WEBVTT - Warsh’s Fed, Vibe Coding, Geothermal Energy, Vegas Bets Big 0:00:02.480 --> 0:00:07.000 Bloomberg Audio Studios, podcasts, radio news. 0:00:21.000 --> 0:00:23.760 This is Wall Street Week. I'm David Weston bringing you 0:00:23.800 --> 0:00:27.080 stories of capitalism. If you can think it, you can 0:00:27.160 --> 0:00:30.400 code it. Generatve AI brings the brave new world of 0:00:30.520 --> 0:00:34.720 vibe coding to a project near you. Plus, we're going 0:00:34.760 --> 0:00:37.120 to need a lot more power to run all that 0:00:37.240 --> 0:00:39.800 new vibe coding, and it turns out that we may 0:00:39.880 --> 0:00:42.280 get some of it from the ground under your feet 0:00:43.240 --> 0:00:48.400 and Las Vegas. It's not just for gambling anymore. But 0:00:48.520 --> 0:00:51.760 we start with the FED, which held its monetary policy 0:00:51.800 --> 0:00:54.279 meetings this week even as it waits for its new 0:00:54.360 --> 0:00:57.320 chair to be confirmed by the Senate. Glenn Hubbard was 0:00:57.360 --> 0:01:00.320 the head of President George W. Bush's National out of 0:01:00.400 --> 0:01:03.240 Council and then went on to run the Columbia Business School, 0:01:03.440 --> 0:01:07.520 where he is now on the faculty. So, Glenn, we 0:01:07.600 --> 0:01:11.640 had the Federal Open Market Committee meeting and we heard 0:01:11.640 --> 0:01:15.440 from Chair Pal they basically didn't do anything. Did they 0:01:15.480 --> 0:01:16.160 have a choice? 0:01:16.600 --> 0:01:19.240 No, But I think they did do a few things. 0:01:19.240 --> 0:01:21.560 So they certainly didn't change the funds rate. I don't 0:01:21.560 --> 0:01:25.120 think anybody expected that, or at least reasonably. The battle, 0:01:25.160 --> 0:01:28.560 of course, the descents that happened. Were over the FED statement, 0:01:28.640 --> 0:01:31.160 or at least what people perceive the FED statement to be. 0:01:31.400 --> 0:01:34.360 Is there an easing bias? Is there a tightening bias? 0:01:34.880 --> 0:01:37.600 I think it's fair to say, under the hood, members 0:01:37.640 --> 0:01:41.960 of the FOMC are conflicted. Some view that the next 0:01:42.000 --> 0:01:43.600 move is going to have to be flat for an 0:01:43.600 --> 0:01:47.720 extended period, possibly even up. Others like Governor Myron have 0:01:48.000 --> 0:01:50.720 a different sign. So that's really what happened. 0:01:51.080 --> 0:01:52.520 One of the things I couldn't figure out is the 0:01:52.520 --> 0:01:55.360 easing bias language that they descended from the three members. 0:01:55.680 --> 0:01:57.360 I looked at the same of carefully. I didn't find 0:01:57.400 --> 0:01:58.200 those words in there. 0:01:58.440 --> 0:02:01.080 No, I don't think it's there in black and white. 0:02:01.360 --> 0:02:04.000 I think probably this is a reading the room kind 0:02:04.080 --> 0:02:08.000 of observation of people going on record that we really 0:02:08.040 --> 0:02:12.000 are uncomfortable with a view that at the next FOMC 0:02:12.320 --> 0:02:14.040 meeting or maybe even the meeting after that, that a 0:02:14.160 --> 0:02:15.240 cut is forthcoming. 0:02:16.000 --> 0:02:18.000 They did say in the statement that there was extreme 0:02:18.160 --> 0:02:20.640 uncertainty coming out of the Middle East because of the 0:02:20.639 --> 0:02:25.160 Iran war. Is that what is driving the concern about 0:02:25.240 --> 0:02:28.120 rates or was there a pre existing issue with inflation? 0:02:28.240 --> 0:02:30.640 I would have to give you the classic economist answer 0:02:30.840 --> 0:02:33.920 yes and no. So yes, the uncertainty is a problem. 0:02:34.000 --> 0:02:37.200 Oil prices are high, the US economy is less energy intensive, 0:02:37.240 --> 0:02:41.040 but it will still raise at least headline inflation. Let's 0:02:41.080 --> 0:02:45.360 remember inflation was stock well above the fed's target before 0:02:45.400 --> 0:02:49.760 February twenty eighth, So Iran adds to the problems the 0:02:49.800 --> 0:02:52.720 FED faces, but it's not really the core problem. I 0:02:52.760 --> 0:02:55.519 think what's vexing the FED is why it was more 0:02:55.560 --> 0:02:58.480 difficult to get inflation down before Iran. But yes, the 0:02:58.560 --> 0:02:59.639 uncertainty matters. 