Constraining Agents for Software Development with Don Demcsak 的完整文字稿

文字稿 🔗

WEBVTT 1 00:00:01.169 --> 00:00:04.790 How'd you like to listen to. NET Rocks with no ads? Easy. 2 00:00:05.370 --> 00:00:08.769 Become a patron. For just $ 5 a month, you get 3 00:00:08.890 --> 00:00:11.720 access to a private RSS feed where all the shows 4 00:00:11.800 --> 00:00:14.900 have no ads. $ 20 a month will get you that 5 00:00:15.119 --> 00:00:18.839 and a special. NET Rocks patron mug. Sign up now 6 00:00:18.899 --> 00:00:36.549 at patreon.dotnetrocks.com. Hey, and welcome back to. NET Rocks. I'm 7 00:00:36.590 --> 00:00:37.210 Carl Franklin. 8 00:00:37.289 --> 00:00:37.950 And I'm Richard Kaplan. 9 00:00:38.049 --> 00:00:41.250 Our friend Don XML, I mean, Don Demsack is here. 10 00:00:41.270 --> 00:00:41.530 Oh, jeez. 11 00:00:42.170 --> 00:00:44.590 Or is he Don Jason? Are you Don Jason now? 12 00:00:45.109 --> 00:00:45.770 That's my brother. 13 00:00:48.700 --> 00:00:51.880 At least he didn't say Don Yammel. God knows. No. 14 00:00:51.899 --> 00:00:52.659 No. 15 00:00:53.479 --> 00:00:58.119 This is going to be a good one. This is 16 00:00:58.259 --> 00:01:01.460 episode 2018. And as we always do at the top 17 00:01:01.500 --> 00:01:04.040 of the show, let's talk about things that happened that year. 18 00:01:05.120 --> 00:01:07.799 I get all the depressing ones and Richard gets all 19 00:01:07.840 --> 00:01:09.760 the happy ones in space and tech. 20 00:01:09.780 --> 00:01:11.459 Just have to pick carefully. 21 00:01:11.480 --> 00:01:15.849 Yeah, I guess so. But, you know, these are important events, right? 22 00:01:15.950 --> 00:01:16.250 Are they? 23 00:01:16.390 --> 00:01:17.310 School shootings. 24 00:01:17.750 --> 00:01:18.030 Are they? 25 00:01:18.170 --> 00:01:20.689 Robert Mueller's Russia investigation. 26 00:01:20.909 --> 00:01:21.290 Oh, yeah. 27 00:01:21.469 --> 00:01:21.689 Well. 28 00:01:21.730 --> 00:01:25.590 Brett Kavanaugh, Supreme Court confirmation battle. 29 00:01:25.689 --> 00:01:25.870 Yeah. 30 00:01:25.930 --> 00:01:30.620 The 2018 midterm elections when the Democrats took over. The 31 00:01:30.659 --> 00:01:32.640 Democrats strike back. Did they? 32 00:01:32.799 --> 00:01:33.079 Okay. 33 00:01:33.400 --> 00:01:33.840 Well, okay. 34 00:01:33.959 --> 00:01:34.239 I don't know. 35 00:01:37.599 --> 00:01:40.829 Trump and Kim Jong-un, still crazy after all these years. 36 00:01:41.269 --> 00:01:41.549 Nice. 37 00:01:42.030 --> 00:01:42.590 Still friends. 38 00:01:42.609 --> 00:01:45.290 I think that was one of the first meetings, wasn't it? 39 00:01:45.709 --> 00:01:46.290 Yeah, yeah. 40 00:01:46.670 --> 00:01:54.019 June 12th in Singapore. Jamal Khashoggi, as Trump said, these 41 00:01:54.040 --> 00:02:00.260 things happen. He was killed and dissolved in a bathtub 42 00:02:00.400 --> 00:02:00.959 of acid. 43 00:02:01.219 --> 00:02:01.659 Or something. 44 00:02:01.680 --> 00:02:02.040 Okay. Okay. 45 00:02:02.849 --> 00:02:07.480 Or something. That's what we think. California wildfires happens every year, 46 00:02:07.560 --> 00:02:11.520 but 2018 was particularly bad. The Me Too movement and 47 00:02:11.539 --> 00:02:15.639 the Larry Nassar scandal. And just to put a little 48 00:02:15.659 --> 00:02:17.879 fine point on it, hurricanes Florence and Michael. 49 00:02:18.120 --> 00:02:18.280 Right. 50 00:02:18.439 --> 00:02:19.000 Those are good ones. 51 00:02:19.120 --> 00:02:21.139 Catastrophic flooding in the Carolinas. 52 00:02:21.319 --> 00:02:21.590 Yeah. 53 00:02:21.840 --> 00:02:24.650 And the Florida panhandle. All right, Richard. That's enough. 54 00:02:24.889 --> 00:02:28.449 Well, it was also, remember that cave rescue in Thailand? Oh, yeah. 55 00:02:28.750 --> 00:02:31.449 Those students went in and there was the flash flood. 56 00:02:31.530 --> 00:02:32.289 That's when that happened. 57 00:02:32.310 --> 00:02:33.210 That was a good story. 58 00:02:33.430 --> 00:02:33.789 It was. 59 00:02:33.870 --> 00:02:36.129 I mean, one of the rescuers lost their lives trying 60 00:02:36.150 --> 00:02:38.909 to save those kids. Right. But they were successful. Do 61 00:02:39.270 --> 00:02:42.090 you remember when, I think that's when Elon started going crazy. Yeah. 62 00:02:42.189 --> 00:02:44.509 Remember he really wanted to be involved in that. It's like, 63 00:02:44.530 --> 00:02:48.580 you're not actually qualified. I'll build a submersible. It's like, A, 64 00:02:48.590 --> 00:02:50.719 you don't know how to build a submersible. And B, 65 00:02:51.280 --> 00:02:53.939 that's not going to work in a cave either. And 66 00:02:53.960 --> 00:02:56.759 then he got really snarky and started calling people pedophiles 67 00:02:56.800 --> 00:02:59.599 and things. Yeah. I don't know. It was one of 68 00:02:59.620 --> 00:03:02.969 those moments because this is the same year that the 69 00:03:03.030 --> 00:03:05.370 Falcon Heavy flies for the first time and Elon put 70 00:03:05.409 --> 00:03:09.430 his red Tesla Roadster on top of it as the payload, 71 00:03:09.949 --> 00:03:11.250 which was pretty cool. 72 00:03:11.330 --> 00:03:13.930 I thought that was the cool Elon doing that. 73 00:03:13.969 --> 00:03:17.580 Yeah. But it's interesting to think that's the same year 74 00:03:17.639 --> 00:03:21.259 he goes nuts on this cave rescue thing so i 75 00:03:21.280 --> 00:03:22.439 mean things are just getting. 76 00:03:22.240 --> 00:03:27.719 Weirder and weirder yeah weirdness weirdness abides um also i 77 00:03:27.759 --> 00:03:31.370 want to mention aretha franklin and anthony bourdain died in 78 00:03:31.370 --> 00:03:35.210 2018 yeah very sad also paul allen very sad about 79 00:03:35.250 --> 00:03:39.310 have you there's a new movie like a documentary about anthony. 80 00:03:39.189 --> 00:03:41.930 There's an anthony bourdain movie coming out imminently yeah so 81 00:03:41.949 --> 00:03:43.500 it hasn't come out yet I think it's just out 82 00:03:43.520 --> 00:03:44.319 brand new. Okay. 83 00:03:44.520 --> 00:03:45.240 Yeah. I want to see that. 84 00:03:45.300

Load full transcript

--> 00:03:45.479 Yeah. 85 00:03:45.740 --> 00:03:49.169 Also George H.W. Bush and John McCain died. Yeah. 86 00:03:49.370 --> 00:03:51.830 All right, Richard, save us from this misery. So I 87 00:03:52.129 --> 00:03:54.270 mentioned the Falcon heavy flight, which, yeah, that was pretty cool. 88 00:03:54.289 --> 00:03:56.090 We ended up doing a geek out about it. I mean, 89 00:03:56.110 --> 00:03:59.229 it was just an extraordinary rocket strapping three side by side. 90 00:04:00.210 --> 00:04:03.030 Originally he had planned on doing something called asparagus staging, 91 00:04:03.060 --> 00:04:05.060 where he was actually going to feed the fuel from 92 00:04:05.120 --> 00:04:08.610 the wing of boosters into the center so that when 93 00:04:08.669 --> 00:04:12.050 the outer boosters were empty and separated, the center booster 94 00:04:12.069 --> 00:04:15.849 would be completely full. Never happened. Instead, they do it 95 00:04:15.889 --> 00:04:18.149 the simpler way, which is simply to throttle down the 96 00:04:18.189 --> 00:04:21.319 center booster so it doesn't burn as much fuel. That 97 00:04:21.360 --> 00:04:23.740 being said, Falcon Heavy, you'd be never real happy with 98 00:04:23.779 --> 00:04:26.720 it The simple reason that it has a lot of power, 99 00:04:26.759 --> 00:04:28.860 but it doesn't have a lot of payload space. And 100 00:04:28.879 --> 00:04:31.959 so there's relatively few cases where it's needed. And he 101 00:04:31.980 --> 00:04:35.250 was already talking about what we now know is Starship, uh, 102 00:04:35.370 --> 00:04:38.250 and starting to do prototypes around all of that. Uh, 103 00:04:38.259 --> 00:04:42.310 we have a lot of cool, uh, satellites up in 2018, 104 00:04:42.310 --> 00:04:46.750 starting with tests, the transiting exoplanet survey satellite. So this 105 00:04:46.769 --> 00:04:51.129 was the only, the third of the dedicated finding exoplanet 106 00:04:51.149 --> 00:04:55.069 satellites that the, the successor of Kepler. Kepler had launched 107 00:04:55.100 --> 00:04:59.240 back in 2009 and found 2,600 exoplanets. That was Kepler's 108 00:04:59.300 --> 00:05:02.620 mission was to do deep surveys to see how common 109 00:05:02.839 --> 00:05:06.759 are planets. And the answer is very common. Most stars 110 00:05:06.839 --> 00:05:11.100 have planets. TESS as a survey satellite is not designed 111 00:05:11.139 --> 00:05:15.160 to observe as closely, but to cover much larger tracks 112 00:05:15.279 --> 00:05:19.040 of the sky. And fast forwarding eight years now, it's 113 00:05:19.079 --> 00:05:23.759 found over 8,000 potential objects that generally need additional surveying. 114 00:05:24.269 --> 00:05:28.790 because it is a survey spacecraft. But of those 8,000, 115 00:05:28.790 --> 00:05:31.629 almost 1,000 have not been confirmed as exoplanets. So if 116 00:05:31.649 --> 00:05:36.689 you're keeping count, we're up around 6,300 confirmed exoplanets so 117 00:05:36.750 --> 00:05:41.569 far from all of our different collection methodologies. This is 118 00:05:41.610 --> 00:05:44.550 not how Star Trek was supposed to work. Remember in 119 00:05:44.569 --> 00:05:46.470 Star Trek, you had to fly to the solar system 120 00:05:46.490 --> 00:05:48.250 to see the planets? No, we're going to have a 121 00:05:48.300 --> 00:05:50.620 map of all of them long before we can get 122 00:05:50.639 --> 00:05:53.579 to any of them. There's a couple of asteroid missions. 123 00:05:53.639 --> 00:05:59.759 Hayabusa 2 in June got to Rigu. And in December, 124 00:06:00.420 --> 00:06:03.420 OSIRIS-REx gets to Bennu, which is cool. We start to 125 00:06:03.480 --> 00:06:06.379 explore asteroids sort of on a regular basis. In August, 126 00:06:06.430 --> 00:06:09.870 the Parker Solar Probe, the first time we've ever tried 127 00:06:09.910 --> 00:06:13.310 to fly a satellite to the sun. to actually do 128 00:06:13.370 --> 00:06:16.769 measurements of its outer atmosphere. That's a very difficult thing 129 00:06:16.810 --> 00:06:18.550 to do. You have to go really, really fast to 130 00:06:18.569 --> 00:06:18.810 do that. 131 00:06:18.829 --> 00:06:21.269 Remember Richard Pryor's bit, the first man to walk on 132 00:06:21.310 --> 00:06:21.649 the sun? 133 00:06:21.670 --> 00:06:25.009 Ooh, ah, ah, ooh, how? 134 00:06:26.180 --> 00:06:28.720 So to make the solar probe go fast enough to 135 00:06:28.759 --> 00:06:30.240 get that close to the sun, they launched on a 136 00:06:30.259 --> 00:06:32.620 Delta IV, which is one of the highest performance rockets 137 00:06:32.660 --> 00:06:36.339 ever made. It's largely obsolete now and retired and broke 138 00:06:36.360 --> 00:06:37.740 a bunch of speed records to try and get it 139 00:06:37.779 --> 00:06:40.259 down there. It'll take a while to actually get close 140 00:06:40.300 --> 00:06:42.019 enough and actually start to do sampling of the sun. 141 00:06:42.620 --> 00:06:45.439 Also in August was when on the International Space Station, 142 00:06:45.540 --> 00:06:50.279 Soyuz MS-09, this is the vehicle that they transport astronauts 143 00:06:50.339 --> 00:06:52.300 to and from the space station, which at that time, 144 00:06:52.319 --> 00:06:55.990 because the shuttle is no longer flying and Crew Dragon 145 00:06:56.019 --> 00:06:58.610 doesn't exist yet, it's still in development, was the only 146 00:06:58.649 --> 00:07:00.990 way to get to the station. They find a two 147 00:07:01.089 --> 00:07:05.389 millimeter hole in the orbital module of the Soyuz. And 148 00:07:05.410 --> 00:07:07.230 there was a lot of controversy about whether it was 149 00:07:07.350 --> 00:07:09.750 drilled because it was sort of a perfect hole or 150 00:07:09.790 --> 00:07:12.569 was it just that right hole? you know, penetrator that 151 00:07:12.610 --> 00:07:15.189 it went through there. Anyway, fixed with a piece of 152 00:07:15.509 --> 00:07:16.430 cape on tape. 153 00:07:16.449 --> 00:07:17.430 Could have been mice. 154 00:07:17.750 -->

