1
00:00:00,386 --> 00:00:01,653
Token 烧得最快的人

2
00:00:01,946 --> 00:00:02,932
凭什么最值钱

3
00:00:03,786 --> 00:00:06,713
硅谷大厂最近流行一个奇怪的排行榜

4
00:00:07,105 --> 00:00:08,212
不比代码写得好

5
00:00:08,465 --> 00:00:10,153
不比产品上线快

6
00:00:10,305 --> 00:00:12,413
反倒比一件听起来荒唐的事

7
00:00:12,906 --> 00:00:15,352
谁调用 AI 消耗的 Token 最快

8
00:00:15,826 --> 00:00:16,533
消耗越快

9
00:00:16,905 --> 00:00:19,353
工程师反而越被视作高手

10
00:00:19,832 --> 00:00:20,140
哈喽

11
00:00:20,392 --> 00:00:20,860
大家好

12
00:00:21,152 --> 00:00:22,160
我是王利杰

13
00:00:22,473 --> 00:00:24,160
先解释下什么是 Token

14
00:00:24,392 --> 00:00:26,439
你可以把它当成 AI 的燃料

15
00:00:26,672 --> 00:00:28,119
每次让 AI 干活

16
00:00:28,192 --> 00:00:30,360
从读取指令到生成结果

17
00:00:30,473 --> 00:00:33,080
消耗的计算资源都用 Token 计量

18
00:00:33,352 --> 00:00:34,340
Token 烧得越多

19
00:00:34,633 --> 00:00:35,992
AI 干的活越多

20
00:00:35,992 --> 00:00:37,559
你付的钱也越多

21
00:00:38,031 --> 00:00:38,857
有粉丝问我

22
00:00:39,190 --> 00:00:40,617
这也太反常识了吧

23
00:00:41,110 --> 00:00:43,357
公司不都追求降本增效

24
00:00:43,430 --> 00:00:45,018
Token 消耗越少越好吗

25
00:00:45,391 --> 00:00:46,338
怎么反过来了

26
00:00:46,631 --> 00:00:48,458
烧得越快反而越厉害

27
00:00:49,071 --> 00:00:52,157
这个问题触及了 AI 时代的核心逻辑

28
00:00:53,490 --> 00:00:54,476
想象一个场景

29
00:00:54,929 --> 00:00:56,817
工地上有两台挖掘机

30
00:00:56,929 --> 00:00:59,097
一台一天烧两百升柴油

31
00:00:59,170 --> 00:01:00,856
另一台只烧二十升

32
00:01:01,130 --> 00:01:01,877
单看油耗

33
00:01:02,170 --> 00:01:04,017
你会觉得省油那台划算

34
00:01:04,289 --> 00:01:07,336
可如果烧两百升的那台挖了一千方土

35
00:01:07,449 --> 00:01:09,476
烧二十升的只挖了五十方呢

36
00:01:09,970 --> 00:01:11,336
答案就反过来了

37
00:01:11,529 --> 00:01:12,636
柴油不是浪费

38
00:01:12,889 --> 00:01:14,336
是产出的影子

39
00:01:14,880 --> 00:01:16,767
Token 就是 AI 的柴油

40
00:01:17,240 --> 00:01:19,127
烧得多不代表浪费

41
00:01:19,240 --> 00:01:23,967
只代表你有能力让机器全速运转 持续满负荷工作

42
00:01:24,360 --> 00:01:25,187
而大多数人

43
00:01:25,481 --> 00:01:26,928
压根没这个能力

44
00:01:27,240 --> 00:01:28,288
我们来看看

45
00:01:28,321 --> 00:01:31,368
从普通程序员到顶级 AI 工程师

46
00:01:31,440 --> 00:01:33,047
中间差了几道坎

47
00:01:33,889 --> 00:01:34,475
第一道坎

48
00:01:34,769 --> 00:01:36,095
叫基础水平

49
00:01:36,288 --> 00:01:39,856
普通工程师拿到 AI 编程工具就开始写代码

50
00:01:40,008 --> 00:01:42,035
这就是现在很火的“氛围编程”