0:03:00.080 --> 0:03:03.079 What about inflation expectations because I've seen some indications now 0:03:03.120 --> 0:03:04.640 they may be rising in the United States and even 0:03:04.720 --> 0:03:05.280 in Europe. 0:03:05.720 --> 0:03:08.880 So far, I think they're relatively anchored. That's a good 0:03:08.919 --> 0:03:11.160 news for the FED, and that's far more important for 0:03:11.200 --> 0:03:14.680 the FED than whatever the inflation number by whatever measure 0:03:14.760 --> 0:03:17.400 they look at is today. But I think the risk 0:03:17.600 --> 0:03:21.240 is if the war continues and prices remain elevated, and 0:03:21.240 --> 0:03:23.600 of course we still have tariffs working their way through 0:03:23.639 --> 0:03:26.880 the system. Even if the ultimate effect is to stabilize, 0:03:27.680 --> 0:03:30.120 that's going to be an inflation problem and that could 0:03:30.200 --> 0:03:31.440 unhinge expectations. 0:03:32.080 --> 0:03:34.280 So what does this mean for the new chair coming 0:03:34.280 --> 0:03:35.360 in Kevin. 0:03:35.160 --> 0:03:36.760 Walsh good luck. 0:03:37.320 --> 0:03:40.760 I mean Kevin Walsh is actually well suited for this moment. 0:03:41.320 --> 0:03:46.440 He has outstanding small py political skills, consensus building skills. 0:03:46.800 --> 0:03:49.200 He's very smart, he knows these issues. I think his 0:03:49.360 --> 0:03:52.040 challenge is going to be looking at the men and 0:03:52.080 --> 0:03:55.520 women around the table, whether it's the governors or the FMC, 0:03:56.000 --> 0:03:58.120 and figuring out how to bring them together. That's going 0:03:58.200 --> 0:04:01.560 to require two things. His skills, which as I said, 0:04:01.560 --> 0:04:04.000 he has, but another's the theory of the case. I mean, 0:04:04.000 --> 0:04:07.160 how do you talk about the economy. I think that'll 0:04:07.200 --> 0:04:09.040 have to be job one for our new chair. 0:04:09.520 --> 0:04:12.800 Some are interpreting that three person descent as putting a 0:04:12.840 --> 0:04:17.000 marker down really almost against President Trump. With President Trump's 0:04:17.000 --> 0:04:19.719 pressing for lower rates, don't go too far, too fast. 0:04:20.040 --> 0:04:21.800 I'm not sure that that's true. I mean, it is 0:04:21.880 --> 0:04:24.520 true President Trump would prefer lower rates. He certainly has 0:04:24.680 --> 0:04:27.279 told us all that. I think it's more a matter 0:04:27.360 --> 0:04:30.760 of a sincere difference of opinion and policy. And given 0:04:30.800 --> 0:04:33.640 where we are right now, with the labor market softening 0:04:33.640 --> 0:04:36.559 a little yet we have a robust economy, inflation is stuck. 0:04:36.600 --> 0:04:39.240 You can see why people are wondering, is the glass 0:04:39.320 --> 0:04:42.960 half full or half empty. When you have uncertainty, more 0:04:43.000 --> 0:04:46.440 often than not, watching and waiting until you get more 0:04:46.440 --> 0:04:48.920 information is the right answer, and I think it is here. 0:04:49.320 --> 0:04:51.560 Is it a symmetric risk right now as you look 0:04:51.600 --> 0:04:54.200 at it, or is there a real risk of stagflation 0:04:54.360 --> 0:04:57.279 where you actually could hurt growth and still have inflation. 0:04:57.360 --> 0:04:59.839 Well, there's certainly a risk of stagflation, but I would 0:04:59.839 --> 0:05:03.400 be more worried about inflation at the moment. The underlying 0:05:03.440 --> 0:05:07.360 economy is very good and the GDP numbers we got 0:05:07.360 --> 0:05:10.599 for the first quarter are reassuring still about the pace 0:05:10.680 --> 0:05:13.760 of AI investment and about final sales, even though the 0:05:13.839 --> 0:05:16.320 headline print is a little below expectations. 