00:07:19.910 You know, and because it was the Orbridge module, it 155 00:07:19.930 --> 00:07:21.829 doesn't have to survive re-entry, so it's not that big 156 00:07:21.850 --> 00:07:23.509 of a deal. But yeah, it kicked up a lot 157 00:07:23.550 --> 00:07:27.569 of speculation. And if that wasn't enough, the very next 158 00:07:27.649 --> 00:07:31.250 Soyuz in October, this is Soyuz MS-10, it was only 159 00:07:31.290 --> 00:07:33.889 carrying two passengers. Normally, the Soyuz can carry three, but 160 00:07:33.930 --> 00:07:39.569 occasionally fly with less. In this case, it had Alexei Ovichin, And, uh, 161 00:07:39.680 --> 00:07:42.740 from Roscosmos and Nick Hague from NASA, they were going 162 00:07:42.759 --> 00:07:44.699 to be part of the exhibition 57 to the international 163 00:07:44.720 --> 00:07:49.160 space station. But two minutes after launch, there was a failure. 164 00:07:49.399 --> 00:07:52.199 And Soyuz is one of the most reliable, most flown 165 00:07:52.220 --> 00:07:53.500 spacecraft of all time. 166 00:07:53.579 --> 00:07:53.839 Yeah. 167 00:07:53.899 --> 00:07:56.220 This was one of the very rare failures and it 168 00:07:56.240 --> 00:08:00.189 was a failure that had never occurred before. There was 169 00:08:00.199 --> 00:08:02.810 something wrong in the sensor of one of the strap-on boosters. 170 00:08:02.829 --> 00:08:05.449 There are four strap-on boosters that famously separate from a 171 00:08:05.490 --> 00:08:08.170 Soyuz and do what they call the Korolev cross. As 172 00:08:08.189 --> 00:08:10.649 they come flying off. And it didn't happen this time. 173 00:08:10.709 --> 00:08:12.459 And the one booster hung up. And then when it 174 00:08:12.500 --> 00:08:17.360 did break away, it struck the central booster core. And 175 00:08:17.420 --> 00:08:20.560 that ended the going into space. As Nick Hague said, 176 00:08:20.620 --> 00:08:22.860 as soon as the bouncing around started, I guess we're 177 00:08:22.899 --> 00:08:27.139 not going to the space station. But at this point, 178 00:08:27.160 --> 00:08:29.639 they're already 50 kilometers up. And that's too high for 179 00:08:29.699 --> 00:08:33.659 the escape tower. So it's already been ejected. But there is... 180 00:08:34.269 --> 00:08:38.549 Mode 2, Ascent Abort, a system that had been tested 181 00:08:38.690 --> 00:08:41.899 but never used. And this was the first and only 182 00:08:41.940 --> 00:08:42.919 time it's ever been used. 183 00:08:43.100 --> 00:08:45.159 So they basically turned it around and re-entered? 184 00:08:45.259 --> 00:08:47.620 It's more complicated than that because you've got an exploding 185 00:08:47.679 --> 00:08:49.679 rocket behind you, so you don't want to be near that. 186 00:08:49.750 --> 00:08:52.789 So instead of abort motors fire and push the Soyuz 187 00:08:52.830 --> 00:08:56.049 assembly away from the booster, and then you have to 188 00:08:56.210 --> 00:09:00.370 get into... descent mode because you're carrying the propulsion module 189 00:09:00.409 --> 00:09:03.309 as well as descent module and the orbital module so 190 00:09:03.370 --> 00:09:05.370 those each have to be pushed off and then it 191 00:09:05.409 --> 00:09:07.480 has to switch over to descent mode get the heat 192 00:09:07.500 --> 00:09:12.259 shield pointed down and actually land because of the angle 193 00:09:12.279 --> 00:09:14.259 of ascent they only get to about 90 kilometers so 194 00:09:14.279 --> 00:09:16.059 they spend some time in free fall still going up 195 00:09:16.120 --> 00:09:17.139 the explosion happened. 196 00:09:16.940 --> 00:09:19.620 At 50 dude you should make movies This is great. 197 00:09:20.179 --> 00:09:20.799 It's crazy. 198 00:09:21.000 --> 00:09:22.580 It's like I'm on the edge of my seat. 199 00:09:22.679 --> 00:09:25.000 And they're in free fall, and now they have to reorient, 200 00:09:25.019 --> 00:09:27.110 and they do a ballistic return, which means those poor 201 00:09:27.120 --> 00:09:30.950 guys pulled about seven Gs as they came back down. 202 00:09:32.289 --> 00:09:35.830 Worked perfectly. Both astronauts walked away. The crazy part is 203 00:09:35.870 --> 00:09:40.889 realizing from launch at Balkanur to on the ground about 204 00:09:40.889 --> 00:09:44.019 200 kilometers downrange in Kazakhstan, 19 minutes. 205 00:09:44.419 --> 00:09:44.879 Oh, my God. 206 00:09:45.289 --> 00:09:45.710 Goodness. 207 00:09:46.029 --> 00:09:49.379 It was fast and it's all automated. Like basically they 208 00:09:49.399 --> 00:09:52.659 were having trouble keeping up with the control systems working 209 00:09:52.700 --> 00:09:53.000 through it. 210 00:09:53.059 --> 00:09:55.259 That's definitely not enough time for a feature film. 211 00:09:55.470 --> 00:09:57.970 No, it's pretty quick, but I'm sure there's a bit 212 00:09:57.990 --> 00:09:59.970 of story to go around it. Anyway, it worked flawlessly. 213 00:10:00.659 --> 00:10:03.019 and they would eventually get to space on a different 214 00:10:03.039 --> 00:10:08.600 spacecraft because that one weren't no good no more. In November, 215 00:10:08.740 --> 00:10:10.899 we have the 20th anniversary of the International Space Station 216 00:10:10.919 --> 00:10:14.580 because the Zarya module went up in 1998. We're also 217 00:10:14.679 --> 00:10:18.299 fully over into commercial resupply now. So Roscoff's mostly due 218 00:10:18.350 --> 00:10:22.570 to five deliveries in 2018. But SpaceX does three on 219 00:10:22.629 --> 00:10:26.370 Cargo Dragon. Orbital ATK does three with Cygnus. And JAXA 220 00:10:26.409 --> 00:10:31.909 has one HTV delivery. So a total of, what is that, 12. 221 00:10:31.909 --> 00:10:34.450 And also in November, the InSight lander lands on Mars. 222 00:10:34.509 --> 00:10:37.200 First thing to have landed on Mars since Perseverance. And 223 00:10:37.299 --> 00:10:41.440 it's doing measurements of Mars quakes and interior mapping of Mars. 224 00:10:41.990 --> 00:10:45.990 by looking at reverberations that

travel through the planet. And finally, 225 00:10:46.110 --> 00:10:50.720 in December, China launches the Chang'e 4 relay satellite in 226 00:10:50.840 --> 00:10:53.779 preparation for their landing, putting a lander and rover down 227 00:10:53.799 --> 00:10:55.480 on the far side of the moon. They need a 228 00:10:55.519 --> 00:10:57.159 satellite to be able to communicate with the far side 229 00:10:57.179 --> 00:10:59.460 of the moon. So that goes up in 2018. The 230 00:10:59.500 --> 00:11:02.539 big mission will be in 2019. Ready for compute stories? 231 00:11:02.879 --> 00:11:03.200 Yes. 232 00:11:03.720 --> 00:11:05.379 And you always think I have to give you all 233 00:11:05.399 --> 00:11:07.620 the good news. So let's start off in January with 234 00:11:07.720 --> 00:11:10.000 Meltdown Inspector. Okay. 235 00:11:10.039 --> 00:11:10.159 Okay. 236 00:11:10.809 --> 00:11:18.870 So these were discovered potential exploits using a speculative execution 237 00:11:18.889 --> 00:11:23.700 that exists in both AMD and Intel processors. And there 238 00:11:23.740 --> 00:11:25.740 was no security around that. So the theory was that 239 00:11:25.759 --> 00:11:27.879 if you did stuff exactly right, you might be able 240 00:11:27.899 --> 00:11:30.740 to grab data from another virtual machine that was running through. 241 00:11:31.100 --> 00:11:34.179 This exploit has never been found in the wild. Even 242 00:11:34.259 --> 00:11:36.519 attempts to try and utilize it have been largely futile. 243 00:11:36.559 --> 00:11:39.990 But at the time, they really freaked out. And putting 244 00:11:40.029 --> 00:11:44.860 security rules in place caused substantial performance penalties. I think 245 00:11:44.879 --> 00:11:48.720 to talk about literally 20% of available compute disappeared to 246 00:11:48.759 --> 00:11:50.200 fix spectrum and meltdown. 247 00:11:50.480 --> 00:11:50.960 Not good. 248 00:11:51.000 --> 00:11:51.679 Remember that? 249 00:11:51.720 --> 00:11:52.019 Wow. 250 00:11:52.399 --> 00:11:54.759 And adding to my, I don't only have good news 251 00:11:55.039 --> 00:11:59.379 stories in 2018, Google officially removes their don't be evil 252 00:11:59.460 --> 00:12:02.100 from their code of conduct. I mean, they didn't add 253 00:12:02.120 --> 00:12:04.139 be evil, but they just took out the don't be. 254 00:12:04.759 --> 00:12:08.919 And then in March, uh, The Cambridge Analytica scandal. Oh, yes. 255 00:12:09.259 --> 00:12:12.000 Where Cambridge Analytica had mined Facebook data in a way 256 00:12:12.019 --> 00:12:15.220 that not even Facebook had thought about and was very 257 00:12:15.279 --> 00:12:19.620 effective at doing things like creating ads for individuals where 258 00:12:19.639 --> 00:12:22.649 the individual didn't realize they're the only ones seeing the ad. 259 00:12:22.970 --> 00:12:26.389 A very effective strategy for getting a group of English 260 00:12:26.429 --> 00:12:29.110 people to vote against their best interests and exit the EU. 261 00:12:29.269 --> 00:12:31.230 Now we have Facebook and nobody cares. 262 00:12:31.250 --> 00:12:32.169 There you go. Yeah. 263 00:12:32.330 --> 00:12:32.929 Targeted ad. 264 00:12:33.029 --> 00:12:37.110 And in April, Belgium declares... video game loot box systems 265 00:12:37.149 --> 00:12:40.590 to be gambling and illegal for minors, which it is. 266 00:12:40.610 --> 00:12:46.549 I don't disagree with them. Interesting problem. Yeah. Don't manipulate children, please. Well, 267 00:12:46.889 --> 00:12:50.029 not too heavily. And then in May, the GDPR is enacted. 268 00:12:50.490 --> 00:12:53.210 It's been in development for a few years. This is 269 00:12:53.250 --> 00:12:56.129 why you have to keep agreeing to cookies as well 270 00:12:56.169 --> 00:12:57.500 as a bunch of other things like the right to 271 00:12:57.529 --> 00:13:02.840 be forgotten and serious fines for not being forthright with breaches. 272 00:13:03.120 --> 00:13:03.299 Yeah. 273 00:13:03.639 --> 00:13:05.500 Um, over on the Microsoft side of things, it may, 274 00:13:05.519 --> 00:13:10.570 we have build 2018, which ships.net core 2.1. But also, 275 00:13:10.590 --> 00:13:12.429 and I realized this in hindsight, when I was putting 276 00:13:12.450 --> 00:13:15.210 this together is sort of the first wave of AI stuff. 277 00:13:15.230 --> 00:13:17.710 Remember visual studio and telecode? I think we did a 278 00:13:17.769 --> 00:13:18.269 show on it. 279 00:13:18.370 --> 00:13:18.830 Yeah. Yeah. 280 00:13:18.850 --> 00:13:18.990 Yeah. 281 00:13:19.090 --> 00:13:21.429 Where it wasn't an, you know, we're really talking about having, 282 00:13:21.529 --> 00:13:24.320 helping to generate code for testing and things. It's also 283 00:13:24.360 --> 00:13:27.940 the year they ship ml.net machine learning.net for, for doing 284 00:13:27.980 --> 00:13:28.700 models that way. 285 00:13:28.720 --> 00:13:29.059 Right. 286 00:13:29.350 --> 00:13:32.409 And certainly last but not least, for the build announcements, 287 00:13:32.480 --> 00:13:35.059 one of the ones that I jumped over was Azure Sphere. 288 00:13:35.120 --> 00:13:37.620 Do you remember this? This is when Microsoft starts building 289 00:13:37.659 --> 00:13:41.500 their own flavor of a Linux operating system for IoT. 290 00:13:42.570 --> 00:13:45.370 It's subsequently been phased out now, but for a few 291 00:13:45.409 --> 00:13:48.149 years here, it's like, Microsoft makes their own Linux. I 292 00:13:48.210 --> 00:13:48.549 love that. 293 00:13:48.710 --> 00:13:48.870 Mm-hmm. 294 00:13:49.490 --> 00:13:51.210 Later in the year, in October, well, actually it was 295 00:13:51.250 --> 00:13:54.320 announced in June, but completed in October, Microsoft acquires GitHub 296 00:13:54.340 --> 00:13:59.120 for $ 7. 5 billion. And our friend, Nat Friedman, formerly of Xamarin, 297 00:13:59.179 --> 00:14:02.679 now a Microsoft employee, becomes the new CEO of GitHub. 298 00:14:02.860 --> 00:14:04.700 We get Core 2.2 at the

