WEBVTT 00:00.000 --> 00:01.500 Welcome to The Explainer. 00:01.500 --> 00:04.300 Today, we're diving deep into AI co-pilots. 00:04.300 --> 00:05.960 And we're not just talking about the hype. 00:05.960 --> 00:08.460 We're looking at how these tools are really changing the game 00:08.460 --> 00:10.920 for software developers, the good, the bad, 00:10.920 --> 00:12.540 and everything in between. 00:12.540 --> 00:15.180 All right, let's just jump right in with a quote that, 00:15.180 --> 00:17.600 well, it pretty much says it all. 00:17.600 --> 00:18.680 Imagine this. 00:18.680 --> 00:20.920 You ship a feature in three days, 00:20.920 --> 00:24.100 a feature that your team thought would take two weeks. 00:24.100 --> 00:25.520 Think about that for a second. 00:25.520 --> 00:27.220 That's not just a little bit faster. 00:27.220 --> 00:29.060 That's a total transformation. 00:29.360 --> 00:30.600 And how they do it? 00:30.600 --> 00:32.960 It wasn't about working crazy hours or cutting corners. 00:32.960 --> 00:33.800 Nope. 00:33.800 --> 00:36.240 It was about an engineer who made their AI co-pilot 00:36.240 --> 00:38.120 a true partner in the process. 00:38.120 --> 00:39.600 But here's the thing. 00:39.600 --> 00:42.600 That quote, it's not just about speed. 00:42.600 --> 00:44.040 That's the real kicker. 00:44.040 --> 00:46.040 Because when you bring these tools in, 00:46.040 --> 00:48.640 you're also introducing brand new kinds of bugs, 00:48.640 --> 00:51.520 you're forcing a complete overhaul of team processes, 00:51.520 --> 00:53.660 and you're fundamentally changing what it even means 00:53.660 --> 00:55.100 to be a good developer. 00:55.100 --> 00:57.680 It's a whole new playbook. 00:57.700 --> 01:00.580 Okay, so to really get our heads around this shift, 01:00.580 --> 01:02.500 let's do a little reality check. 01:02.500 --> 01:05.300 We need to be super clear on what these tools are, 01:05.300 --> 01:08.500 and maybe even more importantly, what they aren't. 01:08.500 --> 01:10.140 So you might be thinking, hang on, 01:10.140 --> 01:11.980 haven't we had autocomplete forever? 01:11.980 --> 01:12.940 And yeah, you're right, 01:12.940 --> 01:15.260 but this is a whole different beast. 01:15.260 --> 01:17.340 The secret sauce here is that these tools 01:17.340 --> 01:20.220 use large language models to understand your intent, 01:20.220 --> 01:21.880 not just your syntax. 01:21.880 --> 01:23.940 You can literally just tell it what you want 01:23.940 --> 01:24.780 in plain English, 01:24.780 --> 01:27.660 and poof, it translates that into working code. 01:28.480 --> 01:30.040 It's a completely different level of partnership. 01:30.040 --> 01:32.900 And this is absolutely crucial. 01:32.900 --> 01:35.080 These things are not magic ones. 01:35.080 --> 01:37.760 They're not gonna make big architectural decisions for you 01:37.760 --> 01:40.800 or understand your company's unique business goals. 01:40.800 --> 01:42.460 They can suggest stuff for sure, 01:42.460 --> 01:44.800 but they don't replace human judgment. 01:44.800 --> 01:45.800 Think of it this way. 01:45.800 --> 01:48.160 The developer is still very much the pilot. 01:48.160 --> 01:51.420 The AI is the co-pilot, not the one flying the plane. 01:51.420 --> 01:53.520 And if you're wondering if this is just a fad, 01:53.520 --> 01:57.200 well, just look at this number, 76%. 01:57.220 --> 01:59.860 That's not some small group of early adopters. 01:59.860 --> 02:02.620 That's the vast majority of professional developers 02:02.620 --> 02:04.420 already using these tools. 02:04.420 --> 02:05.780 So yeah, the question isn't 02:05.780 --> 02:08.620 if teams will use them anymore, it's how. 02:08.620 --> 02:11.580 This is just part of the modern developers toolkit now. 02:11.580 --> 02:14.900 Okay, but where are we actually seeing the wins? 02:14.900 --> 02:16.940 This is where it gets really fascinating. 02:16.940 --> 02:19.540 For all that boring, repetitive stuff, 02:19.540 --> 02:21.560 writing the same old boilerplate code, 02:21.560 --> 02:22.900 churning out unit tests, 02:22.900 --> 02:25.780 the speed up is huge, a 50% boost. 02:25.800 --> 02:28.440 But when you get to the really gnarly problems, 02:28.440 --> 02:30.460 like hunting down a super tricky bug, 02:30.460 --> 02:32.480 the benefit is a lot smaller. 02:32.480 --> 02:33.640 So what's that tell us? 02:33.640 --> 02:36.440 The AI is an amazing accelerator for the grunt work, 02:36.440 --> 02:38.880 but it's not a silver bullet for the hard stuff. 