Astra, having two really capable models

67
00:02:51,500 --> 00:02:56,600
[แปลไทย 67] that I can actually use for long periods

68
00:02:53,800 --> 00:02:58,300
[แปลไทย 68] like Soul and like Opus. is really

69
00:02:56,600 --> 00:03:00,900
[แปลไทย 69] useful just dayto-day more than just

70
00:02:58,300 --> 00:03:03,600
[แปลไทย 70] having a even more powerful fable or

71
00:03:00,900 --> 00:03:05,000
[แปลไทย 71] more powerful Astra. So, let's get

72
00:03:03,600 --> 00:03:06,900
[แปลไทย 72] started. And if you're looking to level

73
00:03:05,000 --> 00:03:09,100
[แปลไทย 73] up your own agents, check out my project

74
00:03:06,900 --> 00:03:10,700
[แปลไทย 74] agent wikis. Here you get access to

75
00:03:09,100 --> 00:03:12,400
[แปลไทย 75] knowledge races I use every day for

76
00:03:10,700 --> 00:03:14,600
[แปลไทย 76] researching these videos and building

77
00:03:12,400 --> 00:03:16,600
[แปลไทย 77] projects. Topics include Hermes agent,

78
00:03:14,600 --> 00:03:19,800
[แปลไทย 78] hyperframes, local AI tools, and many

79
00:03:16,600 --> 00:03:21,900
[แปลไทย 79] more. All standard wikis are free, but

80
00:03:19,800 --> 00:03:24,800
[แปลไทย 80] if you sign up for Agent Wikis Pro, you

81
00:03:21,900 --> 00:03:26,500
[แปลไทย 81] can access Excel Wikis as well as custom

82
00:03:24,800 --> 00:03:28,500
[แปลไทย 82] skills I've developed with my agents

83
00:03:26,500 --> 00:03:30,600
[แปลไทย 83] over months of working together. Pro is

84
00:03:28,500 --> 00:03:32,200
[แปลไทย 84] only $9.99 a month. And if you sign up

85
00:03:30,600 --> 00:03:33,800
[แปลไทย 85] now, you're locked into that price for

86
00:03:32,200 --> 00:03:35,300
[แปลไทย 86] life. I'm currently working on

87
00:03:33,800 --> 00:03:37,100
[แปลไทย 87] specialized agent profiles and

88
00:03:35,300 --> 00:03:38,800
[แปลไทย 88] automation workflow templates. So once

89
00:03:37,100 --> 00:03:40,600
[แปลไทย 89] those are released, the price for the

90
00:03:38,800 --> 00:03:43,400
[แปลไทย 90] protier will increase. So consider

91
00:03:40,600 --> 00:03:44,700
[แปลไทย 91] signing up today on agentwiks.com.

92
00:03:43,400 --> 00:03:47,500
[แปลไทย 92] Thank you all for your support. Now,

93
00:03:44,700 --> 00:03:49,400
[แปลไทย 93] back to the video. Okay, to start out

94
00:03:47,500 --> 00:03:50,900
[แปลไทย 94] we're going to do a very quick kind of

95
00:03:49,400 --> 00:03:53,500
[แปลไทย 95] presentations about the unique features

96
00:03:50,900 --> 00:03:55,400
[แปลไทย 96] of these two models. And I had they

97
00:03:53,500 --> 00:03:57,800
[แปลไทย 97] themselves make these presentations just

98
00:03:55,400 --> 00:04:00,200
[แปลไทย 98] based on their introductory blogs. So

99
00:03:57,800 --> 00:04:03,000
[แปลไทย 99] even though this is a presentation part,

100
00:04:00,200 --> 00:04:05,100
[แปลไทย 100] it's also kind of a a test to see how

101
00:04:03,000 --> 00:04:06,900
[แปลไทย 101] well they can make slides. Uh so we're

102
00:04:05,100 --> 00:04:11,000
[แปลไทย 102] going to start with Opus 5.5 and then

103
00:04:06,900 --> 00:04:13,000
[แปลไทย 103] we're going to move on to the GBT6 soul.

104
00:04:11,000 --> 00:04:15,000
[แปลไทย 104] Um, so the headline for Opus, like I

105
00:04:13,000 --> 00:04:17,800
[แปลไทย 105] said, is that it has Fable 5.1 level

106
00:04:15,000 --> 00:04:22,400
[แปลไทย 106] work, but a lot cheaper.

107
00:04:17,800 --> 00:04:25,600
[แปลไทย 107] And input is now $4 for a million tokens

108
00:04:22,400 --> 00:04:28,200
[แปลไทย 108] output 20. And generates 30% faster,

109
00:04:25,600 --> 00:04:31,500
[แปลไทย 109] they say. And we should be seeing Sonnet

110
00:04:28,200 --> 00:04:33,900
[แปลไทย 110] 5.5 and Haiku 5.5. We haven't had a

111
00:04:31,500 --> 00:04:36,200
[แปลไทย 111] Haiku model in a long time. Those are

112
00:04:33,900 --> 00:04:37,700
[แปลไทย 112] coming in the coming weeks. They say

113
00:04:36,200 --> 00:04:39,900
[แปลไทย 113] it's the best aligned model they've

114
00:04:37,700 --> 00:04:41,800
[แปลไทย 114] shipped. uh strongest score on their

115
00:04:39,900 --> 00:04:43,800
[แปลไทย 115] automated behavioral audit yet, which

116
00:04:41,800 --> 00:04:46,600
[แปลไทย 116] would be great. One of the big issues

117
00:04:43,800 --> 00:04:48,600
[แปลไทย 117] with Opus 5 was that it had some really

118
00:04:46,600 --> 00:04:50,900
[แปลไทย 118] terrible alignment issues. Like it would

119
00:04:48,600 --> 00:04:52,400
[แปลไทย 119] just do stuff that you didn't ask it for

120
00:04:50,900 --> 00:04:54,800
[แปลไทย 120] and it wouldn't do stuff that you did

121
00:04:52,400 --> 00:04:57,800
[แปลไทย 121] ask it for. So clear and more agentic

122
00:04:54,800 --> 00:04:59,700
[แปลไทย 122] puts key info up front. And this was an

123
00:04:57,800 --> 00:05:03,500
[แปลไทย 123] example they had in uh their Twitter

124
00:04:59,700 --> 00:05:05,400
[แปลไทย 124] post. You can see Opus 5, it was all

125
00:05:03,500 --> 00:05:08,100
[แปลไทย 125] this extra information and they're

126
00:05:05,400 --> 00:05:10,200
[แปลไทย 126] saying Cloud Opus 5.5. That's the key

127
00:05:08,100 --> 00:05:12,200
[แปลไทย 127] info up front and then kind of the

128
00:05:10,200 --> 00:05:14,400
[แปลไทย 128] details later, which would be great. I

129
00:05:12,200 --> 00:05:16,900
[แปลไทย 129] still struggle with these models the way

130
00:05:14,400 --> 00:05:19,100
[แปลไทย 130] they write sometimes. Um, in terms of

131
00:05:16,900 --> 00:05:24,000
[แปลไทย 131] benchmarks, top of the table on aentic

132
00:05:19,100 --> 00:05:27,800
[แปลไทย 132] coding, uh, Opus 5.5 beating GBT6 Astra