end of the year 299 00:14:04.720 --> 00:14:05.600 in December. Yeah. 300 00:14:05.779 --> 00:14:07.360 A couple other compute stories. 301 00:14:08.019 --> 00:14:10.649 This is the year that Elon Musk attempts to take 302 00:14:10.750 --> 00:14:13.529 over OpenAI from his partners and gets booted out of it. 303 00:14:13.529 --> 00:14:15.509 And he's still angry about that and purse fights with 304 00:14:15.549 --> 00:14:16.769 Sam Altman on regular basis. 305 00:14:16.950 --> 00:14:19.110 Yeah, because... That's where he is right now. 306 00:14:19.309 --> 00:14:22.539 Because reasons, because I realized all the same, the same 307 00:14:22.600 --> 00:14:25.759 year that he flies the biggest rocket in history into 308 00:14:25.799 --> 00:14:28.860 space and puts a sports car into space, which still orbiting, 309 00:14:28.879 --> 00:14:32.259 you know, between the earth and Mars also, you know, 310 00:14:32.370 --> 00:14:36.649 call some guy trying to save children in Thailand, a pedophile, uh, 311 00:14:36.730 --> 00:14:39.899 also tries to hijack the AI company. Yeah. Uh, We're 312 00:14:39.940 --> 00:14:43.210 coming into the period of excitement as we're going through 313 00:14:43.230 --> 00:14:47.470 these next few years in quantum computing. And in 2018, 314 00:14:47.470 --> 00:14:50.610 there was a very important paper published, the NISC paper, 315 00:14:50.669 --> 00:14:54.250 which means Noisy Intermediate Scale Quantum, this paper by John Preskill. 316 00:14:55.149 --> 00:14:59.549 This is the point where companies are starting to build qubits, 317 00:14:59.929 --> 00:15:03.289 build these chandelier dipped in liquid helium style computers that 318 00:15:03.309 --> 00:15:06.250 have about 50 qubits or so forth. And in this paper, 319 00:15:06.269 --> 00:15:08.490 he talks about it's all going to be about error correction. 320 00:15:09.220 --> 00:15:10.860 that we're getting out of the lab. These things are 321 00:15:10.879 --> 00:15:12.340 going to be able to do certain levels of compute, 322 00:15:12.360 --> 00:15:15.610 but without really fault tolerance in this, you're not going 323 00:15:15.629 --> 00:15:20.049 to be able to get much done. And that sets 324 00:15:20.149 --> 00:15:22.529 up Google, who's driving a lot of this stuff with 325 00:15:22.570 --> 00:15:26.549 their quantum toolkit, aimed at researchers for testing what prototype 326 00:15:26.570 --> 00:15:28.940 hardware could do, which is funny because next year they'll 327 00:15:29.240 --> 00:15:32.919 actually produce the second market process. So obviously Google was anticipated. 328 00:15:33.419 --> 00:15:33.919 And that's what I got. 329 00:15:34.250 --> 00:15:34.570 All right. 330 00:15:34.610 --> 00:15:38.730 Well, before we bring Don on, let's talk about Better 331 00:15:38.750 --> 00:15:40.149 Know Framework. Roll the music. 332 00:15:40.230 --> 00:15:40.490 Awesome. 333 00:15:40.809 --> 00:15:43.870 All right, man. 334 00:15:44.230 --> 00:15:51.399 What do you got? Guess who's back in Better Know Framework? 335 00:15:51.539 --> 00:15:57.080 I think I know. He's Australian. It's Simon. 336 00:15:57.320 --> 00:15:57.860 It's Simon. 337 00:15:58.539 --> 00:16:05.009 So he belongs to our Appfee Next Slack channel. And 338 00:16:05.210 --> 00:16:08.519 he basically knew that I had done some work into 339 00:16:08.600 --> 00:16:14.480 looking into serializing link queries and passing them from client 340 00:16:14.519 --> 00:16:18.820 to server and pretty much gave up. And there was 341 00:16:18.860 --> 00:16:23.049 a whole smattering of GitHub repos out there of people 342 00:16:23.090 --> 00:16:26.149 who attempted it. And I tried like 50% of them 343 00:16:26.169 --> 00:16:29.090 and they all failed. And so he came up with 344 00:16:29.129 --> 00:16:33.559 this thing called SCRY, S-C-R-Y, or maybe it's SCARY. I 345 00:16:33.580 --> 00:16:33.799 don't know. 346 00:16:33.879 --> 00:16:34.559 Or Scree. 347 00:16:34.759 --> 00:16:35.379 Or Scree? 348 00:16:35.440 --> 00:16:35.700 Yeah. 349 00:16:35.779 --> 00:16:38.860 He doesn't have a pronunciation legend, but he says it's 350 00:16:38.919 --> 00:16:43.070 a type-safe serializable link from a client to a server 351 00:16:43.159 --> 00:16:47.470 side EF core model. And I had told him, and 352 00:16:47.750 --> 00:16:52.590 both Mateus Carvalho and I had worked on this before, 353 00:16:52.629 --> 00:16:56.169 and Mateus came up with something, and I came up 354 00:16:56.190 --> 00:17:01.740 with something, AVN repository, that basically tells Did the same, 355 00:17:01.779 --> 00:17:07.279 but I used classes and serialized those classes for filtering 356 00:17:07.380 --> 00:17:11.099 and basically all aspects of a SQL query, a link query, right? 357 00:17:11.640 --> 00:17:14.660 A where, an order by, all of that stuff just 358 00:17:14.720 --> 00:17:16.269 turned into data and passed in. 359 00:17:16.289 --> 00:17:16.809 And it works. 360 00:17:17.349 --> 00:17:23.150 But as Simon pointed out, it's not as efficient as 361 00:17:23.190 --> 00:17:26.289 what he's doing. Yeah. So this is, and he has a, 362 00:17:26.950 --> 00:17:28.650 there's two links. We're going to give you a link 363 00:17:28.670 --> 00:17:32.109 to the scary or the scry or scree project. And 364 00:17:32.170 --> 00:17:36.549 then there's this doc querying MD composing a query at runtime. 365 00:17:37.990 --> 00:17:41.079 And so he just talks about how to do that, 366 00:17:41.150 --> 00:17:47.480 how to create projections and computer projection members. It's really cool. And, 367 00:17:47.700 --> 00:17:49.039 and it works interesting. 368 00:17:49.180 --> 00:17:50.299 And so, yeah. 369 00:17:50.420 --> 00:17:54.960 And, and so, yeah. Um, Simon, you win the prize. 370 00:17:55.220 --> 00:17:55.960 You're awesome. 371 00:17:56.319 --> 00:17:59.240 You're the first person I know who has actually done this. 372 00:17:59.849 --> 00:18:02.930 And maybe I just resigned and said, I can't be done.

373 00:18:02.990 --> 00:18:06.190 And maybe it was done, you know, long before that. 374 00:18:06.309 --> 00:18:06.670 I don't know. 375 00:18:06.950 --> 00:18:09.190 But that reminds me of link to rest. 376 00:18:09.650 --> 00:18:12.690 Oh, link to rest. Well, that sounds exactly like what 377 00:18:12.730 --> 00:18:13.230 I was doing. 378 00:18:13.410 --> 00:18:14.049 Yeah. Yeah. 379 00:18:14.069 --> 00:18:16.170 Because it was the inverse. 380 00:18:16.430 --> 00:18:16.670 Right. 381 00:18:16.769 --> 00:18:21.059 Where you, well, you created from a link query, you 382 00:18:21.220 --> 00:18:23.740 could create a query string. 383 00:18:23.980 --> 00:18:24.269 Right. 384 00:18:24.420 --> 00:18:26.119 But you couldn't put it back on the other side 385 00:18:26.140 --> 00:18:28.750 and turn it into a link query, which is, which 386 00:18:28.789 --> 00:18:30.529 is what he's doing. And what I wanted to do 387 00:18:30.589 --> 00:18:34.109 as well, right? Because you've got a form and it's 388 00:18:34.150 --> 00:18:36.289 got a whole bunch of search parameters. And if your 389 00:18:36.309 --> 00:18:38.289 data is local, great link query. 390 00:18:38.789 --> 00:18:41.009 Well, no, so it did it the other way too. 391 00:18:41.450 --> 00:18:42.690 If it's behind an API. 392 00:18:43.529 --> 00:18:48.039 Link to REST, it would, it was an ASP.NET. It 393 00:18:48.059 --> 00:18:52.369 was NBC, ASP.NET. It would, take that query string and 394 00:18:52.410 --> 00:18:54.630 then turn it into a link. 395 00:18:54.950 --> 00:19:00.160 Oh, okay. Well, that's great. Link to rest, rest to link. Okay. Well, 396 00:19:00.339 --> 00:19:02.960 check out scary or scree or scry or whatever it is. 397 00:19:03.960 --> 00:19:06.349 And ask Simon how to pronounce this thing. 398 00:19:06.369 --> 00:19:06.990 Cause I don't know. 399 00:19:07.509 --> 00:19:08.990 I'm sure there's a difference, but yeah. 400 00:19:09.109 --> 00:19:09.369 Right. 401 00:19:09.609 --> 00:19:11.509 I just remember link to rest. 402 00:19:11.930 --> 00:19:13.910 Well, Richard, who's talking to us today? 403 00:19:14.029 --> 00:19:16.650 Grabbed a comment off show 1994, the one we did 404 00:19:16.670 --> 00:19:19.470 with our friend Brady Gaster, talking about Squad. A lot 405 00:19:19.490 --> 00:19:20.630 of good comments on that show. 406 00:19:20.710 --> 00:19:20.890 Yeah. 407 00:19:21.130 --> 00:19:24.650 And this one comes from Paula Pinto, who says, while 408 00:19:24.769 --> 00:19:27.349 interesting as always, the whole Squad enthusiasm was pretty much 409 00:19:27.390 --> 00:19:30.890 focused on single developer workflow. Now not having to depend 410 00:19:30.930 --> 00:19:33.400 on anyone else to do related development tasks. It was 411 00:19:33.539 --> 00:19:37.039 all about their nicknamed agents without any ethical discussion of 412 00:19:37.079 --> 00:19:40.640 what tooling like Squad will have in developer teams. I've 413 00:19:40.660 --> 00:19:43.119 already observed how teams

