The Joy of Coding Is Gone: My Honest Take on AI — Transcript

Full transcript

1. 0:00I've spent 10 years working in tech, and
2. 0:01in the last few months AI has changed
3. 0:03the job so dramatically that I barely
4. 0:06recognize it anymore. And I don't mean
5. 0:08that AI makes writing code faster. I
6. 0:10mean that the actual work I spent 10
7. 0:12years getting good at is now something I
8. 0:13can largely hand off to an agent. One of
9. 0:15my friends, probably the best software
10. 0:17engineer I personally know, got laid off
11. 0:19earlier this year. He's been in the
12. 0:21industry for a long ass time, and he has
13. 0:23basically decided he's never going back.
14. 0:25Not because he can't learn new tools,
15. 0:27but because software engineering is no
16. 0:28longer the job he wants to do. He loved
17. 0:31programming. He loved solving difficult
18. 0:33problems and building things, and that
19. 0:34hands-on moment-to-moment
20. 0:36problem-solving is increasingly being
21. 0:38replaced by managing AI agents. I
22. 0:41completely understand where he's coming
23. 0:42from because I'm experiencing the same
24. 0:44thing. And what's remarkable is how
25. 0:47quickly this has happened. At this
26. 0:49point, if I read something about AI
27. 0:51coding that was written before May of
28. 0:532026, I would absolutely consider that
29. 0:56outdated. That's how fast the tools have
30. 0:58changed.
31. 0:59Previously, AI was an assistant. Like in
32. 1:02in cursor, you just get to write a
33. 1:04function or explain something, debug an
34. 1:06issue, but you were still doing the
35. 1:07actual engineering. It was It was
36. 1:09autocomplete. But now you can give an AI
37. 1:11a task, it can explore a code base,
38. 1:13investigate the problem, run tests,
39. 1:15iterate on its own work. It's not
40. 1:16perfect, but it's good enough that for
41. 1:19many common programming tasks, doing the
42. 1:21work manually is becoming difficult to
43. 1:23justify. And that's a fundamental shift.
44. 1:27We're no longer just using tools to help
45. 1:29us program. We're increasingly
46. 1:31supervising systems that do the
47. 1:32programming for us. That changes which
48. 1:35skills are valuable, and it changes what
49. 1:37the job actually feels like. My biggest
50. 1:40strength in tech was always being able
51. 1:42to solve really difficult, deep
52. 1:45technical problems. I could go into a
53. 1:47complicated system and find issues that
54. 1:49other people couldn't. Things where
55. 1:50somebody might get sidetracked or find a
56. 1:52workaround or stop investigating before
57. 1:54they actually reached the root cause.
58. 1:57I enjoyed going those extra five or six
59. 1:59layers down until I genuinely understood
60. 2:02what was happening. That was my
61. 2:03specialty, and it was something I spent
62. 2:05a decade developing.
63. 2:07But, it turns out that this is also one
64. 2:10of the better uses for AI agents. They
65. 2:13can search through an enormous amount of
66. 2:15code, trace dependencies, investigate
67. 2:16different hypotheses, work through
68. 2:18problems without the same limitations on
69. 2:20attention and patience that people have.
70. 2:23They're not always successful,
71. 2:24obviously, but they're increasingly
72. 2:26capable of doing work that previously
73. 2:28required a great deal of experience, and
74. 2:30that's an uncomfortable realization.
75. 2:34Because if the thing that made you
76. 2:36particularly valuable was your ability
77. 2:38to solve problems when other people
78. 2:39couldn't, what happens when everybody
79. 2:41has access to solving those problems?
80. 2:44Your experience is still useful. You can
81. 2:46evaluate the results, identify mistakes,
82. 2:48direct the agent more effectively, but
83. 2:51the value of that experience has
84. 2:53changed. And beyond the economic
85. 2:55implications,
86. 2:57there's something else we're losing.
87. 3:00There's a certain mechanical joy to
88. 3:03writing code. You start with a problem,
89. 3:05you figure out a solution, and you
90. 3:06gradually construct something that's
91. 3:08elegant and complete. You understand
92. 3:10every piece because you personally
93. 3:12worked through it. And for a lot of
94. 3:14programmers, that process is the entire
95. 3:16reason they got into software
96. 3:17development. They like making things.
97. 3:20But now, if I have a feature to
98. 3:22implement, I have to ask myself why I
99. 3:24would spend two hours manually writing
100. 3:26something an agent could produce much
101. 3:27faster.
102. 3:29Even if I did enjoy those two hours, the
103. 3:31business justification increasingly
104. 3:33isn't there, and that's the problem. The
105. 3:36most productive way to do the job is
106. 3:38becoming less and less like the activity
107. 3:40that made people want to do the job in
108. 3:42the first place. Of course, you can
109. 3:44still write code manually. Nobody's
110. 3:45stopping you. But there's a difference
111. 3:47between doing something as a hobby and
112. 3:49having a professional career built
113. 3:51around that craft. And I think we're
114. 3:53underestimating how many people in this
115. 3:55industry actually care about that
116. 3:58distinction. Here's something I think
117. 3:59gets overlooked in conversations about
118. 4:01the future of programming. A lot of
119. 4:03extremely good software engineers don't
120. 4:07particularly care about the product.
121. 4:09They don't care about product market fit
122. 4:12or customers or building a business.