133
00:05:24,000 --> 00:05:30,900
[แปลไทย 133] in terminal bench 4.0 and incursor bench

134
00:05:27,800 --> 00:05:33,500
[แปลไทย 134] higher than Fable 5. than obviously the

135
00:05:30,900 --> 00:05:37,100
[แปลไทย 135] earlier models. This is another frontier

136
00:05:33,500 --> 00:05:39,200
[แปลไทย 136] code uh the main set. You could see this

137
00:05:37,100 --> 00:05:43,200
[แปลไทย 137] is on the different reasoning levels,

138
00:05:39,200 --> 00:05:46,500
[แปลไทย 138] but it is generally higher than Astra or

139
00:05:43,200 --> 00:05:48,200
[แปลไทย 139] even Fable on most of these. And you can

140
00:05:46,500 --> 00:05:49,400
[แปลไทย 140] see here like even on medium, which is

141
00:05:48,200 --> 00:05:51,200
[แปลไทย 141] kind of the default, the one that I

142
00:05:49,400 --> 00:05:53,300
[แปลไทย 142] would usually go for, you're getting a

143
00:05:51,200 --> 00:05:55,800
[แปลไทย 143] much higher score

144
00:05:53,300 --> 00:05:57,900
[แปลไทย 144] and at a lower cost even than some of

145
00:05:55,800 --> 00:06:00,500
[แปลไทย 145] these models.

146
00:05:57,900 --> 00:06:02,100
[แปลไทย 146] So that's encouraging if that's true.

147
00:06:00,500 --> 00:06:05,000
[แปลไทย 147] And you could see it leads on knowledge

148
00:06:02,100 --> 00:06:08,000
[แปลไทย 148] work but trails on science still uh

149
00:06:05,000 --> 00:06:12,200
[แปลไทย 149] humanity's last exam. It was above fable

150
00:06:08,000 --> 00:06:14,800
[แปลไทย 150] and you could see still be behind Astra

151
00:06:12,200 --> 00:06:17,500
[แปลไทย 151] um in terms of the science bench here.

152
00:06:14,800 --> 00:06:19,100
[แปลไทย 152] So that's it. Very quick presentation

153
00:06:17,500 --> 00:06:21,900
[แปลไทย 153] because we have a lot to go through. But

154
00:06:19,100 --> 00:06:24,700
[แปลไทย 154] the headline is that the benchmarks look

155
00:06:21,900 --> 00:06:27,400
[แปลไทย 155] really great performance on Fable levels

156
00:06:24,700 --> 00:06:29,600
[แปลไทย 156] but much much cheaper. Uh cheaper even

157
00:06:27,400 --> 00:06:34,200
[แปลไทย 157] than Opus 5 and certainly much cheaper

158
00:06:29,600 --> 00:06:36,600
[แปลไทย 158] than Fable. Uh so now we have GBT6 Soul

159
00:06:34,200 --> 00:06:41,100
[แปลไทย 159] and this is the presentation they made

160
00:06:36,600 --> 00:06:44,200
[แปลไทย 160] at a nice little orb here. So

161
00:06:41,100 --> 00:06:47,500
[แปลไทย 161] let's see uh what we have for them. So,

162
00:06:44,200 --> 00:06:50,600
[แปลไทย 162] GBT6 Soul boasts Frontier Work at lower

163
00:06:47,500 --> 00:06:52,900
[แปลไทย 163] token prices. This also has the low $2.

164
00:06:50,600 --> 00:06:55,000
[แปลไทย 164] It is cheaper than Opus, uh, basically

165
00:06:52,900 --> 00:06:58,600
[แปลไทย 165] half the price. So, you have to consider

166
00:06:55,000 --> 00:07:02,300
[แปลไทย 166] that. And it is cheaper than even GBT

167
00:06:58,600 --> 00:07:04,700
[แปลไทย 167] 5.6 Soul, which had the same $420.

168
00:07:02,300 --> 00:07:07,000
[แปลไทย 168] So, you're seeing some competition here

169
00:07:04,700 --> 00:07:09,300
[แปลไทย 169] between OpenAI Enthropic trying to get

170
00:07:07,000 --> 00:07:13,000
[แปลไทย 170] these kind of mid-tier, lower tier

171
00:07:09,300 --> 00:07:14,200
[แปลไทย 171] models, lowering their prices.

172
00:07:13,000 --> 00:07:16,900
[แปลไทย 172] So, we don't know how long this will

173
00:07:14,200 --> 00:07:19,300
[แปลไทย 173] last. It can always change, but as of

174
00:07:16,900 --> 00:07:22,200
[แปลไทย 174] now, that's what we get.

175
00:07:19,300 --> 00:07:24,200
[แปลไทย 175] So, better work per dollar. Uh, these

176
00:07:22,200 --> 00:07:26,400
[แปลไทย 176] are some of the benchmarks. And you can

177
00:07:24,200 --> 00:07:30,200
[แปลไทย 177] see they posted this graph in the GBT6

178
00:07:26,400 --> 00:07:32,900
[แปลไทย 178] soul. You can see this yellow line. Uh,

179
00:07:30,200 --> 00:07:36,000
[แปลไทย 179] scoring basically similar to Claude

180
00:07:32,900 --> 00:07:40,100
[แปลไทย 180] Fable 5.1, still below Astra, but

181
00:07:36,000 --> 00:07:44,700
[แปลไทย 181] basically at a similar level. Um and

182
00:07:40,100 --> 00:07:47,000
[แปลไทย 182] then above uh GPD 5.6 soul but at a

183
00:07:44,700 --> 00:07:51,000
[แปลไทย 183] lower cost significantly lower than

184
00:07:47,000 --> 00:07:53,000
[แปลไทย 184] Fable and a bit lower than

185
00:07:51,000 --> 00:07:55,400
[แปลไทย 185] uh 5.6

186
00:07:53,000 --> 00:07:57,300
[แปลไทย 186] in terms of long horizon coding which is

187
00:07:55,400 --> 00:08:02,600
[แปลไทย 187] important for me. It has code that

188
00:07:57,300 --> 00:08:06,900
[แปลไทย 188] survives review. Um it had 68.8% 8% on

189
00:08:02,600 --> 00:08:10,600
[แปลไทย 189] deep WSC and this is slightly under

190
00:08:06,900 --> 00:08:11,700
[แปลไทย 190] table 5 but at 80% lower cost per task.

191
00:08:10,600 --> 00:08:14,100
[แปลไทย 191] So that's the thing. It's not going to

192
00:08:11,700 --> 00:08:17,400
[แปลไทย 192] be the best model out there, but it's

193
00:08:14,100 --> 00:08:20,400
[แปลไทย 193] the cost per task. I just really want a

194
00:08:17,400 --> 00:08:22,300
[แปลไทย 194] model that is capable of long horizon

195
00:08:20,400 --> 00:08:25,000
[แปลไทย 195] coding. It doesn't have to be crazy. It

196
00:08:22,300 --> 00:08:26,900
[แปลไทย 196] doesn't have to be Fable 5 levels.