51
00:01:42,448 --> 00:01:43,975
不用自己一行行写

52
00:01:44,049 --> 00:01:46,327
用人话告诉 AI 想要什么功能

53
00:01:46,329 --> 00:01:47,296
让它代劳

54
00:01:47,856 --> 00:01:48,683
听起来美好

55
00:01:49,056 --> 00:01:49,963
实际操作呢

56
00:01:50,576 --> 00:01:52,704
AI 写的东西一测就有问题

57
00:01:53,017 --> 00:01:55,023
把报错丢回去让它修

58
00:01:55,177 --> 00:01:57,583
修完再测又冒出新毛病

59
00:01:57,696 --> 00:01:58,736
改个三四轮

60
00:01:58,737 --> 00:02:01,103
AI 还可能把之前改好的弄坏

61
00:02:01,177 --> 00:02:02,624
只好推倒重来

62
00:02:02,776 --> 00:02:04,856
这种边做边返工的模式

63
00:02:04,856 --> 00:02:07,055
一个月下来用包月套餐算

64
00:02:07,056 --> 00:02:09,064
大概消耗一百美元 Token

65
00:02:09,474 --> 00:02:10,340
一百美元不多

66
00:02:10,673 --> 00:02:11,780
但买到的是什么

67
00:02:12,233 --> 00:02:14,201
是一个效率极低的劳动力

68
00:02:14,354 --> 00:02:16,753
AI 九成时间没在推进项目

69
00:02:16,754 --> 00:02:18,320
全在修自己犯的错

70
00:02:18,689 --> 00:02:19,277
第二道坎

71
00:02:19,570 --> 00:02:21,136
叫工作流优化

72
00:02:21,409 --> 00:02:23,717
聪明的工程师很快发现秘诀

73
00:02:24,170 --> 00:02:25,777
问题不在 AI 笨

74
00:02:25,850 --> 00:02:28,576
而在交代任务的方式太粗糙

75
00:02:28,770 --> 00:02:30,596
直接说“帮我做个网站”