123. 4:14They just like solving hard technical
124. 4:16problems. And I'm talking about
125. 4:18genuinely excellent engineers, people
126. 4:21who can design incredibly complicated
127. 4:23systems and build things most other
128. 4:25people couldn't. Think of the Steve
129. 4:27Wozniak archetype, somebody who's
130. 4:29motivated primarily by the engineering
131. 4:31itself, the satisfaction of making
132. 4:33something work. The tech industry was
133. 4:36built in large part by people like that.
134. 4:38And now increasingly we're telling those
135. 4:40people that the way to remain valuable
136. 4:42is to become product focused. Understand
137. 4:45the customer, define the requirements,
138. 4:47figure out what should be built, and let
139. 4:49the agents handle the implementation.
140. 4:52Those are absolutely valuable skills,
141. 4:54but what if you don't want to do that?
142. 4:56What if the whole reason you became a
143. 4:58programmer was that you enjoyed
144. 5:00programming?
145. 5:02That's essentially where my friend has
146. 5:04landed. He's an exceptional engineer,
147. 5:06but the work is shifting away from the
148. 5:08thing he actually enjoys.
149. 5:10And I think we're going to see more of
150. 5:12that. Not necessarily because people
151. 5:14can't adapt, but because they don't want
152. 5:17the job they're asked to adapt to.
153. 5:19People often compare AI coding tools to
154. 5:22the invention of power tools. They'll
155. 5:24say that carpenters didn't disappear
156. 5:26when electric saws were invented, but I
157. 5:28think that analogy completely misses the
158. 5:30scale of what's happening. An electric
159. 5:33planer helps a carpenter work faster.
160. 5:35They're still making the furniture. What
161. 5:37we're experiencing is closer to
162. 5:40replacing the carpenter's workshop with
163. 5:42a fully automated furniture factory. You
164. 5:45describe what you want, the machine
165. 5:46produces it, and you inspect the
166. 5:48results. You might need to adjust
167. 5:50something or tell the machines to try
168. 5:52again, but you're no longer personally
169. 5:54constructing the thing. You've gone from
170. 5:56being a craftsman to being a factory
171. 5:58floor manager, essentially just yelling
172. 6:01at the machines. And I think that's a
173. 6:03much more accurate description of a
174. 6:05gentic software development.
175. 6:08There's a video I seen, I think it's the
176. 6:09Hisense factory for
177. 6:13it's television manufacturing. It's
178. 6:14incredibly sophisticated automated clean
179. 6:16room facility where a bunch of the
180. 6:18production happens without human hands
181. 6:20touching anything. That's what this
182. 6:21feels like. Except that metaphorically,
183. 6:23we were carving wood with hand planers
184. 6:25just a few months ago, and suddenly
185. 6:27we're operating these automated
186. 6:29factories.
187. 6:30It's an extraordinary technological
188. 6:32achievement, but if you enjoyed being a
189. 6:34craftsman, becoming the factory manager
190. 6:36isn't necessarily an improvement. And
191. 6:39this creates some genuinely difficult
192. 6:41questions about the future of software
193. 6:42careers. If the actual construction of
194. 6:45software becomes much cheaper, companies
195. 6:47may need fewer people doing that work.
196. 6:50Maybe cheaper software creates enormous
197. 6:52new demand and more opportunities.
198. 6:54That's possible. But we shouldn't assume
199. 6:57that automatically happens or that the
200. 6:59new opportunities will require the same
201. 7:01skills. The advice now is that engineers
202. 7:03should focus on architecture, judgment,
203. 7:06business problems, and directing AI. And
204. 7:09I think that's probably correct from a
205. 7:10career perspective, but we're also
206. 7:13describing a profession that is becoming
207. 7:15fundamentally different from the one
208. 7:17many of us entered.
209. 7:19For somebody who loves product
210. 7:20development, this might be the most
211. 7:21exciting time in history. You can build
212. 7:24things with a fraction of the resources
213. 7:26you previously needed. And for somebody
214. 7:28who loves the technical craft itself, it
215. 7:31can feel like the career is
216. 7:32disappearing. And I think both reactions
217. 7:35are completely reasonable.
218. 7:37I don't know where this settles. The
219. 7:39technology is changing too quickly to
220. 7:41confidently predict what the industry
221. 7:43looks like even a year from now.
222. 7:45What I do know is that I spent 10 years
223. 7:47developing a particular kind of
224. 7:49expertise and in the span of a few
225. 7:51months, the way that expertise is
226. 7:52applied has changed dramatically. And
227. 7:55when I see one of the best engineers
228. 7:56I've ever known deciding that he doesn't
229. 7:58work in software anymore, I find that
230. 8:00concerning.
231. 8:01Because he didn't stop loving
232. 8:03engineering.
233. 8:05The job stopped being the kind of
234. 8:07engineering he loved.
235. 8:09I'm curious how other people in tech are
236. 8:11experiencing this. Are AI agents making
237. 8:13your work more enjoyable or are you
238. 8:14starting to miss the parts of
239. 8:16programming that you used to do
240. 8:17yourself?
241. 8:19Let me know in the comments.
242. 8:21And if you want more of my thoughts
243. 8:23about tech careers navigating this new
244. 8:25AI age, join my newsletter using the
245. 8:28link in the description.
246. 8:29Because I think we're going to have to
247. 8:31rethink a lot of what we assume about
248. 8:33building a career in this industry.
249. 8:36And I don't think we're going back.