WEBVTT 1 00:00:00.270 --> 00:00:01.620 Hey there, Eden here. 2 00:00:01.620 --> 00:00:05.520 And this video is all about FOMO, the fear of missing out, 3 00:00:05.520 --> 00:00:09.570 and in the ever-changing rapid-pace environment 4 00:00:09.570 --> 00:00:10.830 that we live in. 5 00:00:10.830 --> 00:00:12.090 And specifically I'm talking 6 00:00:12.090 --> 00:00:15.210 about the AI development environment. 7 00:00:15.210 --> 00:00:19.170 So there is this blog which Andrej Karpathy wrote 8 00:00:19.170 --> 00:00:22.260 at the end of 2025, which really resonated 9 00:00:22.260 --> 00:00:24.420 with a lot of people, including myself. 10 00:00:24.420 --> 00:00:27.330 And I think it encompasses a lot of ideas 11 00:00:27.330 --> 00:00:29.967 which I actually agree and believe in. 12 00:00:29.967 --> 00:00:33.930 And this ex post is Andrej's sharing his feelings 13 00:00:33.930 --> 00:00:36.750 on this environment that we live in. 14 00:00:36.750 --> 00:00:38.550 So he says that, "I've never felt 15 00:00:38.550 --> 00:00:40.950 this much behind as a programmer." 16 00:00:40.950 --> 00:00:45.950 So this is very true and I can totally resonate with that. 17 00:00:45.990 --> 00:00:47.430 And I feel it every day 18 00:00:47.430 --> 00:00:51.300 because there are new things coming out every day 19 00:00:51.300 --> 00:00:55.080 and it's really hard, sorry, it's impossible 20 00:00:55.080 --> 00:00:56.730 to keep track on everything 21 00:00:56.730 --> 00:00:59.587 that is happening with all of the AI development. 22 00:00:59.587 --> 00:01:02.670 "The profession is being dramatically refactored 23 00:01:02.670 --> 00:01:04.140 as the bits contributed 24 00:01:04.140 --> 00:01:08.100 by the programmer are increasingly sparse and between. 25 00:01:08.100 --> 00:01:08.933 I have a sense 26 00:01:08.933 --> 00:01:11.640 that I could be 10X more powerful 27 00:01:11.640 --> 00:01:13.770 if I just properly string together 28 00:01:13.770 --> 00:01:16.920 what has become available over the last year, 29 00:01:16.920 --> 00:01:19.140 and failure to claim the boost 30 00:01:19.140 --> 00:01:22.500 feels decidedly like skill issue." 31 00:01:22.500 --> 00:01:25.590 And here he is referring to the FOMO feeling 32 00:01:25.590 --> 00:01:28.380 that we have all been living with, 33 00:01:28.380 --> 00:01:30.810 because we see tons of announcements 34 00:01:30.810 --> 00:01:32.700 and we really don't have the time 35 00:01:32.700 --> 00:01:34.260 to go and dive into everything. 36 00:01:34.260 --> 00:01:38.070 He says that there's a new programmable layer of abstraction 37 00:01:38.070 --> 00:01:41.130 to master in addition to the usual layers below, 38 00:01:41.130 --> 00:01:44.700 involving agents, sub-agents, their prompts, 39 00:01:44.700 --> 00:01:48.570 context, memory, modes, permissions, 40 00:01:48.570 --> 00:01:52.590 tools, plugins, skills, hooks, MCP, LSP, 41 00:01:52.590 --> 00:01:55.980 slash commands, workflows, IDE integrations, 42 00:01:55.980 --> 00:02:00.450 and the need to build an all-encompassing mental model 43 00:02:00.450 --> 00:02:02.250 for strengths and pitfalls 44 00:02:02.250 --> 00:02:07.050 of fundamentally stochastic, fallible, unintelligible 45 00:02:07.050 --> 00:02:11.130 and changing entities suddenly intermingled 46 00:02:11.130 --> 00:02:14.820 with what used to be good old-fashioned engineering. 47 00:02:14.820 --> 00:02:16.350 And the TL;DR for this 48 00:02:16.350 --> 00:02:19.320 is that the software stack has changed. 