have been downsized. Thank the SaaS 414 00:19:43.160 --> 00:19:45.789 products for placing the need for backend developers, AI tooling 415 00:19:45.809 --> 00:19:47.670 doing the work of team members that used to do 416 00:19:47.690 --> 00:19:50.890 translations or asset creations for CMS platforms. Now we get 417 00:19:50.930 --> 00:19:54.099 tools like Squad reducing even further the amount of humans 418 00:19:54.140 --> 00:19:56.480 in the loop, getting celebrated for what a single developer 419 00:19:56.500 --> 00:19:58.559 could choose while taking into account the repercussions it has 420 00:19:59.000 --> 00:20:00.839 on those that are deemed no longer needed to be 421 00:20:00.859 --> 00:20:01.740 kept around as employees. 422 00:20:01.819 --> 00:20:03.259 Unless you didn't have a team to begin with. 423 00:20:03.319 --> 00:20:05.720 In a late capitalism-driven society, if $ 5 can do what... 424 00:20:06.690 --> 00:20:08.829 work of 20 developers thanks to tooling like squad the 425 00:20:08.869 --> 00:20:10.609 remaining 15 are out of the company like an old 426 00:20:10.630 --> 00:20:12.910 detroit they will need to find work that isn't yet taken. 427 00:20:12.710 --> 00:20:15.880 By ai unless you didn't start with a team yeah 428 00:20:16.000 --> 00:20:17.299 you were always one guy. 429 00:20:18.039 --> 00:20:20.819 But i also think it's kind of a short-sighted mindset 430 00:20:20.859 --> 00:20:24.299 too because i'm seeing more nobody team's not getting laid 431 00:20:24.339 --> 00:20:26.400 out but just working on the other stuff they hadn't 432 00:20:26.440 --> 00:20:29.160 gotten to like the joke for years was none of 433 00:20:29.220 --> 00:20:30.750 us would get to the bottom of our to-do list 434 00:20:32.079 --> 00:20:35.559 And now we're getting through the feature lists and starting 435 00:20:35.579 --> 00:20:38.619 to get into tech debt. And then other features start 436 00:20:38.700 --> 00:20:43.460 to appear. It's only the shortest-sighted of companies that would 437 00:20:43.500 --> 00:20:46.640 work on one thing and not think, what are the 438 00:20:46.680 --> 00:20:48.309 other things we could do? You've got a group of 439 00:20:48.349 --> 00:20:51.049 people who are now more powerful. Utilize them. 440 00:20:51.069 --> 00:20:52.970 You take all those people on the team, give them 441 00:20:53.049 --> 00:20:57.279 all a unique task, give them their AI tools. And 442 00:20:57.500 --> 00:20:59.789 set them free. And before you know it, you're through 443 00:20:59.880 --> 00:21:00.619 all of those things. 444 00:21:00.869 --> 00:21:01.089 Yeah. 445 00:21:01.200 --> 00:21:02.460 Technical debt's a big one. 446 00:21:02.589 --> 00:21:05.150 I also think there is a larger conversation about what 447 00:21:05.210 --> 00:21:09.069 the shape of teams looks like with these tools going forward. Right. 448 00:21:09.289 --> 00:21:10.430 We're still in early days. 449 00:21:10.789 --> 00:21:14.170 I've been running into folks who, in the startup community, 450 00:21:14.230 --> 00:21:16.390 are saying they don't need developers anymore. 451 00:21:16.970 --> 00:21:19.190 Yeah. That doesn't seem true. 452 00:21:20.710 --> 00:21:23.650 Somebody I ran into, he just came back from two 453 00:21:23.670 --> 00:21:27.670 weeks of a startup incubator, and he was saying, Of 454 00:21:27.750 --> 00:21:31.490 the 20 or so startups there, none of them had developers. 455 00:21:31.569 --> 00:21:31.809 Yeah. 456 00:21:31.930 --> 00:21:32.150 Right. 457 00:21:32.309 --> 00:21:33.289 That's easy to understand. 458 00:21:33.869 --> 00:21:35.950 Sure. All right, Paulo. Thank you so much for your comment. 459 00:21:36.049 --> 00:21:37.410 And a copy of Music Code Buy is on its 460 00:21:37.430 --> 00:21:38.569 way to you. And if you'd like a copy of 461 00:21:38.589 --> 00:21:42.369 Music Code Buy, write a comment on the website at. netronics. com 462 00:21:41.690 --> 00:21:43.500 or on the Facebooks. We publish every show there. And 463 00:21:43.539 --> 00:21:45.000 if you comment there and I read it on the show, 464 00:21:45.460 --> 00:21:46.690 we'll send you a copy of Music Code Buy. 465 00:21:46.700 --> 00:21:52.599 Or you could just go to musictocodebuy.net and get it yourself. MP3, WAV, 466 00:21:52.700 --> 00:21:56.240 and FLAC formats. Go ahead. You know you want it. 467 00:21:59.170 --> 00:22:01.809 All right, let's bring on Don Demsack. Don is a 468 00:22:01.849 --> 00:22:05.809 technology leader dedicated to strengthening the tech ecosystem in Central 469 00:22:05.849 --> 00:22:09.430 Pennsylvania and beyond. His day job is with a major 470 00:22:09.470 --> 00:22:14.279 technology company's infrastructure group where he helps CIOs and CTOs 471 00:22:14.380 --> 00:22:19.140 think like COOs. His nights and weekends go to volunteering 472 00:22:19.200 --> 00:22:21.700 as a board member of the Technology Council of Central 473 00:22:21.720 --> 00:22:26.569 PA and as an active member of Tech Lancaster. You 474 00:22:26.589 --> 00:22:29.970 can find his latest ramblings over on Substack. This is 475 00:22:30.039 --> 00:22:32.420 his fourth trip to. NET Rocks and the first in 476 00:22:32.460 --> 00:22:33.400 almost 20 years. 477 00:22:33.940 --> 00:22:34.740 It's been too long. 478 00:22:35.019 --> 00:22:40.140 Dude, I remember your first episode. It was during the 479 00:22:40.200 --> 00:22:44.140 Rory years. So it's episode 62. Rory years were 50 480 00:22:44.140 --> 00:22:49.970 to 100. And he somehow convinced you to come up 481 00:22:50.009 --> 00:22:52.130 to the studio. And you were there in the studio 482 00:22:52.170 --> 00:22:53.250 for us, with us. 483 00:22:53.769 --> 00:22:54.589 Yep. And you remember. 484 00:22:55.099 --> 00:22:57.539 Richard Roy used to do the Google weirdos thing. 485 00:22:57.720 --> 00:22:58.000 Right. 486 00:22:59.799 --> 00:23:03.039 So just for people who don't remember, he would look, 487 00:23:03.059 --> 00:23:08.460 he had a blog, popular blog,

neopolian.com. And it was 488 00:23:08.779 --> 00:23:14.640 hysterical and very, very surreal. And he took it down 489 00:23:14.680 --> 00:23:16.809 because he, he wanted to be taken more seriously as 490 00:23:16.829 --> 00:23:19.230 a developer and everybody just wanted comic relief, but he 491 00:23:19.269 --> 00:23:23.670 was very funny. Anyway, he would look at the logs and, 492 00:23:24.390 --> 00:23:28.890 of what people typed into Google and then landed themselves 493 00:23:29.109 --> 00:23:31.609 on his blog, right? He could tell that it came 494 00:23:31.650 --> 00:23:36.579 from Google and the string is in the log, right? 495 00:23:36.599 --> 00:23:37.150 The URL. 496 00:23:38.130 --> 00:23:42.099 And so he would read these and they were just crazy, 497 00:23:42.119 --> 00:23:45.500 what people were looking for. But one of them, and 498 00:23:45.519 --> 00:23:49.740 I remember this specifically, Don, tell Don XML to come 499 00:23:49.799 --> 00:23:51.039 home and take out the garbage. 500 00:23:52.809 --> 00:23:56.930 You just made my wife's week. That is exactly. Make 501 00:23:57.049 --> 00:23:58.470 sure Carl brings that up. 502 00:23:59.190 --> 00:24:00.759 Oh, well, how could we not? 503 00:24:00.779 --> 00:24:03.000 I mean, it's crazy. 504 00:24:03.259 --> 00:24:06.259 But yeah, you used to be Don XML. And then 505 00:24:06.299 --> 00:24:09.759 I think by your third episode, which was 271, it 506 00:24:09.799 --> 00:24:11.740 was Don Demsack on Link to XML. 507 00:24:12.109 --> 00:24:16.599 Yeah. And who did the session before Mike? Do you remember? 508 00:24:16.619 --> 00:24:18.880 Uh, two 70 to 70. 509 00:24:18.880 --> 00:24:20.619 Let's see. I got the list right here. 510 00:24:22.099 --> 00:24:23.779 It's funny. It's all online. That's weird. 511 00:24:23.839 --> 00:24:25.740 Yes, it is. I can tell you. 512 00:24:25.759 --> 00:24:28.779 No, I got it right here. It was Eric Meyer, 513 00:24:29.220 --> 00:24:29.980 Eric Meyer. 514 00:24:30.400 --> 00:24:34.359 Oh, why the Eric Meyer on link on link? Yeah. 515 00:24:34.380 --> 00:24:35.539 The guy who helped create it. 516 00:24:35.599 --> 00:24:37.200 Something he knows a little thing about. 517 00:24:37.509 --> 00:24:37.740 Yeah. 518 00:24:37.809 --> 00:24:42.029 Just a little. Um, if you haven't, he, or, uh, 519 00:24:42.069 --> 00:24:46.339 been paying attention to him, uh, He recently has a 520 00:24:46.380 --> 00:24:48.980 video up on YouTube of a presentation. 521 00:24:49.000 --> 00:24:49.079 AI. 522 00:24:50.640 --> 00:24:58.319 Engineering, around how AI agents got teeth once we allowed 523 00:24:58.440 --> 00:25:03.279 them to do more than just chat when he gave 524 00:25:03.299 --> 00:25:06.579 them access to tools. And he throws out a whole 525 00:25:06.759 --> 00:25:11.529 theory of how we can lock down AI agents by 526 00:25:11.589 --> 00:25:16.039 making them do... mathematical proofs that the stuff that they're 527 00:25:16.059 --> 00:25:22.609 doing won't negatively impact something. It is an awesome video. 528 00:25:22.950 --> 00:25:25.269 Yeah, we need more of that. Absolutely. 529 00:25:25.269 --> 00:25:26.549 100%. 530 00:25:26.549 --> 00:25:30.990 I saw an article that you wrote on LinkedIn, which 531 00:25:31.029 --> 00:25:35.369 was also on your sub stack. Everything that I've seen 532 00:25:35.589 --> 00:25:39.259 from you has been thought-provoking, but this was about the 533 00:25:39.279 --> 00:25:48.039 new MCP 2.0 and how you can't make a stateful 534 00:25:48.279 --> 00:25:53.299 MCP server anymore if you're using 2.0. And so, yeah, 535 00:25:53.339 --> 00:25:55.589 we talked about that. I talked about that with Rocky 536 00:25:55.890 --> 00:26:02.730 on Code with AI. And Richard found your substack and 537 00:26:02.809 --> 00:26:05.529 found some good stuff that you're thinking of and talking about. 538 00:26:05.849 --> 00:26:06.650 And here you are. 539 00:26:06.890 --> 00:26:11.319 Yeah, it ties right back to what Eric is talking about. 540 00:26:11.380 --> 00:26:13.240 It was funny because I've been working on this for 541 00:26:13.279 --> 00:26:20.819 a bit. But this whole idea of how do we 542 00:26:20.980 --> 00:26:22.940 govern these agents better? 543 00:26:23.099 --> 00:26:23.559 Right. Sure. 544 00:26:23.700 --> 00:26:28.259 Everybody. Now, for the last two years, people have been 545 00:26:28.480 --> 00:26:32.279 really complaining about governance. But the big thing everybody now 546 00:26:32.380 --> 00:26:34.240 is complaining about is tokenomics. 547 00:26:34.500 --> 00:26:35.920 Tokenomics. 548 00:26:36.079 --> 00:26:39.420 That's the big thing that everybody is looking because now 549 00:26:39.440 --> 00:26:41.119 there's a cost per token. 550 00:26:41.279 --> 00:26:41.519 Right. 551 00:26:41.599 --> 00:26:46.710 Well, those two things are kind of related. about two 552 00:26:46.730 --> 00:26:49.809 years ago, just over two years ago, through my day job, 553 00:26:50.049 --> 00:26:54.750 I got to see a demo of some agentic AI. 554 00:26:55.509 --> 00:27:01.509 And the demo that they created was really thought-provoking because 555 00:27:01.529 --> 00:27:09.549 it didn't just show what the agents were working on, 556 00:27:09.950 --> 00:27:12.859 but it actually, at the bottom of it, showed the 557 00:27:12.990 --> 00:27:16.599 conversations the agents were actually having with each other. And 558 00:27:17.099 --> 00:27:20.960 I was like, wow, that looks like a Teams or 559 00:27:21.000 --> 00:27:26.839 a Slack conversation. But I'd never thought of agentic AI 560 00:27:27.180 --> 00:27:32.160 in that way. We're software developers. We think about APIs, 561 00:27:32.519 --> 00:27:36.539 application programming interfaces. Now there's this whole idea of a 562 00:27:37.220 --> 00:27:38.700 natural language interface. 563 00:27:38.900 --> 00:27:42.690 Or semi-natural, like The way those tools talk to each 564 00:27:42.789 --> 00:27:47.329 other is not that normal, but I