0:05:17.120 --> 0:05:20.040 Kevin worsh comes in saying that he thinks there is 0:05:20.080 --> 0:05:22.600 a room for rate cuts because of AI. 0:05:23.320 --> 0:05:24.200 How does that work? 0:05:25.800 --> 0:05:28.360 Well, the story, if it's correct, is what I would 0:05:28.360 --> 0:05:30.640 come more of a long run story. So AI in 0:05:30.680 --> 0:05:34.680 the long run, certainly, as the potential to be disinflationary 0:05:34.839 --> 0:05:39.440 reduces the costs, particularly in service producing sectors. Think about 0:05:39.480 --> 0:05:42.640 the costs of producing what lawyers do, or accountants do, 0:05:42.760 --> 0:05:46.000 or engineers or dare I say, economists, that could all 0:05:46.040 --> 0:05:49.720 be disinflationary. The problem is that's not today, it's not tomorrow, 0:05:49.920 --> 0:05:51.560 and it may not even be a year or two 0:05:51.560 --> 0:05:56.200 from now. The immediate effect of AI is the aggregate 0:05:56.240 --> 0:05:59.800 demand effect of building data centers. The investment numbers that 0:06:00.000 --> 0:06:03.840 showed up so robust in GDP all a SQL that raises, 0:06:04.040 --> 0:06:07.280 not lowers the real interest rate, and inflation is stuck. 0:06:07.760 --> 0:06:11.960 So I think incoming chair Worsh's story makes sense as 0:06:12.000 --> 0:06:15.280 a potential long run hypothesis. I don't think it could 0:06:15.320 --> 0:06:17.560 be a theory of the case for cutting rates now. 0:06:17.839 --> 0:06:19.480 To cut rates now, you'd have to have a view 0:06:19.520 --> 0:06:23.240 that the economy's weakening is such that that tramps small 0:06:23.320 --> 0:06:25.280 teeth the inflationary pressures. 0:06:26.440 --> 0:06:30.039 There are other parts of the fed's job besides just rates. 0:06:31.320 --> 0:06:34.039 What do we expect from Kevin Warsh's chair in the 0:06:34.120 --> 0:06:35.080 other parts. 0:06:35.320 --> 0:06:37.120 Well, I think he's signaled that he wants to take 0:06:37.120 --> 0:06:39.080 a look at least in a couple of areas. One 0:06:39.200 --> 0:06:41.520 is the size of the Fed's balance sheet, which of 0:06:41.560 --> 0:06:46.120 course bloomed enormously since the global financial crisis. That's not 0:06:46.200 --> 0:06:48.080 really as simple as saying do I want a big 0:06:48.120 --> 0:06:50.960 balance sheet or a small balance sheet? Because the FED 0:06:51.040 --> 0:06:53.960 has proven to be the market maker of last resort 0:06:54.000 --> 0:06:57.080 in the treasury market multiple times, and the treasury market 0:06:57.200 --> 0:07:01.599 is the market for the world's safe asset. Regulation perhaps 0:07:01.720 --> 0:07:05.039 unwittingly made it hard for private market makers to do 0:07:05.080 --> 0:07:07.359 the job they could be doing. So a FED balance 0:07:07.360 --> 0:07:10.080 sheet is pretty important. But I do expect Kevin worsh 0:07:10.120 --> 0:07:13.040 to spend a lot of time there the other's financial regulation, 0:07:13.480 --> 0:07:17.080 especially since the Global financial crisis, the FED got more 0:07:17.200 --> 0:07:20.920 territory in financial regulation, so there may be a look, 0:07:21.000 --> 0:07:22.720 do we have the right rules? Do we have the 0:07:22.840 --> 0:07:27.040 right capital regime? I would expect incoming Chairwash to take 0:07:27.080 --> 0:07:30.320 a hard to take a hard look at that a 0:07:30.360 --> 0:07:34.840 warning though, regulation, unlike monetary policy, is a political subject, 0:07:35.200 --> 0:07:38.480 and so do expect Congress. Do expect the President to, 0:07:39.240 --> 0:07:42.040 with good merits, have political views to balance. 0:07:42.720 --> 0:07:44.640 For the question that's been raised by some other members 0:07:44.640 --> 0:07:47.560 the FED right now is the relationship of the regional 0:07:48.760 --> 0:07:51.560 Fed to the national Fed. Then, in fact, maybe there 0:07:51.560 --> 0:07:55.920 should be more power or authority or responsibility taken into Washington. 0:07:55.960 --> 0:07:56.720 Does that make sense? 