49 00:02:19.320 --> 00:02:24.320 We no longer write code, we write prompts that write code. 50 00:02:24.540 --> 00:02:26.550 We write prompts that execute 51 00:02:26.550 --> 00:02:29.730 in agents or subagents which invoke tools, 52 00:02:29.730 --> 00:02:31.500 and we get the artifact, 53 00:02:31.500 --> 00:02:33.510 and we are becoming the orchestrators. 54 00:02:33.510 --> 00:02:35.790 And the skill of software engineering 55 00:02:35.790 --> 00:02:38.700 is transferring from being a good coder 56 00:02:38.700 --> 00:02:40.350 of knowing the right syntax 57 00:02:40.350 --> 00:02:42.300 and how to use the technologies, 58 00:02:42.300 --> 00:02:46.980 but to have a more higher-layer orchestrator role. 59 00:02:46.980 --> 00:02:49.980 So we now manage a team of agents 60 00:02:49.980 --> 00:02:52.350 where each agent is doing something else. 61 00:02:52.350 --> 00:02:53.880 And I really like the comparison 62 00:02:53.880 --> 00:02:57.000 of a team lead which manages engineers. 63 00:02:57.000 --> 00:02:58.440 So the team lead goes 64 00:02:58.440 --> 00:03:00.660 and tells the engineers what to do. 65 00:03:00.660 --> 00:03:04.020 The team leader reviews their work, gives them feedback, 66 00:03:04.020 --> 00:03:06.570 and these loops goes on and on and on 67 00:03:06.570 --> 00:03:08.640 until we get the final product. 68 00:03:08.640 --> 00:03:11.010 And the same work 69 00:03:11.010 --> 00:03:15.750 a team lead is doing with engineers, we do with our agents. 70 00:03:15.750 --> 00:03:18.900 We give the agents tasks, we see how they're doing, 71 00:03:18.900 --> 00:03:21.450 we sometimes give them the correct tools 72 00:03:21.450 --> 00:03:24.330 and the correct skills to perform those tasks. 73 00:03:24.330 --> 00:03:28.110 And the work shifted more to review 74 00:03:28.110 --> 00:03:31.200 than to the actual writing and implementation. 75 00:03:31.200 --> 00:03:33.030 And this is a fundamental shift 76 00:03:33.030 --> 00:03:34.920 that we are experiencing right now. 77 00:03:34.920 --> 00:03:37.285 Now, all of the tech stacks of agents, 78 00:03:37.285 --> 00:03:39.270 subagents, MCP, LSP, slash command, 79 00:03:39.270 --> 00:03:42.270 workflows, IDE integrations, this is all true, 80 00:03:42.270 --> 00:03:43.680 and this is the new tech stack. 81 00:03:43.680 --> 00:03:47.250 This is the new programming language of the future. 82 00:03:47.250 --> 00:03:50.520 And I think we have great privilege 83 00:03:50.520 --> 00:03:55.080 to experience this firsthand as everything rapidly evolves. 84 00:03:55.080 --> 00:03:59.280 So we are actually witnessing a paradigm shift 85 00:03:59.280 --> 00:04:01.890 and a new abstraction layer 86 00:04:01.890 --> 00:04:05.910 that we are going to work in order to create software. 87 00:04:05.910 --> 00:04:08.760 So if in the beginning of software, 88 00:04:08.760 --> 00:04:10.350 people used to have punch cards, 89 00:04:10.350 --> 00:04:11.430 which they give the machine 90 00:04:11.430 --> 00:04:14.640 and this would run the code, then we had Assembly, 91 00:04:14.640 --> 00:04:18.180 then we had C++, which abstracted Assembly, 92 00:04:18.180 --> 00:04:20.940 then we had Python, which abstracted C++, 93 00:04:20.940 --> 00:04:23.190 so now we're in an era 94 00:04:23.190 --> 00:04:26.790 where programming languages are abstracted 95 00:04:26.790 --> 00:04:29.970 and English now is the new abstraction. 96 00:04:29.970 --> 00:04:33.630 And the real challenge is for us to orchestrate 97 00:04:33.630 --> 00:04:36.120 how to work with those new abstractions 98 00:04:36.120 --> 00:04:38.520 and to learn to work with their limitations 99 00:04:38.520 --> 00:04:40.200 and to solve those limitations 100 00:04:40.200 --> 00:04:42.420 and to continuously push the limits. 