197
00:08:25,000 --> 00:08:29,000
[แปลไทย 197] It doesn't even need to be Astro levels,

198
00:08:26,900 --> 00:08:30,600
[แปลไทย 198] but it does need to be competent. um

199
00:08:29,000 --> 00:08:32,299
[แปลไทย 199] because I can have those bigger models

200
00:08:30,600 --> 00:08:34,200
[แปลไทย 200] create detailed plans and kind of

201
00:08:32,299 --> 00:08:38,799
[แปลไทย 201] orchestrate things. I just need a

202
00:08:34,200 --> 00:08:40,799
[แปลไทย 202] reliable worker and um 5.6 Soul was not

203
00:08:38,799 --> 00:08:42,700
[แปลไทย 203] always that and Opus 5 was certainly not

204
00:08:40,799 --> 00:08:45,100
[แปลไทย 204] that. And that's it. More attempts, less

205
00:08:42,700 --> 00:08:48,700
[แปลไทย 205] cost per run. And I'll probably be using

206
00:08:45,100 --> 00:08:51,000
[แปลไทย 206] Luna um on some lower level automation

207
00:08:48,700 --> 00:08:52,800
[แปลไทย 207] tasks as well. But in this video, we're

208
00:08:51,000 --> 00:08:56,900
[แปลไทย 208] just going to look at Soul. So let's

209
00:08:52,800 --> 00:08:56,900
[แปลไทย 209] that's enough of the charts.

210
00:08:57,000 --> 00:09:01,800
[แปลไทย 210] So the first one I want to do this is a

211
00:08:58,500 --> 00:09:05,600
[แปลไทย 211] newer task but this is uh from a post

212
00:09:01,800 --> 00:09:08,500
[แปลไทย 212] from uh Ivon uh Fioranti

213
00:09:05,600 --> 00:09:13,100
[แปลไทย 213] and the prompt is quite a big one and

214
00:09:08,500 --> 00:09:15,000
[แปลไทย 214] it's from uh Majid Banzapore apparently

215
00:09:13,100 --> 00:09:17,500
[แปลไทย 215] and then it is to create this kind of

216
00:09:15,000 --> 00:09:21,100
[แปลไทย 216] diarama and this is in 3.js this was

217
00:09:17,500 --> 00:09:23,000
[แปลไทย 217] made using high 4 I believe but this is

218
00:09:21,100 --> 00:09:26,200
[แปลไทย 218] the idea 3.js JS that you can kind of

219
00:09:23,000 --> 00:09:30,100
[แปลไทย 219] zoom in. It's a cherry blossom kind of

220
00:09:26,200 --> 00:09:31,800
[แปลไทย 220] bonsai tree. That would be a cool use of

221
00:09:30,100 --> 00:09:33,800
[แปลไทย 221] 3.js and something a little bit

222
00:09:31,800 --> 00:09:36,700
[แปลไทย 222] different than what we've done before.

223
00:09:33,800 --> 00:09:38,900
[แปลไทย 223] So that is going to be it. Um I'll put

224
00:09:36,700 --> 00:09:40,400
[แปลไทย 224] this post in

225
00:09:38,900 --> 00:09:43,200
[แปลไทย 225] the description because the prompt is

226
00:09:40,400 --> 00:09:47,500
[แปลไทย 226] quite large. It's down here from ivon.

227
00:09:43,200 --> 00:09:49,000
[แปลไทย 227] Uh but now we are in claude code and

228
00:09:47,500 --> 00:09:53,700
[แปลไทย 228] let's change our model here. Why are we

229
00:09:49,000 --> 00:09:55,200
[แปลไทย 229] on opus 4.8? Here we go. 5.5.

230
00:09:53,700 --> 00:09:56,700
[แปลไทย 230] So

231
00:09:55,200 --> 00:09:58,600
[แปลไทย 231] I'm going to paste in this very large

232
00:09:56,700 --> 00:10:01,300
[แปลไทย 232] prompt.

233
00:09:58,600 --> 00:10:03,800
[แปลไทย 233] And you can see it.

234
00:10:01,300 --> 00:10:06,000
[แปลไทย 234] That's what it is.

235
00:10:03,800 --> 00:10:07,900
[แปลไทย 235] So that is going to be using cloud code.

236
00:10:06,000 --> 00:10:10,700
[แปลไทย 236] And you can see here we are in OpenAI

237
00:10:07,900 --> 00:10:12,700
[แปลไทย 237] codeex GBT6 soul. So I'll give it the

238
00:10:10,700 --> 00:10:15,400
[แปลไทย 238] same prompt

239
00:10:12,700 --> 00:10:17,500
[แปลไทย 239] and we are going to see how well it

240
00:10:15,400 --> 00:10:19,300
[แปลไทย 240] does.

241
00:10:17,500 --> 00:10:21,700
[แปลไทย 241] Okay. So, we have our first and this is

242
00:10:19,300 --> 00:10:24,300
[แปลไทย 242] from Claude.

243
00:10:21,700 --> 00:10:26,100
[แปลไทย 243] And you could see here

244
00:10:24,300 --> 00:10:28,500
[แปลไทย 244] looks pretty good. I mean, it looks very

245
00:10:26,100 --> 00:10:32,000
[แปลไทย 245] realistic. Uh, so now we're going to

246
00:10:28,500 --> 00:10:34,200
[แปลไทย 246] scroll in. This is a heavy process. I'm

247
00:10:32,000 --> 00:10:36,900
[แปลไทย 247] not sure how well how smoothly this will

248
00:10:34,200 --> 00:10:39,200
[แปลไทย 248] move, but take a look. Scroll in a

249
00:10:36,900 --> 00:10:40,800
[แปลไทย 249] little bit.

250
00:10:39,200 --> 00:10:44,100
[แปลไทย 250] See, it looks very realistic. This is

251
00:10:40,800 --> 00:10:46,900
[แปลไทย 251] all 3.js, which is amazing.

252
00:10:44,100 --> 00:10:50,300
[แปลไทย 252] Like the bark looks really good.

253
00:10:46,900 --> 00:10:50,300
[แปลไทย 253] the flowers.

254
00:10:52,400 --> 00:10:56,800
[แปลไทย 254] So, obviously it's a bit laggy because

255
00:10:54,000 --> 00:10:59,300
[แปลไทย 255] we have two versions of this. Codex also

256
00:10:56,800 --> 00:11:01,700
[แปลไทย 256] finished this at the same time. So, look

257
00:10:59,300 --> 00:11:04,700
[แปลไทย 257] at that detail though. That is pretty

258
00:11:01,700 --> 00:11:04,700
[แปลไทย 258] impressive.

259
00:11:09,300 --> 00:11:14,000
[แปลไทย 259] That looks really good actually. Scroll

260
00:11:11,700 --> 00:11:15,400
[แปลไทย 260] back out a little bit. Uh so, back to

261
00:11:14,000 --> 00:11:17,800
[แปลไทย 261] the front here. Let's try a few of these

262
00:11:15,400 --> 00:11:19,400
[แปลไทย 262] adjustments.

263
00:11:17,800 --> 00:11:23,000
[แปลไทย 263] So you can see the whole kind of scene

264
00:11:19,400 --> 00:11:25,800
[แปลไทย 264] here. So it's currently at

265
00:11:23,000 --> 00:11:28,800
[แปลไทย 265] noon. Try morning. Should change the

266
00:11:25,800 --> 00:11:30,800
[แปลไทย 266] lighting a little bit. So this is Opus

267
00:11:28,800 --> 00:11:33,200
[แปลไทย 267] 5.5.