76
00:02:31,130 --> 00:02:32,928
AI 按自己的理解去做

77
00:02:32,929 --> 00:02:34,856
大概率不是你想要的

78
00:02:34,890 --> 00:02:37,136
然后就是无休止地修改

79
00:02:37,290 --> 00:02:40,289
但如果先花一小时把开发流程

80
00:02:40,290 --> 00:02:43,976
和多智能体的角色和协作规范定义清楚

81
00:02:44,010 --> 00:02:46,256
再把网站拆成五个模块

82
00:02:46,290 --> 00:02:50,648
把每个模块的功能 输入输出 边界条件写清楚

83
00:02:50,649 --> 00:02:51,808
再交给 AI

84
00:02:51,809 --> 00:02:54,576
一次通过的概率就会大幅提升

85
00:02:55,148 --> 00:02:56,636
这叫工作流设计

86
00:02:56,949 --> 00:02:57,815
输入精准了

87
00:02:57,909 --> 00:02:59,236
AI 返工就少

88
00:02:59,469 --> 00:03:00,136
但要注意

89
00:03:00,509 --> 00:03:02,188
虽然单次返工少了

90
00:03:02,189 --> 00:03:04,636
一天能推进的项目反而多了

91
00:03:04,789 --> 00:03:06,796
总 Token 消耗不降反升

92
00:03:07,590 --> 00:03:09,917
同时你会开始用更好的模型

93
00:03:09,990 --> 00:03:11,178
贵模型犯错少

94
00:03:11,471 --> 00:03:12,877
一次出活率高

95
00:03:13,231 --> 00:03:13,898
算总账

96
00:03:14,150 --> 00:03:14,817
省了时间

97
00:03:15,110 --> 00:03:16,278
花了更多钱

98
00:03:16,671 --> 00:03:17,669
到这个阶段

99
00:03:17,671 --> 00:03:20,078
一个月大概两百到四百美元

100
00:03:20,962 --> 00:03:22,409
这里有个微妙变化

101
00:03:22,642 --> 00:03:23,349
第一层时

102
00:03:23,682 --> 00:03:25,608
工程师输入大量指令

103
00:03:25,642 --> 00:03:27,488
AI 输出一堆半成品

104
00:03:27,522 --> 00:03:29,689
人再输入更多指令去修正

105
00:03:29,721 --> 00:03:31,768
投入大 有效产出小

106
00:03:31,922 --> 00:03:32,629
到了第二层

107
00:03:32,962 --> 00:03:34,801
工程师输入越来越少

108
00:03:34,802 --> 00:03:36,969
AI 产出反而越来越精准

109
00:03:37,202 --> 00:03:39,409
好的工作流就像精密模具

110
00:03:39,481 --> 00:03:40,689
原料放进去

111
00:03:40,721 --> 00:03:42,328
出来的就是标准件

112
00:03:42,522 --> 00:03:45,969
工程师从手工匠人变成了模具设计师

113
00:03:46,242 --> 00:03:49,608
模具设计师的价值不在于亲手做了多少产品

114
00:03:49,642 --> 00:03:53,129
而在于模具能让多少产品自动造出来

115
00:03:53,960 --> 00:03:55,187
真正拉开差距的

116
00:03:55,441 --> 00:03:56,387
是第三道坎

117
00:03:56,840 --> 00:03:57,727
高并发

118
00:03:58,240 --> 00:03:59,967
打个比方你就明白了

119
00:04:00,240 --> 00:04:01,028
前两层里

120
00:04:01,321 --> 00:04:02,368
你是个老板

121
00:04:02,601 --> 00:04:04,208
手下只有一个员工

122
00:04:04,481 --> 00:04:05,227
活交给它

123
00:04:05,680 --> 00:04:06,928
干完你检查

124
00:04:07,120 --> 00:04:08,967
检查完再派下一个

125
00:04:08,1000 --> 00:04:10,847
一次只能干一件事

126
00:04:11,322 --> 00:04:12,848
第三层不一样了

127
00:04:13,001 --> 00:04:15,089
你升级成了公司 CEO

128
00:04:15,162 --> 00:04:16,601
手下不是一个人

129
00:04:16,602 --> 00:04:17,628
而是一支团队

130