101 00:04:42.420 --> 00:04:45.480 And it talks here about good old-fashioned engineering. 102 00:04:45.480 --> 00:04:47.520 And I have to give you an example 103 00:04:47.520 --> 00:04:49.050 from my personal experience. 104 00:04:49.050 --> 00:04:51.570 So when I was a software engineer, 105 00:04:51.570 --> 00:04:54.480 I hated writing regular expressions. 106 00:04:54.480 --> 00:04:55.410 I hated it. 107 00:04:55.410 --> 00:04:56.820 And there is this website, 108 00:04:56.820 --> 00:04:59.340 I think it's called regex101.com, 109 00:04:59.340 --> 00:05:00.420 and this was a website 110 00:05:00.420 --> 00:05:01.980 that really made things easier 111 00:05:01.980 --> 00:05:04.170 when writing regular expressions. 112 00:05:04.170 --> 00:05:06.900 So we had in our thing, this guy, 113 00:05:06.900 --> 00:05:10.020 he was really, really good in writing regexes. 114 00:05:10.020 --> 00:05:13.140 And every time I needed to go and write one, 115 00:05:13.140 --> 00:05:15.510 I would always go and ask him. 116 00:05:15.510 --> 00:05:16.980 And it was really a skill, 117 00:05:16.980 --> 00:05:20.130 and he was renowned to it in the entire company 118 00:05:20.130 --> 00:05:22.200 that he knew how to write regex. 119 00:05:22.200 --> 00:05:25.320 So everyone who needed to write a regex, 120 00:05:25.320 --> 00:05:28.050 if it was something complicated and not trivial, 121 00:05:28.050 --> 00:05:29.700 that guy was the address. 122 00:05:29.700 --> 00:05:32.490 We always went and asked for this engineer's help. 123 00:05:32.490 --> 00:05:35.250 And today, this skill is actually not useful. 124 00:05:35.250 --> 00:05:37.470 It's like learning how to navigate in a map 125 00:05:37.470 --> 00:05:39.000 in order to drive somewhere. 126 00:05:39.000 --> 00:05:41.640 You have Google Maps, you have Waze, 127 00:05:41.640 --> 00:05:44.550 and today, you can simply ask an LLM 128 00:05:44.550 --> 00:05:46.230 or Cloud Code or Cursor 129 00:05:46.230 --> 00:05:47.940 how to write the regular expression 130 00:05:47.940 --> 00:05:49.230 for what you're looking for. 131 00:05:49.230 --> 00:05:54.030 So, the good old-fashioned engineering is changing. 132 00:05:54.030 --> 00:05:56.880 And there are a lot of skills and techniques 133 00:05:56.880 --> 00:06:00.270 and practices that software engineers had 134 00:06:00.270 --> 00:06:02.610 which are no longer relevant. 135 00:06:02.610 --> 00:06:05.010 And I think being a good software engineer 136 00:06:05.010 --> 00:06:08.340 is simply to be a good problem solver, 137 00:06:08.340 --> 00:06:10.320 and simply being curious, 138 00:06:10.320 --> 00:06:12.600 I think those are the most important skills. 139 00:06:12.600 --> 00:06:15.990 And this is valid for the old software engineering era 140 00:06:15.990 --> 00:06:19.290 and it's valid for the future software engineering era. 141 00:06:19.290 --> 00:06:21.450 And I think curiosity today 142 00:06:21.450 --> 00:06:24.420 is having much more impact on your skills. 143 00:06:24.420 --> 00:06:27.210 Because curiosity, this is what driving the change. 144 00:06:27.210 --> 00:06:29.580 This is what driving to push the limits 145 00:06:29.580 --> 00:06:31.110 and to try things new 146 00:06:31.110 --> 00:06:33.810 and to learn how to do things differently 147 00:06:33.810 --> 00:06:35.280 and potentially better. 