268
00:11:30,800 --> 00:11:36,200
[แปลไทย 268] Morning a little bit lighter. Try

269
00:11:33,200 --> 00:11:36,200
[แปลไทย 269] nighttime.

270
00:11:38,400 --> 00:11:44,800
[แปลไทย 270] I don't know how much this lag is uh due

271
00:11:41,800 --> 00:11:48,200
[แปลไทย 271] to the model itself or due to just my

272
00:11:44,800 --> 00:11:52,200
[แปลไทย 272] kind of older you see here the night

273
00:11:48,200 --> 00:11:52,200
[แปลไทย 273] looks good. Look at that.

274
00:11:52,600 --> 00:11:58,300
[แปลไทย 274] So GPT6 soul also finished. And this is

275
00:11:56,100 --> 00:12:01,100
[แปลไทย 275] what we got right now.

276
00:11:58,300 --> 00:12:04,200
[แปลไทย 276] Um you can see it's a bit more of a

277
00:12:01,100 --> 00:12:07,200
[แปลไทย 277] simplistic take. Now, you have to keep

278
00:12:04,200 --> 00:12:09,400
[แปลไทย 278] in mind that Soul is half the price and

279
00:12:07,200 --> 00:12:11,800
[แปลไทย 279] it was faster

280
00:12:09,400 --> 00:12:13,300
[แปลไทย 280] um than Claude.

281
00:12:11,800 --> 00:12:15,700
[แปลไทย 281] So, please take that into

282
00:12:13,300 --> 00:12:17,400
[แปลไทย 282] considerations. It's not a you know,

283
00:12:15,700 --> 00:12:19,100
[แปลไทย 283] they're not the exact same model, but

284
00:12:17,400 --> 00:12:23,100
[แปลไทย 284] they are kind of like the second tier of

285
00:12:19,100 --> 00:12:25,300
[แปลไทย 285] both of the the Big Labs models.

286
00:12:23,100 --> 00:12:27,600
[แปลไทย 286] Um, and you can see obviously there's a

287
00:12:25,300 --> 00:12:30,200
[แปลไทย 287] bit of a difference here,

288
00:12:27,600 --> 00:12:32,400
[แปลไทย 288] perhaps more than just a bit.

289
00:12:30,200 --> 00:12:37,600
[แปลไทย 289] So, you can see some flowers falling,

290
00:12:32,400 --> 00:12:37,600
[แปลไทย 290] but it's not quite as realistic as

291
00:12:37,800 --> 00:12:42,500
[แปลไทย 291] the um the clawed one.

292
00:12:43,300 --> 00:12:48,000
[แปลไทย 292] Not not able to scroll in.

293
00:12:46,400 --> 00:12:49,900
[แปลไทย 293] So, I can move it adjust it like that,

294
00:12:48,000 --> 00:12:52,900
[แปลไทย 294] but I can't like scroll in to get

295
00:12:49,900 --> 00:12:52,900
[แปลไทย 295] closer.

296
00:12:54,000 --> 00:12:57,800
[แปลไทย 296] Let's try nighttime.

297
00:12:59,400 --> 00:13:07,000
[แปลไทย 297] Guess that's night time.

298
00:13:01,900 --> 00:13:10,000
[แปลไทย 298] Uh so yeah, quite a bit different. Um in

299
00:13:07,000 --> 00:13:12,100
[แปลไทย 299] fact, drastically different if I'm being

300
00:13:10,000 --> 00:13:13,800
[แปลไทย 300] honest.

301
00:13:12,100 --> 00:13:15,900
[แปลไทย 301] But you're going to get cost savings

302
00:13:13,800 --> 00:13:19,000
[แปลไทย 302] with Soul. So we have to take that into

303
00:13:15,900 --> 00:13:20,500
[แปลไทย 303] consideration as well. So the next one I

304
00:13:19,000 --> 00:13:22,300
[แปลไทย 304] want to do is going to be a landing

305
00:13:20,500 --> 00:13:25,600
[แปลไทย 305] page. I want to see its front-end design

306
00:13:22,300 --> 00:13:28,100
[แปลไทย 306] work. And I'm gonna do this similar to

307
00:13:25,600 --> 00:13:31,600
[แปลไทย 307] I've done this before, but I have all

308
00:13:28,100 --> 00:13:34,800
[แปลไทย 308] these clips of this like dinosaur

309
00:13:31,600 --> 00:13:37,100
[แปลไทย 309] um that I did dinosaur kind of video

310
00:13:34,800 --> 00:13:39,600
[แปลไทย 310] series that I did

311
00:13:37,100 --> 00:13:44,500
[แปลไทย 311] for my son and I did it locally with

312
00:13:39,600 --> 00:13:46,600
[แปลไทย 312] Miniax uh M3 H3.

313
00:13:44,500 --> 00:13:49,100
[แปลไทย 313] So I wanted them to create a landing

314
00:13:46,600 --> 00:13:51,700
[แปลไทย 314] page using these clips and I have some

315
00:13:49,100 --> 00:13:54,500
[แปลไทย 315] other like photo assets. So, this is

316
00:13:51,700 --> 00:13:57,000
[แปลไทย 316] going to be their front-end design skill

317
00:13:54,500 --> 00:13:59,100
[แปลไทย 317] test. So, the prompt here is use the

318
00:13:57,000 --> 00:14:01,200
[แปลไทย 318] assets in this folder to create a

319
00:13:59,100 --> 00:14:03,000
[แปลไทย 319] beautiful landing page for Sunny Dino.

320
00:14:01,200 --> 00:14:05,500
[แปลไทย 320] Creative direction and style is up to

321
00:14:03,000 --> 00:14:08,200
[แปลไทย 321] you. Only rules are that it must one

322
00:14:05,500 --> 00:14:10,500
[แปลไทย 322] look playful and visually amazing. Uh,

323
00:14:08,200 --> 00:14:12,700
[แปลไทย 323] two, use a video scrolling aspect in

324
00:14:10,500 --> 00:14:14,500
[แปลไทย 324] some way. And three, have at least three

325
00:14:12,700 --> 00:14:17,400
[แปลไทย 325] other interactive elements. Four, run

326
00:14:14,500 --> 00:14:20,400
[แปลไทย 326] perfectly well in browser without any

327
00:14:17,400 --> 00:14:24,600
[แปลไทย 327] extra loading, etc. So, I'm going to

328
00:14:20,400 --> 00:14:26,200
[แปลไทย 328] give that to Codex here or Soul here,

329
00:14:24,600 --> 00:14:28,400
[แปลไทย 329] and I'm going to give the same prompt to

330
00:14:26,200 --> 00:14:31,400
[แปลไทย 330] Opus,

331
00:14:28,400 --> 00:14:33,400
[แปลไทย 331] and we'll see what they they make. Okay,

332
00:14:31,400 --> 00:14:34,900
[แปลไทย 332] so they're both done. Claude was

333
00:14:33,400 --> 00:14:39,600
[แปลไทย 333] actually faster. It only took 11

334
00:14:34,900 --> 00:14:41,800
[แปลไทย 334] minutes. Uh, Codeex here took

335
00:14:39,600 --> 00:14:44,100
[แปลไทย 335] uh 18 almost 19 minutes, which is

336
00:14:41,800 --> 00:14:45,900
[แปลไทย 336] interesting. Usually, the sole model is

337
00:14:44,100 --> 00:14:47,900
[แปลไทย 337] faster. Let's look at them. This is the

338
00:14:45,900 --> 00:14:49,800
[แปลไทย 338] clawed one first. You could see it's

339
00:14:47,900 --> 00:14:51,500
[แปลไทย 339] very colorful.