get what you're saying. 565 00:27:47.430 --> 00:27:49.109 It is English of a sort. 566 00:27:49.750 --> 00:27:52.009 And that, that was the thing that jumped out at 567 00:27:52.029 --> 00:27:59.759 me was it was too generic in a team's chat. 568 00:27:59.920 --> 00:28:05.460 You expect a lot of three letter acronyms and very 569 00:28:05.519 --> 00:28:08.849 specific words. for each of the roles that are there. 570 00:28:09.369 --> 00:28:13.089 Think about chat between a DBA, a front end developer 571 00:28:13.130 --> 00:28:16.970 and a middle and a mid tier developer, all the conversations, 572 00:28:17.119 --> 00:28:19.220 all the acronyms that they wind up using. 573 00:28:19.319 --> 00:28:19.539 Yeah. 574 00:28:19.740 --> 00:28:20.920 That was lacking. 575 00:28:21.259 --> 00:28:24.039 It speaks to a lack of common context between the agents. 576 00:28:24.440 --> 00:28:28.160 And that's where my head was going was how do I, that, 577 00:28:28.220 --> 00:28:32.559 that seems to be missing. Um, and as you guys know, 578 00:28:32.619 --> 00:28:35.259 and you've got, you guys had a bunch of over 579 00:28:35.299 --> 00:28:39.950 the years, especially recently, domain-driven design type folks. I was like, 580 00:28:40.410 --> 00:28:42.009 there's no ubiquitous language. 581 00:28:42.549 --> 00:28:42.750 Right. 582 00:28:43.769 --> 00:28:52.130 What if we made these agents speak the ubiquitous language 583 00:28:52.599 --> 00:28:56.500 of the domain that they're in charge of? 584 00:28:56.819 --> 00:28:57.039 Right. 585 00:28:57.319 --> 00:28:59.240 That would be super cool. 586 00:28:59.599 --> 00:29:02.259 And really what you're doing is bracketing their context, right? 587 00:29:02.279 --> 00:29:04.400 That they need to use the correct language. 588 00:29:04.640 --> 00:29:09.339 Correct. And by doing that, there's side effects, lots of 589 00:29:09.380 --> 00:29:13.140 good side effects. If you can constrain them or control 590 00:29:13.200 --> 00:29:17.029 their language, I'm calling it a controlled natural language. Um, 591 00:29:17.390 --> 00:29:20.750 then they can only do what they can express. 592 00:29:21.450 --> 00:29:21.789 Yeah. 593 00:29:22.089 --> 00:29:23.430 And it's another kind of constraint. 594 00:29:23.910 --> 00:29:28.029 Think about the open AI hacking of hugging face. 595 00:29:28.589 --> 00:29:28.829 Right. 596 00:29:28.970 --> 00:29:34.759 Where it had access via MCP to a tool. figured 597 00:29:34.859 --> 00:29:38.990 out how to do something that wasn't desired by that tool. 598 00:29:39.450 --> 00:29:42.569 But if we had a natural, a constrained or controlled 599 00:29:42.609 --> 00:29:45.190 natural language in front of that, then they wouldn't be 600 00:29:45.250 --> 00:29:47.609 able to ask the things that they shouldn't be able 601 00:29:47.630 --> 00:29:48.009 to do. 602 00:29:48.210 --> 00:29:49.190 Yeah. 603 00:29:49.650 --> 00:29:55.769 So that's, and so the term, these are patterns that I, 604 00:29:56.380 -->

00:30:01.279 I'm using for this is instead of bounded context, domain-driven design, 605 00:30:01.740 --> 00:30:05.710 how about bounded agency? So how do we build walls 606 00:30:05.730 --> 00:30:09.910 around what the domain that that particular agent can do, 607 00:30:11.859 --> 00:30:15.279 not via just security, but actually ways to express it? 608 00:30:15.599 --> 00:30:17.220 If they don't have the language to express it, they 609 00:30:17.259 --> 00:30:17.640 can't do it. 610 00:30:17.759 --> 00:30:18.000 Right. 611 00:30:18.160 --> 00:30:21.859 So if you're adding another layer... Doesn't, uh, doesn't that 612 00:30:21.900 --> 00:30:22.960 affect your tokenomics? 613 00:30:23.259 --> 00:30:30.720 That's the other side is now you can actually decrease the, um, the, 614 00:30:31.220 --> 00:30:37.109 the verbiage that's going across. Um, and one of the 615 00:30:37.349 --> 00:30:41.109 major challenges with a natural language as us as humans 616 00:30:41.890 --> 00:30:43.849 is there's a lots of different ways to say the 617 00:30:43.910 --> 00:30:48.289 exact same thing, which leads to ambiguity. Right. And there's, and, 618 00:30:49.519 --> 00:30:53.299 By constraining it or controlling that language, you actually reduce 619 00:30:53.480 --> 00:31:00.019 the actual word count, which you're going to reduce the tokens. 620 00:31:00.240 --> 00:31:03.740 But then you're also decreasing your ambiguity, too, because you 621 00:31:03.759 --> 00:31:05.359 can only say it a certain way. 622 00:31:05.519 --> 00:31:09.900 Well, aren't you using tokens for that intermediate layer that 623 00:31:09.940 --> 00:31:11.250 does the translation? 624 00:31:11.710 --> 00:31:16.130 So originally, that's where I was thinking. But then I 625 00:31:16.150 --> 00:31:17.470 take it one step further. 626 00:31:17.490 --> 00:31:17.779 If I. 627 00:31:18.880 --> 00:31:24.400 can control the language, and so I can control the syntax. 628 00:31:24.440 --> 00:31:27.670 And this goes back to stuff that you guys know 629 00:31:27.690 --> 00:31:30.690 me from, XML. What did I do with XML? I 630 00:31:30.710 --> 00:31:34.329 created domain-specific languages. What did you do with JSON? Create 631 00:31:34.369 --> 00:31:38.279 domain-specific languages. What do we do with Fluent APIs and LINQ? 632 00:31:38.799 --> 00:31:43.720 We created the Fluent APIs and other languages within code. Well, 633 00:31:44.279 --> 00:31:50.519 with all the stuff that's been going on with chat 634 00:31:50.539 --> 00:31:55.940 GPT and natural language processing and stuff that we were 635 00:31:55.980 --> 00:32:00.160 doing with, with link back in the day. What if 636 00:32:00.299 --> 00:32:05.369 I could put a intent compiler in front of the 637 00:32:05.410 --> 00:32:12.069 tool that then parses that text because it's a constrained 638 00:32:12.109 --> 00:32:16.880 natural language. And instead of feeding everything to the model, 639 00:32:17.980 --> 00:32:23.450 what if I if I knew what words mapped to 640 00:32:23.509 --> 00:32:28.450 specific functions, I could actually call those functions instead of 641 00:32:28.490 --> 00:32:31.890 having to pass it through to the agent who would 642 00:32:31.990 --> 00:32:36.250 eventually pass it through to an MCP server somewhere to 643 00:32:36.569 --> 00:32:40.799 actually call the tool. So now I've actually reduced, dramatically 644 00:32:40.859 --> 00:32:44.000 reduced how much is actually going to the models because 645 00:32:44.019 --> 00:32:45.799 I'm pulling some pieces of that. 646 00:32:46.279 --> 00:32:49.599 From this static glossary kind of thing. But, And I 647 00:32:49.660 --> 00:32:52.000 imagine that the best of both worlds would be if 648 00:32:52.119 --> 00:32:54.859 the word doesn't exist in the glossary and it can't 649 00:32:54.920 --> 00:32:57.619 express it in a more concise way, then it would 650 00:32:57.680 --> 00:33:00.740 go out and get definitions for those words and add 651 00:33:00.759 --> 00:33:01.559 them to the domain. 652 00:33:02.309 --> 00:33:05.950 So you, you, you really want to control that language 653 00:33:06.049 --> 00:33:08.190 so that you're controlling what it can do. 654 00:33:08.210 --> 00:33:09.109 Right. 655 00:33:09.130 --> 00:33:15.720 So you would prefer, uh, to keep to, uh, to 656 00:33:15.940 --> 00:33:19.700 programmatically add the new languages. Don't let it do it 657 00:33:19.880 --> 00:33:20.529 on its own. 658 00:33:20.549 --> 00:33:22.410 Well, at least it could make you a list. 659 00:33:22.650 --> 00:33:23.309 A list of it. 660 00:33:24.529 --> 00:33:26.869 Yeah, a list of things that it didn't understand or whatever. 661 00:33:26.990 --> 00:33:31.410 Correct. And you can publish that language. And that's one 662 00:33:31.450 --> 00:33:35.920 of the things that models are great at, converting from 663 00:33:36.059 --> 00:33:37.400 one language to another language. 664 00:33:37.420 --> 00:33:37.640 Right. 665 00:33:38.079 --> 00:33:41.460 So if you have a way to declare that controlled 666 00:33:41.519 --> 00:33:44.059 natural language, it can convert it automatically. 667 00:33:44.519 --> 00:33:47.680 Yeah, that's right. I could see repos for all sorts 668 00:33:47.720 --> 00:33:53.440 of vertical industries, uh, DSLs in used in that way to, 669 00:33:53.640 --> 00:33:57.880 to rein in your LLMs and your tokenomics. 670 00:33:58.700 --> 00:34:01.299 I make it a heck of a lot easier for that. 671 00:34:01.420 --> 00:34:06.799 And so, um, so I actually built, uh, uh, uh, 672 00:34:07.240 --> 00:34:08.900 I built a couple of different things based on that, 673 00:34:08.920 --> 00:34:14.119 but I actually have a, uh, MVP that you can 674 00:34:14.159 --> 00:34:17.570 go pull down from GitHub. It's called Limelight X. Well, 675 00:34:17.730 --> 00:34:19.250 Rush is out this year, right? 676 00:34:19.469 --> 00:34:20.250 Doing the tour. 677 00:34:20.349 --> 00:34:24.409 So Rush fan had to pull that in. So Limelight X. 678