02:38.880 --> 02:41.600 But all that speed, well, it comes with a catch, 02:41.600 --> 02:43.640 a kind of hidden tax, if you will. 02:43.640 --> 02:45.960 Because when a co-pilot spits out 50 lines of code 02:45.960 --> 02:47.720 in two seconds flat, guess what? 02:47.720 --> 02:50.560 A human still has to read it, understand it, 02:50.560 --> 02:51.480 and approve it. 02:51.480 --> 02:53.600 And that's where this 20% comes in. 02:53.620 --> 02:56.680 That's the extra time teams are spending on code review. 02:56.680 --> 02:59.760 You're basically trading typing time for thinking time, 02:59.760 --> 03:02.420 making sure what the AI wrote is actually correct, 03:02.420 --> 03:03.980 secure, and makes sense. 03:03.980 --> 03:06.480 Which brings us to probably the biggest, 03:06.480 --> 03:09.340 thorniest debate happening in engineering right now. 03:09.340 --> 03:12.820 Are these tools actually making our code better? 03:12.820 --> 03:15.220 Or are they quietly planting ticking time bombs 03:15.220 --> 03:16.640 in our code bases? 03:16.640 --> 03:19.420 And man, you can really see the tension here, right? 03:19.420 --> 03:22.880 On one side, the argument for better code is pretty strong. 03:22.900 --> 03:24.940 You can get everyone writing in a consistent style 03:24.940 --> 03:26.980 and just look at that jump in test coverage. 03:26.980 --> 03:28.340 That's huge. 03:28.340 --> 03:31.100 But then there's the other side of the coin. 03:31.100 --> 03:33.460 The danger of developers just copying and pasting code 03:33.460 --> 03:34.980 they don't really understand. 03:34.980 --> 03:37.060 Or even worse, the AI introducing 03:37.060 --> 03:38.620 these tiny, subtle security holes 03:38.620 --> 03:40.540 that look fine on the surface. 03:40.540 --> 03:42.300 It is the classic double-edged sword. 03:42.300 --> 03:45.500 Okay, so you've got this amazing promise on one hand 03:45.500 --> 03:47.740 and this pretty scary peril on the other. 03:47.740 --> 03:51.200 How do you actually rule this out without blowing things up? 03:51.200 --> 03:52.760 Because it's definitely not as simple 03:52.760 --> 03:56.060 as just giving everyone a license and saying, have fun. 03:56.060 --> 03:57.500 Well, the smart way to do it 03:57.500 --> 03:59.680 seems to be this kind of phased plan. 03:59.680 --> 04:01.920 You don't go big bang, you start small 04:01.920 --> 04:03.980 with a team that's actually excited about it 04:03.980 --> 04:05.940 and you measure everything. 04:05.940 --> 04:08.860 Then, and listen, this is the most important part. 04:08.860 --> 04:12.000 You have to change your process before you scale it out. 04:12.000 --> 04:13.900 Your old code review checklist, 04:13.900 --> 04:15.580 probably not gonna cut it anymore. 04:15.580 --> 04:17.980 You have to build a whole new support system for the tool, 04:17.980 --> 04:19.820 not just throw the tool at your old system 04:19.820 --> 04:20.900 and hope it works. 04:20.920 --> 04:24.200 So all of this, the new tools, the new processes, 04:24.200 --> 04:26.400 it leads to a pretty profound question. 04:26.400 --> 04:28.760 What does it even mean to be a developer now? 04:28.760 --> 04:30.040 Because let's be real, 04:30.040 --> 04:32.840 the skills that made you a rock star coder 10 years ago, 04:32.840 --> 04:35.600 they're not necessarily the ones that matter most today. 04:35.600 --> 04:38.560 So what skills do matter? 04:38.560 --> 04:41.200 The value is shifting big time. 04:41.200 --> 04:43.800 It's becoming less about how fast you can type 04:43.800 --> 04:46.240 or whether you've memorized every piece of syntax. 04:46.240 --> 04:49.320 It's now all about the things the AI can't do. 04:49.320 --> 04:50.700 Can you think about the big picture? 04:50.720 --> 04:51.840 The architecture. 04:51.840 --> 04:54.160 Can you ask the AI smart questions? 04:54.160 --> 04:55.640 And can you look at what it gives you back 04:55.640 --> 04:58.120 and say, nope, that's not quite right? 04:58.120 --> 05:01.100 Basically, the code itself is becoming a commodity. 05:01.100 --> 05:04.000 The real value is in the thinking behind the code. 05:04.000 --> 05:07.000 And that really brings us to the final critical question 05:07.000 --> 05:09.320 you have to ask yourself and your team. 05:09.320 --> 05:11.720 AI co-pilots are here, they're powerful, 05:11.720 --> 05:12.980 and they're not going anywhere. 05:12.980 --> 05:13.820 But at the end of the day, 05:13.820 --> 05:17.960 their impact depends entirely on how you choose to use them. 05:17.960 --> 05:19.880 Are you just using them to type code faster? 05:19.900 --> 05:22.660 Or are you using them to free up your brain to think bigger, 05:22.660 --> 05:25.100 to focus on building truly great systems? 05:25.100 --> 05:27.020 Because that choice right there, 05:27.020 --> 05:28.900 that's what's going to define the next era 05:28.900 --> 05:30.100 of software development.