00:04:18,081 --> 00:04:19,161
有的写代码

131
00:04:19,162 --> 00:04:20,281
有的做设计

132
00:04:20,281 --> 00:04:21,108
有的写文档

133
00:04:21,361 --> 00:04:22,529
有的做测试

134
00:04:22,762 --> 00:04:23,548
更厉害的是

135
00:04:23,762 --> 00:04:25,408
这些员工还能分身

136
00:04:25,481 --> 00:04:27,920
就像孙悟空拔根汗毛吹口气

137
00:04:27,921 --> 00:04:29,168
变出五十个自己

138
00:04:30,130 --> 00:04:31,976
拿我的实际工作流举例

139
00:04:32,169 --> 00:04:34,416
做一个视频需要三十个场景

140
00:04:34,570 --> 00:04:37,077
传统做法是主 AI 逐个处理

141
00:04:37,450 --> 00:04:38,397
先生成图片

142
00:04:38,729 --> 00:04:39,888
再生成音频

143
00:04:39,890 --> 00:04:40,676
再合成视频

144
00:04:41,289 --> 00:04:43,296
做完一个才能做下一个

145
00:04:43,330 --> 00:04:46,376
三十个场景排着队处理要好几个小时

146
00:04:46,570 --> 00:04:47,436
高并发怎么做

147
00:04:48,049 --> 00:04:50,769
设计一个专门处理单个场景的角色

148
00:04:50,770 --> 00:04:54,217
让它自己调用图片 音频 视频合成功能

149
00:04:54,409 --> 00:04:57,097
然后把这个角色分成三十个分身

150
00:04:57,250 --> 00:04:59,248
每个领走一个场景编号

151
00:04:59,250 --> 00:05:00,337
同时开工

152
00:05:00,609 --> 00:05:01,676
原来几小时的活

153
00:05:01,929 --> 00:05:03,416
几分钟就干完了

154
00:05:04,353 --> 00:05:05,760
这就是高并发

155
00:05:06,113 --> 00:05:10,880
目前一个主 AI 最多能同时派五十个分身并行工作

156
00:05:11,272 --> 00:05:11,940
注意到没有

157
00:05:12,512 --> 00:05:15,320
到这一步已经不是一个人写代码了

158
00:05:15,353 --> 00:05:17,800
而是一个人在管理 AI 军团

159
00:05:18,392 --> 00:05:19,880
但这还没到极限

160
00:05:20,141 --> 00:05:20,848
第四道坎

161
00:05:21,141 --> 00:05:22,788
叫全链路压榨

162
00:05:23,062 --> 00:05:25,749
你不再是管一条产品线的 CEO

163
00:05:25,822 --> 00:05:28,468
而是同时管十个 CEO 的董事长

164
00:05:28,822 --> 00:05:31,548
每个 CEO 带着团队跑自己的项目

165
00:05:31,581 --> 00:05:33,269
十条线同时推进

166
00:05:33,382 --> 00:05:36,709
每条线里还有大量分身在高并发协作

167
00:05:37,097 --> 00:05:38,984
更极致的是通宵模式

168
00:05:39,337 --> 00:05:41,696
睡前给 AI 一个复杂大任务

169
00:05:41,698 --> 00:05:45,936
把规则 判断标准 异常处理方案全交代清楚

170
00:05:45,938 --> 00:05:47,224
让它跑一整夜

171
00:05:47,578 --> 00:05:48,496
早上起来

172
00:05:48,498 --> 00:05:50,737
原本需要一个团队干一周的项目

173
00:05:50,737 --> 00:05:51,984
它已经做完了

174
00:05:52,578 --> 00:05:54,085
听起来理想得不真实

175
00:05:54,458 --> 00:05:54,765
对吧

176
00:05:55,498 --> 00:05:57,145
但这里有个致命门槛

177
00:05:57,498 --> 00:05:59,825
如果工作流有一处没设计好

178
00:05:59,857 --> 00:06:01,825
或一条规则没交代到位

179
00:06:01,977 --> 00:06:05,984
AI 凌晨三点遇到无法判断的节点就会停下来等你

180
00:06:06,297 --> 00:06:09,337
第二天早上醒来发现它三点就停工了

181
00:06:09,337 --> 00:06:10,785
白白浪费一整晚

182