340
00:14:49,800 --> 00:14:53,800
[แปลไทย 340] very interactive. If I put my mouse over

341
00:14:51,500 --> 00:14:56,700
[แปลไทย 341] the number uh the letters here, they

342
00:14:53,800 --> 00:14:59,100
[แปลไทย 342] kind of bounce. That's really fun. See,

343
00:14:56,700 --> 00:15:00,900
[แปลไทย 343] shapes are floating around. There's kind

344
00:14:59,100 --> 00:15:02,200
[แปลไทย 344] of like a a dinosaur footprint. I don't

345
00:15:00,900 --> 00:15:05,800
[แปลไทย 345] know if you can tell that, but as I move

346
00:15:02,200 --> 00:15:09,300
[แปลไทย 346] my cursor around. Really, really good.

347
00:15:05,800 --> 00:15:11,800
[แปลไทย 347] That's actually I've done this a couple

348
00:15:09,300 --> 00:15:15,800
[แปลไทย 348] um models. I did this with Fable 5.1 as

349
00:15:11,800 --> 00:15:17,800
[แปลไทย 349] well as Astra and a bunch of others.

350
00:15:15,800 --> 00:15:20,500
[แปลไทย 350] Um, I never saw anything like a a hero

351
00:15:17,800 --> 00:15:22,600
[แปลไทย 351] section that was this fun though. So,

352
00:15:20,500 --> 00:15:24,900
[แปลไทย 352] let's scroll down. See if can fall

353
00:15:22,600 --> 00:15:26,800
[แปลไทย 353] through here. So, this is scroll to

354
00:15:24,900 --> 00:15:29,800
[แปลไทย 354] play. So, this is our scroll movie. See,

355
00:15:26,800 --> 00:15:31,500
[แปลไทย 355] does it actually work? Yep.

356
00:15:29,800 --> 00:15:33,100
[แปลไทย 356] Yeah. So, a lot of sometimes the the

357
00:15:31,500 --> 00:15:35,700
[แปลไทย 357] models have difficulty with this kind of

358
00:15:33,100 --> 00:15:37,800
[แปลไทย 358] scroll movie where I'm scrolling. That's

359
00:15:35,700 --> 00:15:39,500
[แปลไทย 359] fun. It has this little like dinosaur at

360
00:15:37,800 --> 00:15:43,000
[แปลไทย 360] the bottom that kind of bounces around

361
00:15:39,500 --> 00:15:44,800
[แปลไทย 361] as it goes. Um, but this one handles it

362
00:15:43,000 --> 00:15:47,300
[แปลไทย 362] really well.

363
00:15:44,800 --> 00:15:49,200
[แปลไทย 363] really smooth.

364
00:15:47,300 --> 00:15:51,100
[แปลไทย 364] That's excellent.

365
00:15:49,200 --> 00:15:53,800
[แปลไทย 365] Cloud models always do really well with

366
00:15:51,100 --> 00:15:56,600
[แปลไทย 366] um front end, I feel like. So, here we

367
00:15:53,800 --> 00:15:58,300
[แปลไทย 367] go. Drag to spin. Um so, there's

368
00:15:56,600 --> 00:16:01,400
[แปลไทย 368] obviously consistency issues, but that's

369
00:15:58,300 --> 00:16:03,800
[แปลไทย 369] a that's a image model issue. Nothing to

370
00:16:01,400 --> 00:16:05,300
[แปลไทย 370] do with quad.

371
00:16:03,800 --> 00:16:08,400
[แปลไทย 371] So, there we go. These kind of bounce

372
00:16:05,300 --> 00:16:11,000
[แปลไทย 372] around. This is very impressive. I like

373
00:16:08,400 --> 00:16:12,600
[แปลไทย 373] the color and everything here. Which

374
00:16:11,000 --> 00:16:14,200
[แปลไทย 374] shape fits. So, this is one of our

375
00:16:12,600 --> 00:16:20,200
[แปลไทย 375] interactive elements. So you can pick a

376
00:16:14,200 --> 00:16:20,200
[แปลไทย 376] shape, drop it here. It fits.

377
00:16:20,700 --> 00:16:26,300
[แปลไทย 377] Let's try. What if we put this in too

378
00:16:23,300 --> 00:16:26,300
[แปลไทย 378] round?

379
00:16:26,500 --> 00:16:32,900
[แปลไทย 379] What if it's uh pointy? That won't fit.

380
00:16:29,800 --> 00:16:36,600
[แปลไทย 380] So slightly uh got off here, the

381
00:16:32,900 --> 00:16:39,100
[แปลไทย 381] outline, but otherwise that's great.

382
00:16:36,600 --> 00:16:40,700
[แปลไทย 382] It's a great element. Tap tap crack.

383
00:16:39,100 --> 00:16:44,600
[แปลไทย 383] Sunny and mammoth found a nest of

384
00:16:40,700 --> 00:16:44,600
[แปลไทย 384] mystery eggs. Tap three times.

385
00:16:45,100 --> 00:16:50,900
[แปลไทย 385] So that's great, too. Little lizard, the

386
00:16:49,000 --> 00:16:53,600
[แปลไทย 386] frog,

387
00:16:50,900 --> 00:16:54,300
[แปลไทย 387] runaway eggs.

388
00:16:53,600 --> 00:16:57,000
[แปลไทย 388] That's really good with these

389
00:16:54,300 --> 00:16:58,200
[แปลไทย 389] interactive elements.

390
00:16:57,000 --> 00:17:01,500
[แปลไทย 390] I don't know what that is supposed to

391
00:16:58,200 --> 00:17:05,400
[แปลไทย 391] be. It's kind of a blank screen there.

392
00:17:01,500 --> 00:17:07,500
[แปลไทย 392] Is it supposed to play? Not sure. Uh,

393
00:17:05,400 --> 00:17:09,400
[แปลไทย 393] pop a bubble. Magic bubbles float up

394
00:17:07,500 --> 00:17:12,599
[แปลไทย 394] from the pond.

395
00:17:09,400 --> 00:17:15,400
[แปลไทย 395] Uh, let's pick the red one.

396
00:17:12,599 --> 00:17:19,599
[แปลไทย 396] So it plays this little clip of the red

397
00:17:15,400 --> 00:17:21,800
[แปลไทย 397] bubble which comes with strawberry.

398
00:17:19,599 --> 00:17:23,599
[แปลไทย 398] So that's good.