00:34:26.690 --> 00:34:29.840 And so you go there, you can actually go see 679 00:34:29.989 --> 00:34:37.659 how to do this. So you can give it the 680 00:34:37.739 --> 00:34:43.980 natural language, the constrained natural language. give it a text 681 00:34:44.860 --> 00:34:50.469 and it will actually give you the intermediate steps so 682 00:34:50.489 --> 00:34:53.230 you can visualize it. So how do you parse that 683 00:34:53.269 --> 00:34:57.409 apart to an AST? How do you take that AST 684 00:34:58.050 --> 00:35:01.070 and optimize it? Then you take that AST and turn 685 00:35:01.110 --> 00:35:04.650 it into intermediate result, which then gets put into an 686 00:35:04.710 --> 00:35:08.710 expression tree a la link. How do you do that? 687 00:35:08.909 --> 00:35:16.769 Because what I've found is, Most people doing AI engineering, 688 00:35:17.369 --> 00:35:24.210 AI agents, they don't come from our comp sci background. Right. 689 00:35:24.570 --> 00:35:26.989 They come from the dynamic language. 690 00:35:26.809 --> 00:35:27.550 Side of the house. 691 00:35:27.690 --> 00:35:34.500 They're Python and JavaScript type script. So parsing and compiler 692 00:35:34.719 --> 00:35:38.139 theory type stuff is not something that sits in their domain. 693 00:35:38.969 --> 00:35:42.380 While you can do some of that stuff with those languages, 694 00:35:42.420 --> 00:35:45.420 you don't typically do it. So they're not thinking that way. 695 00:35:46.960 --> 00:35:49.619 So how do I get best of both worlds? That's 696 00:35:49.699 --> 00:35:50.599 part of the challenge. 697 00:35:51.059 --> 00:35:52.739 Before we go any further, I think we should take 698 00:35:52.760 --> 00:35:54.880 a little break. You think so, Richard? 699 00:35:55.260 --> 00:35:55.760 Sure, let's do it. 700 00:35:55.760 --> 00:36:00.719 All right, we'll be right back after these very important messages. Hey, 701 00:36:00.780 --> 00:36:04.340 Carl here. You probably know text control is a powerful 702 00:36:04.400 --> 00:36:08.519 library for document editing and PDF generation. But did you 703 00:36:08.559 --> 00:36:11.539 know they're also a strong supporter of the developer community? 704 00:36:12.099 --> 00:36:14.460 It's part of their mission to build and support a 705 00:36:14.519 --> 00:36:18.949 strong developer community by being present, listening to users, and 706 00:36:19.050 --> 00:36:22.670 sharing knowledge at conferences across Europe and the United States. 707 00:36:23.170 --> 00:36:26.269 So if you're heading to a conference soon, check if 708 00:36:26.309 --> 00:36:29.340 Text Control will be there and stop by to say hi. 709 00:36:29.699 --> 00:36:35.920 You can find their full conference calendar at www.textcontrol.com. And 710 00:36:35.960 --> 00:36:38.489 make sure you thank them for supporting. NET Rocks. 711 00:36:42.570 --> 00:36:43.429 And we're back. It's. 712 00:36:43.389 --> 00:36:43.750 NET Rocks. 713 00:36:43.789 --> 00:36:45.210 I'm Richard Campbell. That's Carl Franklin. 714 00:36:45.389 --> 00:36:45.730 Hey. 715 00:36:45.949 --> 00:36:48.590 Talking to our friend Don Demsack a bit about, well, 716 00:36:48.760 --> 00:36:51.460 you just brought up Limelight X as a way to 717 00:36:51.480 --> 00:36:56.920 sort of create a natural language-like set of instructions, which, 718 00:36:56.940 --> 00:36:59.000 I mean, in some ways reminds me of the way 719 00:36:59.039 --> 00:37:03.670 we write specs anyway. Software development is somewhat language. I mean, 720 00:37:03.690 --> 00:37:05.880 you have the programming language is a constraint. But if 721 00:37:05.920 --> 00:37:09.960 you also notice, the average product manager speaks in a 722 00:37:10.099 --> 00:37:14.070 somewhat constrained language. They're trying to communicate with developers, and 723 00:37:14.110 --> 00:37:16.570 developers speak a weird form of English. 724 00:37:18.090 --> 00:37:19.130 Dilbert proved that. 725 00:37:20.289 --> 00:37:23.989 They're really trying to just be very clear on what 726 00:37:24.070 --> 00:37:27.239 the intent is, and I think it serves the LLMs 727 00:37:27.320 --> 00:37:29.900 pretty well to stick with that same kind of language barrier. 728 00:37:29.900 --> 00:37:30.960 100%. 729 00:37:30.960 --> 00:37:35.320 And then domain-driven design never took off. 730 00:37:35.809 --> 00:37:36.030 Yeah. 731 00:37:36.289 --> 00:37:39.110 We all know it works, but one of the side 732 00:37:39.170 --> 00:37:42.949 effects of implementing it is you needed to have a 733 00:37:43.030 --> 00:37:47.780 subject matter expert who can pull that the main logic 734 00:37:48.070 --> 00:37:52.420 out of that, those communications that was very specialized. 735 00:37:52.940 --> 00:37:53.179 Yeah. 736 00:37:53.619 --> 00:38:02.039 Well, nowadays with coding assistance and the spec driven development world, 737 00:38:03.619 --> 00:38:07.760 we are starting to put that context into text documents 738 00:38:07.840 --> 00:38:11.639 that obviously we could use models to start pulling out 739 00:38:11.900 --> 00:38:15.320 that domain-specific language from the specifications. 740 00:38:15.679 --> 00:38:17.909 I would even go one step further and say we're 741 00:38:17.949 --> 00:38:20.550 heading towards, if we're not there already, just not thinking 742 00:38:20.590 --> 00:38:25.369 of it, conversation-based development. In this last episode that I 743 00:38:25.389 --> 00:38:28.710 did with Rocky Latka, he talked about how he and 744 00:38:28.829 --> 00:38:32.250 three other guys sat down to design a system He 745 00:38:32.289 --> 00:38:36.250 whipped out his Android phone, started recording. It made a transcript. 746 00:38:36.690 --> 00:38:39.769 He uploaded that transcript to Claude and said, build me 747 00:38:39.809 --> 00:38:43.920 a spec based on this conversation. Whoa. I mean, talk 748 00:38:43.960 --> 00:38:46.739 about layers, right? 749 00:38:46.860 --> 00:38:48.760 Using the tool as the preprocessor. 750 00:38:49.000

--> 00:38:49.400 Yeah. 751 00:38:50.820 --> 00:38:56.940 It's an interesting space that we're in because that's been 752 00:38:57.000 --> 00:39:00.909 the art of being a developer is has been really 753 00:39:00.989 --> 00:39:06.880 taking these vague requirements and then converting it into code. 754 00:39:07.340 --> 00:39:10.820 What was my favorite, what's the number one thing whenever 755 00:39:10.920 --> 00:39:14.940 I would build something, the business person would typically turn 756 00:39:14.960 --> 00:39:16.760 around and say, I know that's what I asked for, 757 00:39:16.800 --> 00:39:17.920 but that's not what I wanted. 758 00:39:18.340 --> 00:39:18.840 Yeah, right. 759 00:39:20.110 --> 00:39:24.289 Because humans are really, really, really bad at giving requirements. Weird. 760 00:39:24.329 --> 00:39:26.789 Yeah. And in some ways, like you want the domain 761 00:39:26.849 --> 00:39:29.949 expert because they know the space, but they often skip over. 762 00:39:30.179 --> 00:39:32.699 essential parts because they're so knowledgeable in the space. They 763 00:39:32.719 --> 00:39:35.059 don't even realize what they know. 764 00:39:35.280 --> 00:39:35.519 Yeah. 765 00:39:35.659 --> 00:39:35.820 Yeah. 766 00:39:35.980 --> 00:39:39.380 And that's, I mean, because Rocky's been around, he knows 767 00:39:40.000 --> 00:39:43.440 what he needs to put into that. And so how 768 00:39:43.460 --> 00:39:48.570 do you teach that to these new wave of, uh, 769 00:39:50.329 --> 00:39:55.829 coding engineers that are using coding assistants to do that. Well, 770 00:39:57.170 --> 00:40:01.949 we could teach them better requirement skills. And that's one 771 00:40:01.989 --> 00:40:03.989 of the things I started doing a couple of years 772 00:40:04.050 --> 00:40:09.909 ago in prompt engineering was a big rage two years ago. 773 00:40:09.929 --> 00:40:13.289 And now it's context engineering. But when I explain it 774 00:40:14.190 --> 00:40:16.610 to the average person, what a prompt is, I tell 775 00:40:16.630 --> 00:40:21.110 them it's a requirement. Think of it as giving it 776 00:40:21.150 --> 00:40:22.570 a requirement to another human. 777 00:40:22.710 --> 00:40:22.909 Right. 778 00:40:22.969 --> 00:40:26.050 Got to put in all that extra context. And then 779 00:40:26.150 --> 00:40:30.409 because of my agile background, I like to use given 780 00:40:30.449 --> 00:40:34.880 when then, so that as measured by. And then I 781 00:40:34.920 --> 00:40:38.739 would teach people to write their prompts in that structured 782 00:40:38.820 --> 00:40:42.559 form and they get better results out of the engine 783 00:40:42.719 --> 00:40:45.579 because you're giving it all the context that you would need. 784 00:40:45.840 --> 00:40:47.420 Yeah. That's brilliant. 785 00:40:47.559 --> 00:40:50.949 And so that's, And so that's the same problem that 786 00:40:51.250 --> 00:40:58.000 I'm running into with the spec-driven development is while they 787 00:40:58.039 --> 00:41:03.079 have great structure for the documents, how do you build 788 00:41:03.119 --> 00:41:05.559 the tests? Right. What you really kind of need to 789 00:41:05.619 --> 00:41:10.510 be able to do is create, given when then type scenarios, 790 00:41:11.030 --> 00:41:17.409 do more behavior-driven development. So testing for behavior that's expected 791 00:41:17.510 --> 00:41:20.889 or not expected. rather than writing generic tests. 792 00:41:20.949 --> 00:41:24.329 You're right. Domain-driven development is great at defining what we'd 793 00:41:24.449 --> 00:41:28.539 like it to do, but behavior-driven development is much better 794 00:41:28.659 --> 00:41:31.219 at defining what it actually does and if that's what 795 00:41:31.260 --> 00:41:32.059 we really wanted to do. 796 00:41:32.199 --> 00:41:37.199 And then if you could take that, those behavior-driven tests, 797 00:41:37.880 --> 00:41:43.250 now you can also use that to drive the controlled 798 00:41:43.289 --> 00:41:46.769 natural language because you're expressing the behaviors that you're looking for. 799 00:41:47.329 --> 00:41:47.650 And so. 800 00:41:48.800 --> 00:41:54.360 Besides Limelight X, I have a project called VectorRen, V-E-C-T-O-R, 801 00:41:55.400 --> 00:42:01.460 my SVG days, Ren, R-E-N, so two R's. And it 802 00:42:01.590 --> 00:42:06.929 was built with that in mind to show developers how 803 00:42:06.989 --> 00:42:12.250 to build an SVG game, a web-based with JavaScript, with 804 00:42:12.449 --> 00:42:16.469 generic JavaScript, no framework. And so what was the, you 805 00:42:16.489 --> 00:42:19.440 guys remember, what the big game was when we were 806 00:42:19.460 --> 00:42:22.099 trying to do demo SVG back in the day, it 807 00:42:22.119 --> 00:42:23.599 was asteroids. Right. 808 00:42:24.119 --> 00:42:25.590 Oh yeah. So that's right. 809 00:42:25.829 --> 00:42:31.050 There is in half an hour using this methodology, I 810 00:42:31.090 --> 00:42:34.130 describe what the game should do and given one then, 811 00:42:34.150 --> 00:42:40.000 and it generated a simple implementation of asteroids. And you 812 00:42:40.039 --> 00:42:44.239 can go to vector ren.org or go to vector ren, uh, 813 00:42:44.500 --> 00:42:46.989 the GitHub page, uh, out there and you can pull 814 00:42:47.030 --> 00:42:49.949 it down. And it's a very simple way to learn 815 00:42:49.989 --> 00:42:53.909 JavaScript from this point of view, without any of the frameworks, 816 00:42:55.030 --> 00:42:58.800 something simple that, that, that people can actually leverage, uh, 817 00:42:58.940 --> 00:43:03.880 because we have this generation of developers who don't have 818 00:43:03.920 --> 00:43:05.599 the traditional comp side background. 819 00:43:05.619 --> 00:43:05.780 Right. 820 00:43:05.960 --> 00:43:06.179 Yeah. 821 00:43:06.460 --> 00:43:10.159 I'm just thinking about, we're revising this. We've been playing 822 00:43:10.179 --> 00:43:12.510 with what the software development life cycle with