148 00:06:35.280 --> 00:06:37.237 And Andrej continues, he said, 149 00:06:37.237 --> 00:06:41.340 "Clearly some powerful alien tool was handed around, 150 00:06:41.340 --> 00:06:43.260 except it comes with no manual 151 00:06:43.260 --> 00:06:47.340 and everyone has to figure out how to hold it and operate it 152 00:06:47.340 --> 00:06:50.190 while the resulting magnitude nine earthquake 153 00:06:50.190 --> 00:06:51.780 is rocking the profession. 154 00:06:51.780 --> 00:06:55.020 Roll up your sleeves to not fall behind." 155 00:06:55.020 --> 00:06:56.677 And I think the key part here is, 156 00:06:56.677 --> 00:06:58.620 "comes with no manual." 157 00:06:58.620 --> 00:07:03.620 And here, actually I think the imagination is the limit, 158 00:07:04.440 --> 00:07:07.830 because all the technology we're seeing of agents, 159 00:07:07.830 --> 00:07:08.820 of coding agents, 160 00:07:08.820 --> 00:07:10.410 of how impressive they get 161 00:07:10.410 --> 00:07:12.510 is a building block on a building block 162 00:07:12.510 --> 00:07:14.610 on a building block on a building block 163 00:07:14.610 --> 00:07:17.370 that will stack up one after another, 164 00:07:17.370 --> 00:07:19.860 each one using the abstraction below. 165 00:07:19.860 --> 00:07:24.117 And it's really amazing to live and see this in action. 166 00:07:24.117 --> 00:07:25.680 And I was lucky enough 167 00:07:25.680 --> 00:07:29.100 to jump on this train when it began, 168 00:07:29.100 --> 00:07:30.990 and I saw how everything evolved, 169 00:07:30.990 --> 00:07:34.140 and I saw how one abstraction 170 00:07:34.140 --> 00:07:36.360 got better and better and better, 171 00:07:36.360 --> 00:07:39.150 and then there was another layer of abstraction 172 00:07:39.150 --> 00:07:44.150 that was created on top of it, which gave us insane results. 173 00:07:44.340 --> 00:07:48.180 And I think the best way to have a deep understanding 174 00:07:48.180 --> 00:07:51.510 is to go back in time and see how things evolved. 175 00:07:51.510 --> 00:07:55.200 Because when you see how things evolved in this space 176 00:07:55.200 --> 00:07:57.030 and you see the iterations, 177 00:07:57.030 --> 00:07:58.650 for example, estate agents, 178 00:07:58.650 --> 00:08:01.260 it all started from a ReAct prompt. 179 00:08:01.260 --> 00:08:03.900 Then this ReAct prompt evolved to function calling, 180 00:08:03.900 --> 00:08:07.410 then with function calling, we got into agentic workflows. 181 00:08:07.410 --> 00:08:08.790 And then after that, 182 00:08:08.790 --> 00:08:11.970 we are in this deep agents era right now 183 00:08:11.970 --> 00:08:14.647 and task delegation and subagents. 184 00:08:14.647 --> 00:08:17.310 And when you're seeing this firsthand 185 00:08:17.310 --> 00:08:19.590 and when you jump on the train 186 00:08:19.590 --> 00:08:22.590 and you see how abstraction is solving problems 187 00:08:22.590 --> 00:08:24.180 from the early abstraction, 188 00:08:24.180 --> 00:08:26.490 you really get in-depth understanding. 189 00:08:26.490 --> 00:08:30.420 And it really removes a bit this FOMO feeling, 190 00:08:30.420 --> 00:08:33.060 because you know what problems exist today, 191 00:08:33.060 --> 00:08:36.630 and when new tools come, you know what they solve. 192 00:08:36.630 --> 00:08:39.180 Now, it says everyone has to figure out 193 00:08:39.180 --> 00:08:41.250 how to hold it and operate it. 194 00:08:41.250 --> 00:08:44.640 So the key thing here is that, of course, 195 00:08:44.640 --> 00:08:47.880 the best practices are things that are in the making. 196 00:08:47.880 --> 00:08:50.550 And we don't have manuals, 197 00:08:50.550 --> 00:08:54.300 we don't have official documentations of imagination, 198 00:08:54.300 --> 00:08:56.640 because it really boils down to imaginations 199 00:08:56.640 --> 00:08:59.550 and to try things and to experiment. 