00:06:11,842 --> 00:06:13,330
这种事我经历过

183
00:06:13,482 --> 00:06:16,482
有次给 AI 布置复杂的视频制作任务

184
00:06:16,482 --> 00:06:17,510
步骤都交代了

185
00:06:17,683 --> 00:06:18,629
唯独漏了一条

186
00:06:18,922 --> 00:06:21,609
某张图片生成失败时该怎么处理

187
00:06:21,683 --> 00:06:24,970
结果凌晨两点 AI 生成了一张问题图片

188
00:06:25,042 --> 00:06:26,962
不知该跳过还是重试

189
00:06:26,962 --> 00:06:28,330
就停在那儿等我

190
00:06:28,443 --> 00:06:29,770
第二天早上一看

191
00:06:29,802 --> 00:06:31,681
三十个场景只做了六个

192
00:06:31,683 --> 00:06:33,729
剩下时间全在等我一句话

193
00:06:34,505 --> 00:06:38,592
要让 AI 在复杂任务上连续跑十几二十个小时不停

194
00:06:38,666 --> 00:06:41,592
中间不卡壳 不犯方向性错误

195
00:06:41,666 --> 00:06:43,553
该自主判断时能判断

196
00:06:43,585 --> 00:06:46,152
不该擅自做主时知道守规矩

197
00:06:46,265 --> 00:06:49,152
这对工作流的精密程度要求极高

198
00:06:49,505 --> 00:06:51,412
这就像设计一家真正的公司

199
00:06:51,825 --> 00:06:53,972
每个岗位的职责 汇报流程

200
00:06:54,265 --> 00:06:56,784
决策权限 异常上报机制

201
00:06:56,785 --> 00:07:01,272
都得写进虚拟公司的章程 操作手册甚至宪法里

202
00:07:01,426 --> 00:07:02,505
少一条规则

203
00:07:02,505 --> 00:07:05,073
AI 凌晨就可能在那个地方翻车

204
00:07:05,606 --> 00:07:06,813
到了这个级别

205
00:07:06,847 --> 00:07:10,253
工程师的工作节奏和传统程序员完全不同

206
00:07:10,407 --> 00:07:11,974
他不是坐那儿敲键盘

207
00:07:12,087 --> 00:07:14,573
而是同时开着好几条 AI 产品线

208
00:07:14,727 --> 00:07:16,214
时间被切成碎片

209
00:07:16,407 --> 00:07:17,573
AI 跑的时候

210
00:07:17,647 --> 00:07:19,474
他在思考下一个任务怎么安排

211
00:07:19,926 --> 00:07:21,405
哪条线的 AI 停了

212
00:07:21,407 --> 00:07:24,366
他立刻跳过去给新指令或排除卡点

213
00:07:24,366 --> 00:07:25,414
让它继续跑

214
00:07:25,686 --> 00:07:27,773
他自己几乎不做具体执行

215
00:07:27,926 --> 00:07:29,674
所有精力都花在一件事上

216
00:07:30,006 --> 00:07:31,253
让 AI 不要停

217
00:07:32,242 --> 00:07:33,150
这样的工程师

218
00:07:33,523 --> 00:07:34,629
一个月消耗多少

219
00:07:35,283 --> 00:07:36,410
拿我自己举例

220
00:07:36,763 --> 00:07:38,730
我用 Claude Code 包月套餐

221
00:07:38,882 --> 00:07:40,209
每月两百美元

222
00:07:40,362 --> 00:07:41,230
买到的 Token 量

223
00:07:41,482 --> 00:07:44,050
约等于按量计费四千美元的用量

224
00:07:44,322 --> 00:07:47,949
我一个月要消耗两个这样的账号才能把工作时间塞满

225
00:07:48,403 --> 00:07:49,562
一个 Claude Code

226
00:07:49,562 --> 00:07:50,269
一个 Codex

227
00:07:50,562 --> 00:07:51,610
分工合作

228
00:07:51,922 --> 00:07:53,442
折算成按量计费

229
00:07:53,442 --> 00:07:56,209
我每月 AI 算力消耗约八千美元

230
00:07:56,830 --> 00:07:57,957
你可能好奇

231
00:07:58,070 --> 00:07:59,457
Claude Code Max 订阅

232
00:07:59,471 --> 00:08:00,498
后面的“20X”