these tools 823 00:43:12.570 --> 00:43:16.550 is supposed to look like. And, and I'll, I think 824 00:43:16.570 --> 00:43:19.269 there's a lot more energy put into the context up 825 00:43:19.349 --> 00:43:22.929 front just because the tools will generate lots and lots 826 00:43:22.969 --> 00:43:24.809 of code. It's just most of it is not that 827 00:43:24.849 --> 00:43:27.380 good if you haven't really done a good job of 828 00:43:27.510 --> 00:43:28.820 setting up these constraints. 829 00:43:30.019 --> 00:43:33.500 I've been doing a lot of stuff with AI here 830 00:43:33.519 --> 00:43:37.280 in Central Pennsylvania. One of the interesting feedback I'm getting 831 00:43:37.340 --> 00:43:41.539 from non-traditional developers is we don't care. 832 00:43:42.539 --> 00:43:43.619 Yeah. 833 00:43:44.139 --> 00:43:45.920 It gets the job done, and that's all. 834 00:43:45.840 --> 00:43:46.219 I care about. 835 00:43:46.239 --> 00:43:47.699 You're going to care. 836 00:43:50.960 --> 00:43:54.239 I hit up one of the folks I was talking about. 837 00:43:54.280 --> 00:43:59.690 I was talking about, hey, you're just creating dynamic scripts. 838 00:43:59.710 --> 00:44:06.670 This feels a lot like what JavaScript felt like before jQuery, 839 00:44:06.730 --> 00:44:10.949 before we got V8, before we got TypeScript. We're sending around. 840 00:44:10.969 --> 00:44:13.900 Yeah, before we put constraints around for reliability. 841 00:44:14.400 --> 00:44:20.179 And consistency. So, and I was like, you know, from 842 00:44:20.219 --> 00:44:25.570 a tokenomics point of view, every time you send that text, 843 00:44:26.250 --> 00:44:28.550 you're getting charged for it. Yeah. And so there's a 844 00:44:28.610 --> 00:44:31.860 cost because I don't care. It's cheaper for me to 845 00:44:31.929 --> 00:44:34.409 do that than to spend the time to actually write 846 00:44:35.429 --> 00:44:36.250 the compiled code. 847 00:44:36.269 --> 00:44:39.550 For now, anyway, the costs are going up. But often 848 00:44:39.630 --> 00:44:42.050 the folks don't care because they don't know the consequences. 849 00:44:42.329 --> 00:44:44.030 They don't know good engineering principles. 850 00:44:44.469 --> 00:44:47.090 Yeah, they don't know how the software will bite them later. 851 00:44:47.369 --> 00:44:48.670 It's generally what works right now. 852 00:44:48.769 --> 00:44:50.809 And what I'm seeing is a lot of the vibe 853 00:44:50.869 --> 00:44:54.070 coders realizing that they can only go so far of 854 00:44:54.130 --> 00:44:56.730 vibe coding, and then they start going to coding assistants, 855 00:44:57.670 --> 00:45:00.570 but then they don't know how to create these good 856 00:45:00.639 --> 00:45:03.719 engineering principles. And so I think there's a big market 857 00:45:03.860 --> 00:45:04.460 for that. 858 00:45:04.860 --> 00:45:07.219 So the AKA they're starting to care. 859 00:45:07.760 --> 00:45:11.670 And it makes me sad because as a VB guy, 860 00:45:12.190 --> 00:45:14.670 old school VB guy, isn't that what the C plus 861 00:45:14.710 --> 00:45:16.469 plus people complained about us about? 862 00:45:17.150 --> 00:45:17.409 Yep. 863 00:45:17.449 --> 00:45:18.800 Yeah, we weren't real developers. 864 00:45:19.400 --> 00:45:22.480 Well, Yeah, and again, the argument was, is the software 865 00:45:22.530 --> 00:45:24.989 quality sufficient? And let's face it, we built a lot 866 00:45:25.010 --> 00:45:27.469 of bad quality software, and then we learned, hey, software 867 00:45:27.510 --> 00:45:30.429 quality matters. And you could do it with VB, you 868 00:45:30.449 --> 00:45:32.210 just had to follow some good practices. 869 00:45:32.349 --> 00:45:35.969 I think a lot of that nose turning down at 870 00:45:36.409 --> 00:45:40.599 VB developers was a byproduct of what happened between C 871 00:45:41.440 --> 00:45:46.800 developers and basic developers in the very early days of programming. 872 00:45:48.179 --> 00:45:52.500 Basic with go-tos was looked at as horrible. I was 873 00:45:52.599 --> 00:45:57.380 using Visual Basic and then attending a community college and 874 00:45:57.420 --> 00:46:00.920 the computer science lady, I told her that I was 875 00:46:01.239 --> 00:46:03.480 using Visual Basic. She goes, oh, basic. Oh, you can't 876 00:46:03.519 --> 00:46:04.119 say basic. 877 00:46:04.530 --> 00:46:05.849 No go-tos, no go-tos. 878 00:46:05.909 --> 00:46:09.030 I'm like, no, it's completely changed now. You know, it's 879 00:46:09.110 --> 00:46:11.590 come into its own and they still just, you know, 880 00:46:12.210 --> 00:46:14.929 you get an idea in your mind and it's hard 881 00:46:14.969 --> 00:46:15.349 to shake. 882 00:46:15.630 --> 00:46:18.519 And I don't want to be that Get off my 883 00:46:18.579 --> 00:46:20.659 lawn guy. Yeah, right. And I don't want to be 884 00:46:21.300 --> 00:46:21.900 that old guy. 885 00:46:21.920 --> 00:46:22.300 Nobody does. 886 00:46:22.320 --> 00:46:26.920 I've got the gray hair, fine. But I don't want 887 00:46:26.940 --> 00:46:31.920 to be that guy. Same thing with coding assistants. What 888 00:46:31.940 --> 00:46:34.489 I've been talking about to a lot of folks is 889 00:46:34.860 --> 00:46:39.909 coding assistants is the new IDE. I remember back when 890 00:46:40.030 --> 00:46:43.170 we were talking, you know, Vim was it. You didn't 891 00:46:43.329 --> 00:46:44.030 use Studio. 892 00:46:44.489 --> 00:46:44.730 Right. 893 00:46:44.929 --> 00:46:48.489 IntelliSense. Coding is novel. You had to know it and 894 00:46:48.530 --> 00:46:51.530 remember it in your head and use a text editor. 895 00:46:51.989 --> 00:46:53.769 Real developers use text editors. 896 00:46:54.090 --> 00:46:56.010 I think it took Don Box a long time to 897 00:46:56.050 --> 00:46:58.030 come to grips with the fact that people are going 898 00:46:58.050 --> 00:47:00.500 to use IDEs, not them. 899 00:47:01.599 --> 00:47:03.920 It would be interesting to get his opinion on coding. 900

00:47:03.940 --> 00:47:09.820 You see those positions all over the place. I think 901 00:47:09.840 --> 00:47:11.400 you're making a mistake as soon as you're bound to 902 00:47:11.440 --> 00:47:14.159 a tool. In the end, the goal was to deliver 903 00:47:14.199 --> 00:47:15.059 value to customers. 904 00:47:15.869 --> 00:47:19.469 And then there's this whole, and this ties into Run 905 00:47:19.510 --> 00:47:23.769 As Radio, then there's this whole angle from the IT 906 00:47:23.829 --> 00:47:28.909 admin side of the house where IT admins were doing automation, 907 00:47:28.969 --> 00:47:32.510 but they were doing it via a GUI tool because 908 00:47:33.230 --> 00:47:34.909 IT admins didn't write code. 909 00:47:35.409 --> 00:47:35.610 Right. 910 00:47:35.750 --> 00:47:37.369 That was for the developer world. 911 00:47:37.610 --> 00:47:38.190 Right. 912 00:47:38.710 --> 00:47:40.969 In my day job, what I'm starting to see is 913 00:47:41.190 --> 00:47:46.230 more and more IT admins admins leveraging coding assistance to 914 00:47:46.409 --> 00:47:51.170 actually create better code than they could ever written because 915 00:47:51.449 --> 00:47:54.409 all they ever wrote scripts. This is much better code 916 00:47:55.739 --> 00:47:58.400 and it's making their lives a heck of a lot easier. 917 00:47:58.679 --> 00:48:02.519 So that's just, I mean, that's an unintended consequence, but 918 00:48:02.760 --> 00:48:06.260 it's helping, especially in these days when so many companies, 919 00:48:06.380 --> 00:48:10.760 so many on-prem companies who are using VMware and now 920 00:48:11.010 --> 00:48:13.230 they're kind of A little bit of a backlash. They're 921 00:48:13.269 --> 00:48:16.449 all trying to move off, and now they're more open 922 00:48:16.510 --> 00:48:19.269 to how do I do this automation rather than using 923 00:48:19.329 --> 00:48:19.690 a tool. 924 00:48:20.269 --> 00:48:22.909 One quick correction. I think I misspoke. When I was 925 00:48:23.090 --> 00:48:26.670 in a community college, it wasn't VB. It was Quick Basic. 926 00:48:27.150 --> 00:48:31.570 Quick Basic. Quick Basic was compiled, and it was structured, 927 00:48:31.659 --> 00:48:35.480 and it was great back then. And this computer science 928 00:48:35.539 --> 00:48:36.780 person was like, no go-tos. 929 00:48:37.699 --> 00:48:38.019 Don't say it. 930 00:48:39.579 --> 00:48:41.460 Well, and it's the thing, there's nothing wrong with the language. 931 00:48:41.519 --> 00:48:43.320 It just had a construct that was a problem. Hey, 932 00:48:43.360 --> 00:48:46.369 remember option explicit? Like we did learn. 933 00:48:46.989 --> 00:48:47.349 Sure. 934 00:48:48.849 --> 00:48:50.050 On error, resume next. 935 00:48:50.510 --> 00:48:51.809 Yeah. 936 00:48:53.190 --> 00:48:54.190 I don't have any bugs. 937 00:48:55.030 --> 00:48:55.449 What? 938 00:48:55.969 --> 00:48:56.250 Yeah. 939 00:48:56.269 --> 00:48:58.429 So one of the things that I've been starting to 940 00:48:58.550 --> 00:49:02.579 work with is how do I start creating skills to 941 00:49:02.619 --> 00:49:07.420 be able to help teach these non-developers these good habits? 942 00:49:07.460 --> 00:49:10.219 So like one of the things, That's in a beta 943 00:49:10.260 --> 00:49:15.869 right now is a requirements interrogator skill. So help the 944 00:49:15.969 --> 00:49:20.099 average person create that given when then brief so that 945 00:49:20.139 --> 00:49:24.500 they get better requirements when they're trying to build something. 946 00:49:24.760 --> 00:49:25.960 Yeah, you have to wonder if we're not going to 947 00:49:26.039 --> 00:49:28.570 end up with more opinionated agents that are like, I 948 00:49:29.010 --> 00:49:32.389 can't start building this until I have these broader requirements 949 00:49:32.429 --> 00:49:35.389 settled around those kinds of constraints that are going to 950 00:49:35.409 --> 00:49:39.280 matter for security and architecture and reliability and scalability. all 951 00:49:39.300 --> 00:49:42.099 of the abilities that everyone doesn't care about until they 952 00:49:42.139 --> 00:49:42.860 care about them. 953 00:49:43.260 --> 00:49:43.440 Yeah. 954 00:49:44.139 --> 00:49:46.300 One of them I'm playing with right now is, is 955 00:49:46.440 --> 00:49:49.679 creating a skill to create architectural decision records. 956 00:49:50.199 --> 00:49:50.400 Right. 957 00:49:50.699 --> 00:49:54.519 We've never, I mean, we all always wanted to have them, 958 00:49:55.159 --> 00:49:58.739 but now if I'm using a coding assistant, that helps 959 00:49:58.820 --> 00:50:02.019 make my coding system better. It tells the architecture and 960 00:50:02.059 --> 00:50:04.579 the reasons why we picked certain things. 961 00:50:05.909 --> 00:50:06.110 Yeah. 962 00:50:07.409 --> 00:50:07.550 Yeah. 963 00:50:07.949 --> 00:50:10.730 There's lots to all of that. I mean, again, I 964 00:50:10.750 --> 00:50:12.889 feel like we're just groping around right now trying to 965 00:50:12.929 --> 00:50:13.429 figure out. 966 00:50:13.650 --> 00:50:14.909 It's early days. 967 00:50:15.590 --> 00:50:19.010 What are the, yeah, the tools are shifting. Just remember, 968 00:50:19.030 --> 00:50:21.210 we were trying to develop on. NET in the early 969 00:50:21.250 --> 00:50:24.889 days where every time they got a build out, the 970 00:50:25.110 --> 00:50:28.300 editor changed, the framework had changed, and the language had changed. 971 00:50:28.960 --> 00:50:32.500 Everything you're doing, incorrect. Yeah. 972 00:50:32.579 --> 00:50:34.519 I was part of that early days. I don't even 973 00:50:34.579 --> 00:50:36.780 remember if I ever mentioned it to you guys, but 974 00:50:36.940 --> 00:50:44.500 the framework The client that I had back in 99, 2000, 975 00:50:44.500 --> 00:50:51.429 their president was on the board for Microsoft, and I 976 00:50:51.469 --> 00:50:54.670 was building dashboards for him, and I got introduced to 977 00:50:54.710 --> 00:50:55.809 Project Cool.