399
00:17:21,800 --> 00:17:26,599
[แปลไทย 399] Green,

400
00:17:23,599 --> 00:17:26,599
[แปลไทย 400] yellow,

401
00:17:28,200 --> 00:17:33,300
[แปลไทย 401] yellow bubble will come up here. There

402
00:17:30,500 --> 00:17:35,300
[แปลไทย 402] we go. So interactive element is really

403
00:17:33,300 --> 00:17:36,700
[แปลไทย 403] good. Like these are floating around and

404
00:17:35,300 --> 00:17:39,200
[แปลไทย 404] that kind of footprint thing stays with

405
00:17:36,700 --> 00:17:42,400
[แปลไทย 405] the cursor the whole time. Got the whole

406
00:17:39,200 --> 00:17:42,400
[แปลไทย 406] gang here.

407
00:17:42,500 --> 00:17:47,000
[แปลไทย 407] really really impressive. I mean, I

408
00:17:44,700 --> 00:17:51,000
[แปลไทย 408] think this is better than the the Fable

409
00:17:47,000 --> 00:17:54,000
[แปลไทย 409] 5.1 version of this.

410
00:17:51,000 --> 00:17:56,600
[แปลไทย 410] I mean, definitely more fun. Color is

411
00:17:54,000 --> 00:17:59,900
[แปลไทย 411] really good. It nailed all the technical

412
00:17:56,600 --> 00:18:02,600
[แปลไทย 412] elements and this hero section is really

413
00:17:59,900 --> 00:18:05,100
[แปลไทย 413] great. So, yeah, I mean, definitely I

414
00:18:02,600 --> 00:18:08,900
[แปลไทย 414] would say better than Fable 5.1

415
00:18:05,100 --> 00:18:11,800
[แปลไทย 415] in this specific t uh task. Uh so now

416
00:18:08,900 --> 00:18:15,200
[แปลไทย 416] let's check the codeex one finished as

417
00:18:11,800 --> 00:18:19,000
[แปลไทย 417] well. So this is the codeex one

418
00:18:15,200 --> 00:18:21,900
[แปลไทย 418] uh the GBT 6 soul version

419
00:18:19,000 --> 00:18:24,100
[แปลไทย 419] and see it's more simple. I mean it's a

420
00:18:21,900 --> 00:18:28,000
[แปลไทย 420] clean look

421
00:18:24,100 --> 00:18:28,000
[แปลไทย 421] but it's a bit more simple.

422
00:18:28,100 --> 00:18:33,100
[แปลไทย 422] And then we go down here.

423
00:18:31,400 --> 00:18:34,800
[แปลไทย 423] Let's see if the scroll story because

424
00:18:33,100 --> 00:18:37,900
[แปลไทย 424] previously

425
00:18:34,800 --> 00:18:40,200
[แปลไทย 425] um the GBT model that I tried did didn't

426
00:18:37,900 --> 00:18:42,200
[แปลไทย 426] do this well. It struggled with it and a

427
00:18:40,200 --> 00:18:43,700
[แปลไทย 427] lot of the smaller models struggled with

428
00:18:42,200 --> 00:18:45,800
[แปลไทย 428] this as well. So, it actually did the

429
00:18:43,700 --> 00:18:47,800
[แปลไทย 429] scroll movie really well. You can see

430
00:18:45,800 --> 00:18:50,700
[แปลไทย 430] I'm scrolling here and it's actually

431
00:18:47,800 --> 00:18:54,500
[แปลไทย 431] playing the the film.

432
00:18:50,700 --> 00:18:57,400
[แปลไทย 432] So, that's good. Going down here, I like

433
00:18:54,500 --> 00:19:01,600
[แปลไทย 433] the the fonts and everything, the themes

434
00:18:57,400 --> 00:19:06,200
[แปลไทย 434] that it chose. So, these are the videos.

435
00:19:01,600 --> 00:19:07,900
[แปลไทย 435] This is the crew, all the characters.

436
00:19:06,200 --> 00:19:10,800
[แปลไทย 436] Uh, this is one of our interactive

437
00:19:07,900 --> 00:19:12,800
[แปลไทย 437] elements. Pop into a color. So, do we

438
00:19:10,800 --> 00:19:15,900
[แปลไทย 438] pop the green?

439
00:19:12,800 --> 00:19:17,800
[แปลไทย 439] How do I pop it?

440
00:19:15,900 --> 00:19:21,300
[แปลไทย 440] Well, I can't really pop it, I guess. Go

441
00:19:17,800 --> 00:19:23,800
[แปลไทย 441] on, pick a color. Unless there's a way.

442
00:19:21,300 --> 00:19:26,100
[แปลไทย 442] So, that's it.

443
00:19:23,800 --> 00:19:27,500
[แปลไทย 443] Um, so a little bit weak on the

444
00:19:26,100 --> 00:19:31,400
[แปลไทย 444] interactive elements. I guess they're

445
00:19:27,500 --> 00:19:34,600
[แปลไทย 445] calling this an interactive element.

446
00:19:31,400 --> 00:19:36,300
[แปลไทย 446] Not as not as fun as the eggs.

447
00:19:34,600 --> 00:19:37,900
[แปลไทย 447] And this one, it's supposed to pop, I

448
00:19:36,300 --> 00:19:39,800
[แปลไทย 448] would imagine, but it's not doing

449
00:19:37,900 --> 00:19:43,300
[แปลไทย 449] anything.

450
00:19:39,800 --> 00:19:45,000
[แปลไทย 450] So, yeah. I mean, once again,

451
00:19:43,300 --> 00:19:47,900
[แปลไทย 451] it got the scroll movie right, which is

452
00:19:45,000 --> 00:19:50,300
[แปลไทย 452] good. And it's a cleaner look, but I

453
00:19:47,900 --> 00:19:53,800
[แปลไทย 453] mean,

454
00:19:50,300 --> 00:19:54,900
[แปลไทย 454] not really even comparable. Once again,

455
00:19:53,800 --> 00:19:56,600
[แปลไทย 455] I figured there would be a difference

456
00:19:54,900 --> 00:19:59,200
[แปลไทย 456] just because the cost difference between

457
00:19:56,600 --> 00:20:02,000
[แปลไทย 457] these two models. Once again, the Opus

458
00:19:59,200 --> 00:20:04,500
[แปลไทย 458] model is double the price. As of now,

459
00:20:02,000 --> 00:20:07,000
[แปลไทย 459] it's $4 input, whereas the Soul model is

460
00:20:04,500 --> 00:20:09,500
[แปลไทย 460] only two,

461
00:20:07,000 --> 00:20:12,200
[แปลไทย 461] but the the performance is quite

462
00:20:09,500 --> 00:20:15,300
[แปลไทย 462] significantly different. So, I guess it

463
00:20:12,200 --> 00:20:18,000
[แปลไทย 463] depends on what you prioritize

464
00:20:15,300 --> 00:20:19,800
[แปลไทย 464] because it's not like this is bad.

465
00:20:18,000 --> 00:20:21,600
[แปลไทย 465] Um, the interactive elements were weak,

466
00:20:19,800 --> 00:20:25,600
[แปลไทย 466] but it's just a little bit plain. when I

467
00:20:21,600 --> 00:20:25,600
[แปลไทย 467] was looking for something a bit more.