233
00:08:01,070 --> 00:08:01,957
是什么意思

234
00:08:02,270 --> 00:08:02,777
告诉你

235
00:08:03,111 --> 00:08:05,878
包月额度用完进入按量付费后

236
00:08:05,991 --> 00:08:09,517
最好的模型每小时消耗约十到十五美元

237
00:08:09,630 --> 00:08:12,669
一天十小时就是一百到一百五十美元

238
00:08:12,671 --> 00:08:14,957
一个月三千到四千多美元

239
00:08:15,231 --> 00:08:17,717
所以两百美元包月给的额度

240
00:08:17,750 --> 00:08:20,358
差不多等于四千美元按量消耗

241
00:08:20,510 --> 00:08:21,878
这就是那个二十倍

242
00:08:22,190 --> 00:08:24,277
这在硅谷已算不错水平

243
00:08:24,471 --> 00:08:26,918
但 AI 用得最好的那批工程师

244
00:08:26,950 --> 00:08:28,918
需要两到四个这样的账号

245
00:08:29,031 --> 00:08:32,438
一个月折合八千到一万六千美元 AI 算力

246
00:08:32,854 --> 00:08:34,654
现在回头看那个排行榜

247
00:08:34,654 --> 00:08:36,262
就一点都不奇怪了

248
00:08:36,575 --> 00:08:37,442
Token 烧得快

249
00:08:37,695 --> 00:08:38,681
代表三件事

250
00:08:39,254 --> 00:08:41,161
一 工作流设计精良

251
00:08:41,454 --> 00:08:42,561
AI 不会卡壳

252
00:08:43,175 --> 00:08:44,982
二 懂高并发

253
00:08:45,055 --> 00:08:47,321
能同时调度几十个 AI 分身

254
00:08:48,014 --> 00:08:51,781
三 能让 AI 在你睡觉时还全速干活

255
00:08:52,094 --> 00:08:54,441
这三件事说的其实是同一件事

256
00:08:54,935 --> 00:08:58,022
你有能力把 AI 劳动力压榨到极致

257
00:08:58,398 --> 00:08:58,985
发现了吗

258
00:08:59,398 --> 00:09:02,445
这和传统工程师评价标准完全不同

259
00:09:02,798 --> 00:09:04,445
以前衡量工程师

260
00:09:04,518 --> 00:09:05,985
看他自己能写多少代码

261
00:09:06,557 --> 00:09:09,304
现在看的是他能调度多少 AI 劳动力

262
00:09:09,678 --> 00:09:11,804
极限压榨到何种程度

263
00:09:12,117 --> 00:09:13,024
这跟什么很像

264
00:09:13,557 --> 00:09:15,125
跟管理能力很像

265
00:09:15,437 --> 00:09:17,205
职场最经典的跃迁

266
00:09:17,278 --> 00:09:19,625
就是从个人贡献者变成管理者

267
00:09:19,998 --> 00:09:22,164
不再看你自己能干多少活

268
00:09:22,197 --> 00:09:25,125
而是看你能带多大团队一起出活

269
00:09:25,880 --> 00:09:28,159
现在同样的跃迁又发生了

270
00:09:28,160 --> 00:09:29,648
只不过你管的不是人

271
00:09:29,680 --> 00:09:30,487
是 AI

272
00:09:30,800 --> 00:09:32,148
团队里没有一个活人

273
00:09:32,400 --> 00:09:33,648
全是智能体

274
00:09:33,841 --> 00:09:36,227
但作为管理者需要的能力本质没变

275
00:09:36,680 --> 00:09:41,487
拆任务 定流程 设边界 处理异常 做资源调度

276
00:09:41,680 --> 00:09:43,207
顶级 AI 工程师

277
00:09:43,241 --> 00:09:45,847
本质上就是顶级 AI 经理人

278
00:09:46,160 --> 00:09:47,568
所以那个排行榜

279
00:09:47,601 --> 00:09:49,648
比的不是谁更能花钱

280
00:09:49,721 --> 00:09:52,007
而是谁更会当 AI 的老板

281
00:09:52,754 --> 00:09:55,121
但这个故事还有更深一层含义

282
00:09:55,353 --> 00:09:58,793
当一个工程师用每月八千美元 AI 算力

283
00:09:58,794 --> 00:10:01,520
做出了原来十人团队一个月的产出

284
00:10:01,634 --> 00:10:03,260
你算算那十个人的成本

285
00:10:03,713 --> 00:10:06,760
硅谷工程师平均年薪二十万美元左右

286