0:07:57.360 --> 0:08:00.200 I don't think so. I think the decentralized structure the 0:08:00.200 --> 0:08:04.000 FED is a feature, not a bug. Remember the history 0:08:04.080 --> 0:08:06.920 lesson for the FED is it was done to get 0:08:07.000 --> 0:08:09.800 points of view throughout the nation, where a nation of 0:08:09.840 --> 0:08:14.040 different business sectors and different geographies. That said, there is 0:08:14.080 --> 0:08:17.400 a point that I think regional presidents may be communicating 0:08:17.480 --> 0:08:22.080 too much. In any organization, the top of the organization 0:08:22.240 --> 0:08:26.080 needs to set the tone internally and externally. So having 0:08:26.080 --> 0:08:28.600 a lot of debate inside strikes me as a good thing. 0:08:28.880 --> 0:08:31.400 Doing it out in the press or on television, I'm 0:08:31.480 --> 0:08:32.839 less persuaded that's a good thing. 0:08:33.120 --> 0:08:35.959 There's a good deal to talk about what a chair 0:08:36.240 --> 0:08:39.240 wash will do in terms of, for example, forward guidance 0:08:40.040 --> 0:08:43.160 and even number of news conferences that are held the 0:08:43.200 --> 0:08:45.720 dot plots. How much of that is in the authority 0:08:45.720 --> 0:08:47.840 of the chair? Does that need to be voted on 0:08:48.040 --> 0:08:49.480 by the full fed Well, he. 0:08:49.480 --> 0:08:51.840 Would need to get his colleague support, but I suspect 0:08:51.840 --> 0:08:54.000 he could get it. I count myself as among those 0:08:54.040 --> 0:08:56.720 who wonder what the utility of the dot plot is 0:08:57.280 --> 0:09:01.640 and the excessive communication. There's a continuum, if you will. 0:09:01.679 --> 0:09:05.280 Chair Greenspan was quite opaque. He famously said, you know, 0:09:05.280 --> 0:09:07.640 if you thought you understood me, you must be mistaken. 0:09:08.200 --> 0:09:13.400 To Chair Bernanki, who was extremely transparent, I think probably 0:09:13.440 --> 0:09:17.200 an incoming Chairwash wants to be somewhere in between. 0:09:17.280 --> 0:09:17.480 There. 0:09:17.880 --> 0:09:21.079 One thing that incoming chair worsh will have that has 0:09:21.120 --> 0:09:23.160 not been had for a very long time is a 0:09:23.200 --> 0:09:25.640 former chair sitting in the room at least for some 0:09:25.679 --> 0:09:28.160 period of time. What will that do to that dynamic? 0:09:28.200 --> 0:09:28.680 Do you think? 0:09:29.559 --> 0:09:32.360 I'm not sure it's actually going to do very much. 0:09:33.280 --> 0:09:36.520 I think Chair Powell's current Chair Powell's points of view 0:09:36.520 --> 0:09:38.960 are well known. He said them many times. I expect 0:09:39.040 --> 0:09:42.080 he'll repeat those as a governor, and the time he 0:09:42.120 --> 0:09:44.840 has left, I think he also is somebody who would 0:09:44.840 --> 0:09:47.600 be deferential in the way he would be to a 0:09:47.679 --> 0:09:49.880 new chairs. I don't think it's going to change things 0:09:50.320 --> 0:09:53.400 very much. Whether he leaves is a personal decision, and 0:09:53.440 --> 0:09:54.720 he'll have to make it. 0:09:54.600 --> 0:09:56.000 The independence of the FED. 0:09:56.640 --> 0:09:59.840 We also spoke with former US Treasury Secretary Hank Paulson 0:10:00.040 --> 0:10:03.720 about the challenges facing the next FED chair. Paulson agreed 0:10:03.760 --> 0:10:06.480 that Walsh has a hard road ahead, but that he 0:10:06.640 --> 0:10:07.760 is up to the task. 0:10:08.280 --> 0:10:10.840 Well, you have a chairman of the FED, you want 0:10:10.920 --> 0:10:15.520 someone who understands markets, who is a good community gator, 0:10:15.880 --> 0:10:20.240 has a good understanding of sound economic principles, and I 0:10:20.280 --> 0:10:24.720 think Kevin Walsh meets those tests right, and he's been before, 0:10:24.800 --> 0:10:28.760 he's done it. Now, I think his job becomes more 0:10:28.800 --> 0:10:34.079 difficult because there's independence of the perception of independence. So 0:10:34.200 --> 0:10:38.640 I think how this transition between Jay Paul, who's just 0:10:38.840 --> 0:10:43.800 been a star performer and got great credibility and integrity. 