468
00:20:27,000 --> 00:20:33,300
[แปลไทย 468] This is really um encouraging though

469
00:20:31,000 --> 00:20:36,200
[แปลไทย 469] because this opus model looks awesome

470
00:20:33,300 --> 00:20:37,800
[แปลไทย 470] actually which will be great and get a

471
00:20:36,200 --> 00:20:41,800
[แปลไทย 471] lot more out of my usage because it does

472
00:20:37,800 --> 00:20:43,700
[แปลไทย 472] so far seem like close to fable level. I

473
00:20:41,800 --> 00:20:48,200
[แปลไทย 473] did that the cherry blossom one with

474
00:20:43,700 --> 00:20:50,800
[แปลไทย 474] fable as well and it was kind of on par

475
00:20:48,200 --> 00:20:52,500
[แปลไทย 475] with what we got with Opus. But let's do

476
00:20:50,800 --> 00:20:55,000
[แปลไทย 476] one more. So for the last one, it's

477
00:20:52,500 --> 00:20:56,900
[แปลไทย 477] going to be another gamedev one with um

478
00:20:55,000 --> 00:20:59,000
[แปลไทย 478] kind of reused

479
00:20:56,900 --> 00:21:00,600
[แปลไทย 479] elements. If you remember my token burn

480
00:20:59,000 --> 00:21:02,600
[แปลไทย 480] game, which is still underway, but I

481
00:21:00,600 --> 00:21:05,900
[แปลไทย 481] created a bunch of sprites like this for

482
00:21:02,600 --> 00:21:07,000
[แปลไทย 482] the game. So they're all here, and I'm

483
00:21:05,900 --> 00:21:10,600
[แปลไทย 483] basically just going to ask them to

484
00:21:07,000 --> 00:21:12,700
[แปลไทย 484] oneshot a single demo for like a Mega

485
00:21:10,600 --> 00:21:16,000
[แปลไทย 485] Man style platformer. You know, I've

486
00:21:12,700 --> 00:21:18,100
[แปลไทย 486] been doing that Sonic one. Uh but let's

487
00:21:16,000 --> 00:21:19,800
[แปลไทย 487] try a kind of Mega Man style using some

488
00:21:18,100 --> 00:21:21,000
[แปลไทย 488] of those assets. I'm just going to keep

489
00:21:19,800 --> 00:21:22,600
[แปลไทย 489] this prompt simple. I'm just going to

490
00:21:21,000 --> 00:21:24,500
[แปลไทย 490] say, "Use some of the assets in that

491
00:21:22,600 --> 00:21:26,500
[แปลไทย 491] folder to create a single stage Mega Man

492
00:21:24,500 --> 00:21:27,600
[แปลไทย 492] style platformer game. You have creative

493
00:21:26,500 --> 00:21:29,800
[แปลไทย 493] control. I want something that looks

494
00:21:27,600 --> 00:21:33,200
[แปลไทย 494] amazing and fun. No sounds." Very simple

495
00:21:29,800 --> 00:21:35,200
[แปลไทย 495] prompt. So, give that to Opus.

496
00:21:33,200 --> 00:21:39,100
[แปลไทย 496] And I'm going to give this I'm running

497
00:21:35,200 --> 00:21:39,100
[แปลไทย 497] low on usage for

498
00:21:39,100 --> 00:21:44,900
[แปลไทย 498] um for my Codeex account here, so I

499
00:21:42,300 --> 00:21:50,200
[แปลไทย 499] might need to reuse

500
00:21:44,900 --> 00:21:50,200
[แปลไทย 500] one of my um resets that we got today.

501
00:21:50,400 --> 00:21:55,400
[แปลไทย 501] Okay, so soul's done. I will say it only

502
00:21:53,400 --> 00:21:59,600
[แปลไทย 502] took 10 minutes. So Clo's still working

503
00:21:55,400 --> 00:22:01,200
[แปลไทย 503] on this. Here we go. Nova strike.

504
00:21:59,600 --> 00:22:04,200
[แปลไทย 504] So

505
00:22:01,200 --> 00:22:08,500
[แปลไทย 505] I play it full screen here. Start our

506
00:22:04,200 --> 00:22:10,800
[แปลไทย 506] mission. Space to jump. J to shoot.

507
00:22:08,500 --> 00:22:12,700
[แปลไทย 507] K to dash.

508
00:22:10,800 --> 00:22:14,500
[แปลไทย 508] So space. That looks pretty good

509
00:22:12,700 --> 00:22:17,000
[แปลไทย 509] actually.

510
00:22:14,500 --> 00:22:19,700
[แปลไทย 510] J to shoot.

511
00:22:17,000 --> 00:22:22,000
[แปลไทย 511] There we go.

512
00:22:19,700 --> 00:22:25,400
[แปลไทย 512] This looks good.

513
00:22:22,000 --> 00:22:25,400
[แปลไทย 513] There we go.

514
00:22:25,800 --> 00:22:30,400
[แปลไทย 514] Actually looks pretty good. Oh,

515
00:22:31,800 --> 00:22:36,000
[แปลไทย 515] I never did this uh prompt, so I wasn't

516
00:22:33,700 --> 00:22:40,200
[แปลไทย 516] sure how it was going to turn out,

517
00:22:36,000 --> 00:22:43,700
[แปลไทย 517] but this looks pretty good.

518
00:22:40,200 --> 00:22:43,700
[แปลไทย 518] Actually pretty fun.

519
00:22:51,000 --> 00:22:57,200
[แปลไทย 519] Okay, so we beat the boss

520
00:22:53,800 --> 00:22:59,000
[แปลไทย 520] and now we go. Exit open. There we go.

521
00:22:57,200 --> 00:23:01,300
[แปลไทย 521] So, pretty good. Actually, of the test

522
00:22:59,000 --> 00:23:03,400
[แปลไทย 522] that we've had, that was definitely the

523
00:23:01,300 --> 00:23:05,000
[แปลไทย 523] best that we've gotten from the soul

524
00:23:03,400 --> 00:23:07,500
[แปลไทย 524] model.

525
00:23:05,000 --> 00:23:10,400
[แปลไทย 525] So, pretty good.

526
00:23:07,500 --> 00:23:13,700
[แปลไทย 526] And didn't take that long, 10 minutes.

527
00:23:10,400 --> 00:23:17,800
[แปลไทย 527] But let's see what Claude gets.

528
00:23:13,700 --> 00:23:20,200
[แปลไทย 528] The colot is done. I took 22 minutes so

529
00:23:17,800 --> 00:23:24,400
[แปลไทย 529] a bit longer.

530
00:23:20,200 --> 00:23:27,200
[แปลไทย 530] Double actually than soul. So let's see.

531
00:23:24,400 --> 00:23:33,200
[แปลไทย 531] Z is jump.

532
00:23:27,200 --> 00:23:37,900
[แปลไทย 532] Uh shoot is X and then C is slide. Okay.

533
00:23:33,200 --> 00:23:40,000
[แปลไทย 533] Wow. So it even got the kind of Mega Man

534
00:23:37,900 --> 00:23:43,400
[แปลไทย 534] slide move.

535
00:23:40,000 --> 00:23:45,400
[แปลไทย 535] Yeah. I mean it's obviously much better.