00:10:06,794 --> 00:10:09,560
加上福利 保险 办公空间

287
00:10:09,674 --> 00:10:12,800
十个人一个月成本轻松超过二十万美元

288
00:10:12,993 --> 00:10:13,741
八千美元

289
00:10:14,034 --> 00:10:15,321
对二十万美元

290
00:10:15,353 --> 00:10:17,392
这个二十五比一的杠杆率

291
00:10:17,394 --> 00:10:19,920
才是排行榜背后真正在讲的事

292
00:10:20,193 --> 00:10:21,953
这已经不是技术话题

293
00:10:21,953 --> 00:10:24,400
而是经济问题 产业结构问题

294
00:10:24,554 --> 00:10:26,680
也是未来就业市场的信号

295
00:10:27,473 --> 00:10:30,040
谁能最高效调度 AI 劳动力

296
00:10:30,113 --> 00:10:33,440
谁就能以最低成本创造最大产出

297
00:10:34,017 --> 00:10:35,905
以前我们管这叫自动化

298
00:10:36,098 --> 00:10:39,244
但以前的自动化替代的是重复性体力劳动

299
00:10:39,617 --> 00:10:42,185
流水线工人 收费站收银员

300
00:10:42,417 --> 00:10:43,704
今天不一样了

301
00:10:43,738 --> 00:10:45,645
自动化开始替代认知劳动

302
00:10:46,058 --> 00:10:49,324
写代码 做设计 写文案 做研究

303
00:10:49,698 --> 00:10:52,217
这些曾被视作铁饭碗的脑力工作

304
00:10:52,218 --> 00:10:53,584
正被 AI 接管

305
00:10:53,738 --> 00:10:56,645
有没有想过为什么这轮自动化来得这么猛

306
00:10:57,218 --> 00:11:01,425
上一代自动化需要建工厂 买机器人 铺产线

307
00:11:01,537 --> 00:11:02,857
前期投入巨大

308
00:11:02,858 --> 00:11:04,464
只有大公司玩得起

309
00:11:04,617 --> 00:11:05,324
这一代不同

310
00:11:05,698 --> 00:11:08,896
一个人 一台笔记本 一个包月账号

311
00:11:08,897 --> 00:11:11,344
就是一条甚至多条流水线

312
00:11:12,137 --> 00:11:13,704
没有最低投入门槛

313
00:11:13,777 --> 00:11:15,505
没有规模经济限制

314
00:11:15,578 --> 00:11:17,045
扩散速度不是线性的

315
00:11:17,218 --> 00:11:18,584
而是指数级的

316
00:11:19,214 --> 00:11:21,941
驾驭这台认知自动化机器的钥匙

317
00:11:22,053 --> 00:11:23,653
不是写代码的能力

318
00:11:23,653 --> 00:11:28,461
而是设计系统 定义流程 管理智能体团队的能力

319
00:11:28,653 --> 00:11:30,100
如果你是程序员

320
00:11:30,214 --> 00:11:33,500
现在最该练的技能不是学新编程语言

321
00:11:33,573 --> 00:11:36,661
而是学会怎么当好 AI 团队的 CEO

322
00:11:36,854 --> 00:11:38,281
如果你不是程序员

323
00:11:38,734 --> 00:11:40,220
更该关注这件事

324
00:11:40,454 --> 00:11:43,140
这个趋势不会只停在编程领域

325
00:11:43,254 --> 00:11:45,340
任何需要认知能力的岗位

326
00:11:45,454 --> 00:11:47,201
未来都会面临同一个问题

327
00:11:47,893 --> 00:11:49,720
谁能更高效调度 AI

328
00:11:50,173 --> 00:11:51,581
谁就更有价值

329
00:11:51,813 --> 00:11:53,201
不会调度 AI 的人呢

330
00:11:53,734 --> 00:11:56,140
就像挖掘机旁边拿铲子的人

331
00:11:56,933 --> 00:11:57,520
你也能挖

332
00:11:58,134 --> 00:11:59,620
但你和那台机器

333
00:11:59,734 --> 00:12:01,380
根本不在一个维度上

334
00:12:01,840 --> 00:12:02,728
你们觉得

335
00:12:02,760 --> 00:12:05,947
下一个被拿来比烧 Token 速度的行业会是什么

336
00:12:06,441 --> 00:12:08,427
你自己现在处在哪个层级

337
00:12:08,1000 --> 00:12:10,088
评论区聊聊

338
00:12:10,281 --> 00:12:10,988
我是王利杰

339
00:12:11,281 --> 00:12:12,195
我们下期见