200 00:08:59.550 --> 00:09:01.860 So I think this area here 201 00:09:01.860 --> 00:09:05.580 that we're living in, experimentation is key. 202 00:09:05.580 --> 00:09:09.160 And getting our hands dirty, 203 00:09:09.160 --> 00:09:10.620 as it says here, 204 00:09:10.620 --> 00:09:13.800 is very important in order to do this experimentation. 205 00:09:13.800 --> 00:09:15.000 And in my opinion, 206 00:09:15.000 --> 00:09:19.110 this is the moral of the story of this tweet. 207 00:09:19.110 --> 00:09:22.740 It's sort of assuring us and calming us, 208 00:09:22.740 --> 00:09:26.040 saying that, "Hey, we live in this FOMO world, 209 00:09:26.040 --> 00:09:27.960 and even Andrej Karpathy, 210 00:09:27.960 --> 00:09:31.230 who is a renowned computer scientist, 211 00:09:31.230 --> 00:09:33.000 even he feels the FOMO. 212 00:09:33.000 --> 00:09:37.320 Even he feels not knowing enough and falling behind." 213 00:09:37.320 --> 00:09:41.520 And the way to solve it is to get your hands dirty 214 00:09:41.520 --> 00:09:43.200 and to start experimenting. 215 00:09:43.200 --> 00:09:45.510 And in order to solve this FOMO 216 00:09:45.510 --> 00:09:47.370 is to roll up your sleeves 217 00:09:47.370 --> 00:09:51.120 and to start getting your hands dirty and experimenting. 218 00:09:51.120 --> 00:09:52.953 Now, I have to tell you, the FOMO, 219 00:09:52.953 --> 00:09:54.720 this is something here to stay. 220 00:09:54.720 --> 00:09:56.280 You are always going to feel this FOMO. 221 00:09:56.280 --> 00:09:57.720 I feel it myself. 222 00:09:57.720 --> 00:10:02.160 And I think here focus is really, really important as well. 223 00:10:02.160 --> 00:10:04.470 There are lots of tweets every day, 224 00:10:04.470 --> 00:10:06.750 lots of new technologies, new products 225 00:10:06.750 --> 00:10:09.120 planning to solve X and X and Y. 226 00:10:09.120 --> 00:10:12.720 And I think the skill of distilling the noise 227 00:10:12.720 --> 00:10:16.020 and focusing on what's important, 228 00:10:16.020 --> 00:10:18.930 it's really, really important in this era, 229 00:10:18.930 --> 00:10:20.460 because this is a rabbit hole. 230 00:10:20.460 --> 00:10:24.240 If you go into Twitter, you start scrolling around, 231 00:10:24.240 --> 00:10:27.480 then it's going to have tons of content. 232 00:10:27.480 --> 00:10:31.290 And there is this infinite rabbit hole we can go into 233 00:10:31.290 --> 00:10:33.390 if we don't focus. 234 00:10:33.390 --> 00:10:38.390 And my two cents is that when I go through Twitter, 235 00:10:38.790 --> 00:10:40.350 I see what's new. 236 00:10:40.350 --> 00:10:42.270 I don't go through everything. 237 00:10:42.270 --> 00:10:43.920 I don't dive in deep into everything. 238 00:10:43.920 --> 00:10:45.840 If I found something that interests me, 239 00:10:45.840 --> 00:10:47.490 that solves the problem 240 00:10:47.490 --> 00:10:49.500 that I'm really interested in solving, 241 00:10:49.500 --> 00:10:52.890 I go, I dive deep, and I get my hands dirty. 242 00:10:52.890 --> 00:10:54.600 Cool, so that's it for this video. 243 00:10:54.600 --> 00:10:57.180 And I think my main key point here 244 00:10:57.180 --> 00:10:59.670 is that you are not alone with the FOMO. 245 00:10:59.670 --> 00:11:02.400 everyone is experiencing, including myself, 246 00:11:02.400 --> 00:11:04.890 including Andrej Karpathy. 247 00:11:04.890 --> 00:11:07.020 And we need to accept this FOMO 248 00:11:07.020 --> 00:11:09.063 and we need to keep living with it.