536
00:23:43,400 --> 00:23:47,900
[แปลไทย 536] That looks like great. It looks like a

537
00:23:45,400 --> 00:23:50,300
[แปลไทย 537] proper game. Even you could slide. So it

538
00:23:47,900 --> 00:23:55,600
[แปลไทย 538] has the kind of conveyor belt thing to

539
00:23:50,300 --> 00:23:55,600
[แปลไทย 539] get used to the controls. But slide

540
00:23:57,200 --> 00:24:03,100
[แปลไทย 540] jump to the next section. Oh, there's a

541
00:23:59,900 --> 00:24:05,400
[แปลไทย 541] ladder even.

542
00:24:03,100 --> 00:24:10,200
[แปลไทย 542] It genuinely looks like a Mega Man.

543
00:24:05,400 --> 00:24:14,100
[แปลไทย 543] That's pretty amazing for 22 minutes.

544
00:24:10,200 --> 00:24:17,700
[แปลไทย 544] not good at this in game, but

545
00:24:14,100 --> 00:24:17,700
[แปลไทย 545] let's go down here.

546
00:24:26,300 --> 00:24:30,300
[แปลไทย 546] So, the only issue is that there's I

547
00:24:27,900 --> 00:24:33,100
[แปลไทย 547] don't think it's physically possible to

548
00:24:30,300 --> 00:24:34,600
[แปลไทย 548] jump over the spikes. So, some logic

549
00:24:33,100 --> 00:24:37,600
[แปลไทย 549] issues with the game, but it looks

550
00:24:34,600 --> 00:24:40,200
[แปลไทย 550] terrific and it definitely looks better

551
00:24:37,600 --> 00:24:42,200
[แปลไทย 551] than what we got out of Soul. There's a

552
00:24:40,200 --> 00:24:44,300
[แปลไทย 552] lot more elements, the ladder element,

553
00:24:42,200 --> 00:24:47,000
[แปลไทย 553] the different like elevator stuff like

554
00:24:44,300 --> 00:24:50,000
[แปลไทย 554] that. I think souls was pretty good, uh,

555
00:24:47,000 --> 00:24:53,200
[แปลไทย 555] but not quite at this level. So, those

556
00:24:50,000 --> 00:24:55,600
[แปลไทย 556] are our tasks. And you can see here some

557
00:24:53,200 --> 00:24:59,900
[แปลไทย 557] of the usage. I was at 90% when we

558
00:24:55,600 --> 00:25:01,400
[แปลไทย 558] started this. We're only at 92 now.

559
00:24:59,900 --> 00:25:04,600
[แปลไทย 559] And it's been running for a couple

560
00:25:01,400 --> 00:25:07,800
[แปลไทย 560] hours. So, the usage savings is

561
00:25:04,600 --> 00:25:11,900
[แปลไทย 561] significant for using Soul. And this was

562
00:25:07,800 --> 00:25:15,000
[แปลไทย 562] at 70 when we started. Now it's 76 for

563
00:25:11,900 --> 00:25:19,400
[แปลไทย 563] the weekly. These are both $100 plans.

564
00:25:15,000 --> 00:25:23,000
[แปลไทย 564] So, there is a difference in how much

565
00:25:19,400 --> 00:25:25,000
[แปลไทย 565] usage it's going to take. Opus 5 uh 5.5

566
00:25:23,000 --> 00:25:26,600
[แปลไทย 566] is going to take up more of your usage,

567
00:25:25,000 --> 00:25:30,300
[แปลไทย 567] but I think we've seen generally with

568
00:25:26,600 --> 00:25:34,100
[แปลไทย 568] every task basically it's significantly

569
00:25:30,300 --> 00:25:38,800
[แปลไทย 569] higher quality. And Soul seemed okay. It

570
00:25:34,100 --> 00:25:42,600
[แปลไทย 570] seemed some tasks it managed. The first

571
00:25:38,800 --> 00:25:46,000
[แปลไทย 571] game, opening game one, was was honestly

572
00:25:42,600 --> 00:25:49,100
[แปลไทย 572] not good. It was pretty poor and similar

573
00:25:46,000 --> 00:25:52,100
[แปลไทย 573] to earlier generation actually models or

574
00:25:49,100 --> 00:25:54,000
[แปลไทย 574] cheaper open- source models even. Uh the

575
00:25:52,100 --> 00:25:56,900
[แปลไทย 575] cherry blossom tree as well was not very

576
00:25:54,000 --> 00:25:58,800
[แปลไทย 576] good. The landing page was better than

577
00:25:56,900 --> 00:26:01,400
[แปลไทย 577] what I've seen in the past from these

578
00:25:58,800 --> 00:26:04,200
[แปลไทย 578] models. Uh but most impressive was that

579
00:26:01,400 --> 00:26:07,500
[แปลไทย 579] final game at the end. But I have been

580
00:26:04,200 --> 00:26:10,400
[แปลไทย 580] very impressed by Claude Opus 5.5. It is

581
00:26:07,500 --> 00:26:11,800
[แปลไทย 581] double the cost of soul once again. You

582
00:26:10,400 --> 00:26:14,500
[แปลไทย 582] know, I've said that several times, but

583
00:26:11,800 --> 00:26:16,000
[แปลไทย 583] it does bear repeating, but the

584
00:26:14,500 --> 00:26:19,400
[แปลไทย 584] performance has been really impressive

585
00:26:16,000 --> 00:26:21,400
[แปลไทย 585] just across the board. I was very

586
00:26:19,400 --> 00:26:24,100
[แปลไทย 586] skeptical that it was Fable 5.1 level

587
00:26:21,400 --> 00:26:26,600
[แปลไทย 587] work, but honestly, what we saw from

588
00:26:24,100 --> 00:26:28,400
[แปลไทย 588] these tasks today, it was at that level.

589
00:26:26,600 --> 00:26:30,800
[แปลไทย 589] So, I'll continue to work with these a

590
00:26:28,400 --> 00:26:32,300
[แปลไทย 590] lot. These will be two of probably the

591
00:26:30,800 --> 00:26:35,700
[แปลไทย 591] heavier

592
00:26:32,300 --> 00:26:39,300
[แปลไทย 592] uh use models that I use and I'll only

593
00:26:35,700 --> 00:26:42,200
[แปลไทย 593] save Fable and Astra for heavy duty

594
00:26:39,300 --> 00:26:44,400
[แปลไทย 594] coding or engineering work, really

595
00:26:42,200 --> 00:26:46,700
[แปลไทย 595] complicated code bases, stuff like that.

596
00:26:44,400 --> 00:26:49,500
[แปลไทย 596] So, I'm encouraged so far that Opus 5.5

597
00:26:46,700 --> 00:26:53,100
[แปลไทย 597] is really strong and soul as well seems

598
00:26:49,500 --> 00:26:55,600
[แปลไทย 598] like a capable model and um with a ton

599
00:26:53,100 --> 00:26:56,700
[แปลไทย 599] of usage potential.

600
00:26:55,600 --> 00:26:58,700
[แปลไทย 600] That's going to be it for this video.

601
00:26:56,700 --> 00:27:00,900
[แปลไทย 601] Please uh leave a comment. Let me know

602
00:26:58,700 --> 00:27:04,100
[แปลไทย 602] your early thoughts on these two models.

603
00:27:00,900 --> 00:27:04,100
[แปลไทย 603] And I'll see you in the next video.