978 00:50:55.530 --> 00:50:57.969 Before it was announced. Before it became C Sharp? 979 00:50:57.989 --> 00:50:59.269 Yeah, before 2000, the announcement at PDC 2000. 980 00:50:59.250 --> 00:50:59.369 Yeah. 981 00:51:04.820 --> 00:51:06.179 I think it was actually TechEd, but yeah. 982 00:51:06.219 --> 00:51:06.539 TechEd. 983 00:51:06.719 --> 00:51:08.519 Yeah, it was TechEd. That's where I'm at. 984 00:51:08.739 --> 00:51:12.219 But yeah, but I'm also thinking about the craziness that was. 985 00:51:12.420 --> 00:51:18.179 NET Core in, you know, 2018 or 2016. Because it 986 00:51:18.199 --> 00:51:20.010 took them 18 months to get that first version built 987 00:51:20.050 --> 00:51:22.530 and there were so many previews. And every time you 988 00:51:22.550 --> 00:51:25.010 got a preview, everything shuffled on you. And it just 989 00:51:25.030 --> 00:51:28.590 seems an awful lot like trying to work with Anthropica 990 00:51:28.610 --> 00:51:30.730 or OpenAI or any of these tools right now, where 991 00:51:30.829 --> 00:51:33.500 every time you get A new build, new harness, like 992 00:51:33.559 --> 00:51:37.739 there's so many different components that presuming it works the 993 00:51:37.780 --> 00:51:40.690 same way is crazy. It changes. It changes a lot. 994 00:51:40.869 --> 00:51:43.570 And just like with. NET Core, Richard, I think that 995 00:51:43.590 --> 00:51:47.170 there's still a lot of people not dipping their toes 996 00:51:47.250 --> 00:51:51.170 into AI developer-wise for the very same reason. 997 00:51:52.050 --> 00:51:56.110 So have you guys had anybody on for Microsoft Foundry Local? 998 00:51:56.469 --> 00:51:56.840 Not yet. 999 00:51:57.050 --> 00:51:58.519 Not yet. No, it's on my radar. 1000 00:51:59.199 --> 00:52:02.519 I've been intimate with that. So that's part of what 1001 00:52:02.820 --> 00:52:07.639 the bounded context. Part of what I realized is to 1002 00:52:07.960 --> 00:52:12.010 use these smaller models, you had to be very prescriptive 1003 00:52:12.050 --> 00:52:13.329 in the prompts that you were giving. 1004 00:52:13.530 --> 00:52:13.989 Right, right. 1005 00:52:14.030 --> 00:52:16.349 And when I started writing the code, I didn't want 1006 00:52:16.369 --> 00:52:22.070 to put that code as text into my program. I 1007 00:52:22.110 --> 00:52:24.829 wanted to pull it out. And so I created a 1008 00:52:25.880 --> 00:52:30.719 YAML language that did essentially prompt as a function so 1009 00:52:30.760 --> 00:52:33.199 that you could import it in and create it. And 1010 00:52:33.219 --> 00:52:37.619 there's a couple different prompt as a function frameworks out there. 1011 00:52:39.239 --> 00:52:50.130 BAML is one and DPSY, another where it makes it 1012 00:52:50.269 --> 00:52:54.360 easy to put prompts into your programming languages. The Dipsy 1013 00:52:54.400 --> 00:53:01.969 is Python only. BAML is TypeScript and Python. And so 1014 00:53:02.710 --> 00:53:05.289 I built a framework to be able to do that 1015

00:53:05.849 --> 00:53:07.969 because I was trying to run local models. I was 1016 00:53:08.070 --> 00:53:11.780 running local models on what I'm using now, Copilot Plus PC. 1017 00:53:13.699 --> 00:53:17.159 The Copilot Plus PCs were supposed to be the place 1018 00:53:17.199 --> 00:53:19.400 to be able to run these models. It's just not. 1019 00:53:20.010 --> 00:53:21.829 Not as good as it could have been. 1020 00:53:21.849 --> 00:53:26.030 At 40 tops, that's not a lot. Meantime, I'm looking 1021 00:53:26.070 --> 00:53:29.940 at my RTX 5080 at 1,500 tops and saying, why 1022 00:53:30.000 --> 00:53:32.239 aren't I a co-pilot PC? 1023 00:53:32.420 --> 00:53:36.019 So I am super excited for the new RTX Spark laptops. 1024 00:53:36.519 --> 00:53:38.480 I'll be the first to go buy one of them 1025 00:53:38.980 --> 00:53:39.760 when they come out. 1026 00:53:40.360 --> 00:53:42.760 They're actually going to be a laptop, because right now 1027 00:53:42.800 --> 00:53:44.050 it's a NUC. 1028 00:53:44.679 --> 00:53:48.219 Well, the big difference is the DGX Spark, Right. That's 1029 00:53:48.239 --> 00:53:51.820 the one you're thinking is the RTX Spark. Two separate things. 1030 00:53:52.239 --> 00:53:54.840 The DGX is Linux. 1031 00:53:55.219 --> 00:53:55.440 Right. 1032 00:53:55.480 --> 00:53:59.210 The RTX laptop, which is, that was, it was a nook. 1033 00:53:59.969 --> 00:54:03.650 The RTX is a laptop. It is going to run 1034 00:54:03.690 --> 00:54:04.989 Windows on ARM. 1035 00:54:05.329 --> 00:54:05.530 Right. 1036 00:54:05.590 --> 00:54:07.570 So it's, you could think of it as the Uber 1037 00:54:07.989 --> 00:54:09.090 Co-Pilot Plus piece. 1038 00:54:09.429 --> 00:54:12.150 Possibly, yeah. It'll be interesting. It's one of the things 1039 00:54:12.210 --> 00:54:14.230 I've seen over and over again with the DGX Spark, 1040 00:54:14.250 --> 00:54:17.269 but folks are actually using it is, You end up 1041 00:54:17.309 --> 00:54:22.750 running NVIDIA's models because they're so much better optimized for 1042 00:54:22.849 --> 00:54:27.219 their hardware that you just get vastly better responses or results. 1043 00:54:28.099 --> 00:54:32.559 So it's not a universal anything. Where is our ODBC 1044 00:54:34.360 --> 00:54:37.260 so that you really have some equanimity across all of 1045 00:54:37.340 --> 00:54:40.269 these different products? and can really see what all the 1046 00:54:40.389 --> 00:54:42.690 different hardware solutions can be there. I hate that we're 1047 00:54:42.710 --> 00:54:45.519 building dedicated to hardware solutions, but I get it. 1048 00:54:45.699 --> 00:54:49.679 Part of why I created that prompt as a function 1049 00:54:50.059 --> 00:54:53.440 to make it portable, to describe what's the model, what's the, 1050 00:54:54.380 --> 00:54:58.559 am I using VLR? Am I using Foundry Local? What's 1051 00:54:58.599 --> 00:55:01.820 the model? So I can abstract that away so you 1052 00:55:01.840 --> 00:55:03.619 don't have to worry about it and be able to 1053 00:55:03.989 --> 00:55:05.949 switch it in and out. And so I've got a 1054 00:55:06.170 --> 00:55:10.679 framework for that and I'm, I hit my limit with 1055 00:55:10.699 --> 00:55:13.579 the Copilot Plus PC and I'm waiting for these new 1056 00:55:13.679 --> 00:55:17.099 ones to come out. But that's a whole different conversation 1057 00:55:17.119 --> 00:55:20.139 because there's so many people, so many clients and customers 1058 00:55:20.179 --> 00:55:22.610 I talk to that can't use public models. 1059 00:55:22.690 --> 00:55:23.190 Right, right. 1060 00:55:23.389 --> 00:55:25.250 There's whole classes of work that are just not an 1061 00:55:25.349 --> 00:55:27.570 option to be going, leaving your machine in any way. 1062 00:55:28.110 --> 00:55:31.969 Combined with tokens are only getting more expensive and all 1063 00:55:31.989 --> 00:55:36.230 you can eat is disappearing and The methodologies we've been 1064 00:55:36.369 --> 00:55:39.269 using with the sort of adversarial agent model are just 1065 00:55:39.489 --> 00:55:40.389 mo tokens. 1066 00:55:40.809 --> 00:55:45.869 And smaller models hallucinate less. Right. 1067 00:55:45.989 --> 00:55:47.809 They are more bound. 1068 00:55:47.989 --> 00:55:51.469 And then you bound it by a language, a constrained language, 1069 00:55:51.530 --> 00:55:54.909 a controlled language, and you can reduce it even worse. 1070 00:55:55.090 --> 00:55:59.860 That's kind of where I'm going with this. Still early days. 1071 00:56:00.809 --> 00:56:03.789 I do a lot of Foundry local type stuff, running 1072 00:56:03.929 --> 00:56:09.489 to issues, report bugs. There's a whole Viber community. Folks 1073 00:56:09.510 --> 00:56:12.909 who are interested in getting into dipping their toe in 1074 00:56:12.989 --> 00:56:17.469 how do I inject AI LLMs, highly recommend checking it out. 1075 00:56:17.840 --> 00:56:21.079 It's an easy install. And there's a command line tool 1076 00:56:21.119 --> 00:56:23.000 that you can use just to get started to have 1077 00:56:23.019 --> 00:56:25.989 a chat. Have some fun. And then they've got some 1078 00:56:26.030 --> 00:56:28.860 great SDKs, whatever language you want to use from. 1079 00:56:29.030 --> 00:56:29.840 NET to Rust. 1080 00:56:30.719 --> 00:56:31.340 It's out there. 1081 00:56:31.590 --> 00:56:32.690 I highly recommend it. 1082 00:56:32.920 --> 00:56:34.179 This is good stuff, Don. 1083 00:56:34.679 --> 00:56:35.059 Excellent. 1084 00:56:35.139 --> 00:56:36.460 And it got me thinking. 1085 00:56:36.519 --> 00:56:40.599 Yeah, definitely got us thinking. Where can we find you 1086 00:56:40.719 --> 00:56:42.059 online besides your Substack? 1087 00:56:42.860 --> 00:56:49.719 My Substack and LinkedIn. I do both. LinkedIn, it's marketing 1088 00:56:50.059 --> 00:56:54.949 hell right now. People are all complaining. Um, it's not 1089 00:56:54.989 --> 00:56:59.230 where the developers are. Right. And Substack, I tried Reddit. 1090 00:56:59.769 -->

00:57:06.510 I tried, uh, various Discord servers. Um, my local Lancaster 1091 00:57:06.550 --> 00:57:09.670 folks have a great Slack team. So I'm there, Tech Lancaster. 1092 00:57:10.369 --> 00:57:15.360 Those folks are awesome. But Substack feels like the blogging 1093 00:57:15.420 --> 00:57:18.880 from the old days. It's, it's, it's a good community there. 1094 00:57:18.940 --> 00:57:21.199 So in the Lancaster area, do you find that there 1095 00:57:21.239 --> 00:57:22.559 are many Amish developers there? 1096 00:57:23.539 --> 00:57:26.260 There are some that you'd be surprised. 1097 00:57:26.480 --> 00:57:26.820 Wow. 1098 00:57:26.840 --> 00:57:30.739 You'd be surprised. And I wrote up a whole article on, 1099 00:57:31.219 --> 00:57:34.760 because I'm new to the area, Lancaster feels to me 1100 00:57:34.960 --> 00:57:40.889 like Portland, Oregon 20 years ago. They like that weird, 1101 00:57:41.389 --> 00:57:42.829 let's be weird attitude. 1102 00:57:42.849 --> 00:57:43.389 Right. 1103 00:57:43.489 --> 00:57:47.170 They like the back to nature, had the farming, all 1104 00:57:47.210 --> 00:57:48.130 that sort of stuff. 1105 00:57:48.530 --> 00:57:48.710 Yeah. 1106 00:57:48.730 --> 00:57:49.030 But. 1107 00:57:49.360 --> 00:57:52.679 They also like the tech side and there's two massive 1108 00:57:52.719 --> 00:57:56.179 data centers coming in. Coral Reeve is the big AI 1109 00:57:56.920 --> 00:57:59.519 company is putting a big data center and Google's putting 1110 00:57:59.579 --> 00:58:01.739 a big data center in there. So there's a lot 1111 00:58:01.820 --> 00:58:07.239 of innovation going on in that Amish community. It's an 1112 00:58:07.300 --> 00:58:08.510 interesting time. 1113 00:58:08.889 --> 00:58:09.250 Yeah. 1114 00:58:09.349 --> 00:58:12.070 Most people don't think of it that way, but land 1115 00:58:12.130 --> 00:58:14.269 is a heck of a lot cheaper. They can't build 1116 00:58:14.329 --> 00:58:18.159 any more data centers in, um, Northern Virginia. So they're 1117 00:58:18.199 --> 00:58:21.460 moving north into central PA and south into Richmond. 1118 00:58:21.679 --> 00:58:22.099 Okay. 1119 00:58:23.039 --> 00:58:25.590 Well, you have to keep us apprised. And when you 1120 00:58:25.909 --> 00:58:31.449 trade your first AI training for a shoe fly pie, 1121 00:58:31.510 --> 00:58:38.349 let me know. All right, Don, thanks very much. 1122 00:58:38.409 --> 00:58:39.650 Good to catch up with you again. 1123 00:58:39.949 --> 00:58:41.530 Good to catch up. And let's not make it in 1124 00:58:41.550 --> 00:58:42.309 another 20 years. 1125 00:58:42.530 --> 00:58:46.750 Definitely. And we'll talk to you, dear listener, next time on. 1126 00:58:46.679 --> 00:58:47.420 NET Rocks. 1127 00:59:08.389 --> 00:59:09.969 NET Rocks is brought to you by Frankenstein. 1128 00:59:10.409 --> 00:59:14.920 Net and produced by Plop Studios, a full-service audio, video, 1129 00:59:14.960 --> 00:59:18.280 and post-production facility located physically. 1130 00:59:17.820 --> 00:59:20.559 In New London, Connecticut, and, of course, in the cloud. 1131 00:59:21.260 --> 00:59:23.559 Online at pwop.com. 1132 00:59:24.280 --> 00:59:32.179 Visit our website at dotnetrocks.com for RSS feeds, downloads, mobile apps, comments, 1133 00:59:32.619 --> 00:59:35.079 and access to the full archives going back to show 1134 00:59:35.119 --> 00:59:39.440 number one, recorded in September 2002. And make sure you 1135 00:59:39.480 --> 00:59:42.659 check out our sponsors. They keep us in business. Now 1136 00:59:42.679 --> 00:59:43.519 go write some code. 1137 00:59:44.219 --> 00:59:44.860 See you next time.

编辑文字稿