0:10:44.080 --> 0:10:47.760 I think how that transition is handled and how the 0:10:47.840 --> 0:10:51.040 administration deals with it is either going to make Kevin's 0:10:51.160 --> 0:10:56.480 job more difficult or you know, somewhat easier. But Kevin 0:10:56.520 --> 0:11:01.400 won't have an easy job anyway. And I compliment the 0:11:01.480 --> 0:11:03.800 Trump administration. I'm selecting him right. 0:11:05.800 --> 0:11:08.520 Coming up, AI may be coming for a job near you, 0:11:08.920 --> 0:11:22.720 especially if you're a computer programmer. This is a story 0:11:22.760 --> 0:11:25.800 about a double edged sword. At this point, AI is 0:11:25.880 --> 0:11:28.960 promised to do just about everything, but it turns out 0:11:28.960 --> 0:11:31.200 that it's best at creating the very thing that it 0:11:31.280 --> 0:11:34.559 is made of. Code. It was once considered a high 0:11:34.640 --> 0:11:37.760 skill task, but coding is now accessible to just about 0:11:37.800 --> 0:11:41.400 anyone with the help of generative AI. On the one hand, 0:11:41.559 --> 0:11:44.520 this unlocks possibilities for creating a wide range of products 0:11:44.600 --> 0:11:47.760 and businesses that otherwise might never have seen the light 0:11:47.800 --> 0:11:50.640 of day. On the other hand, it's made the future 0:11:50.720 --> 0:11:54.240 less certain for those who write code professionally. Our colleague 0:11:54.320 --> 0:11:56.760 Ed Ludlow brings us the story of what happens when 0:11:56.760 --> 0:11:59.320 we lean in and let AI do the work. 0:12:05.360 --> 0:12:08.800 In Upper Tracts, West Virginia, Jamie Grove owns a boutique 0:12:08.840 --> 0:12:12.560 warehouse helping clients ship out anything from dinosaur bones to 0:12:12.600 --> 0:12:16.959 board games. Last year, he built software that automates shipping 0:12:16.960 --> 0:12:19.480 out packages with help from AI. 0:12:19.920 --> 0:12:21.640 Where's the order at that you're looking at? 0:12:21.640 --> 0:12:21.840 There? 0:12:22.280 --> 0:12:24.160 There's absolutely no way I could have done any of 0:12:24.200 --> 0:12:27.840 this without AI. For the initial run to take some 0:12:27.880 --> 0:12:30.360 spreadsheets that we had that we were working with to 0:12:30.400 --> 0:12:33.920 get an actual workable sample ready to go took me 0:12:34.040 --> 0:12:36.600 less than a day and we were live and we 0:12:36.600 --> 0:12:39.400 were using it right here in the warehouse. 0:12:39.520 --> 0:12:43.040 In Oakland, California. Cynthia Chen wished there was an app 0:12:43.160 --> 0:12:45.920 that collects pictures of different breeds of dogs. 0:12:46.280 --> 0:12:48.679 And then it was when I learned about vibe coding online. 0:12:49.280 --> 0:12:51.440 Is when I thought maybe I should try it myself. 0:12:52.040 --> 0:12:54.600 The first time, you see like this little pop up 0:12:54.640 --> 0:12:57.240 and it says build succeeded, and then you can see 0:12:57.240 --> 0:13:00.280 the app pop up and like it's actually real. That 0:13:00.440 --> 0:13:03.160 was like the sort of magical moment where I was like, 0:13:03.200 --> 0:13:05.160 oh my gosh, this is crazy. I can actually build 0:13:05.160 --> 0:13:05.840 things myself. 0:13:08.000 --> 0:13:11.680 The term vibe coding was coined by Andre Carpathy, a 0:13:11.760 --> 0:13:15.000 founding member of open Ai, to describe a process of 0:13:15.040 --> 0:13:19.040 computer programming akin to having a conversation with a robot. 0:13:19.720 --> 0:13:21.720 Here's what it looks like in practice. Say I want 0:13:21.760 --> 0:13:25.360 to create a website that visualizes and animates a data set. 0:13:25.640 --> 0:13:28.240 I would use a generative AI tool like Claude or 0:13:28.280 --> 0:13:31.319 codex or in this case, Gemini and tell it exactly 0:13:31.320 --> 0:13:34.520 what I want using plain English. AI then writes the 0:13:34.559 --> 0:13:37.480 code for me, and as coding has gotten easier, it's 0:13:37.559 --> 0:13:40.160 led to the creation of more code than ever before. 0:13:40.480 --> 0:13:43.920 Activity on GitHub, the platform used for storing and sharing code, 0:13:44.000 --> 0:13:46.800 has seen a massive increase in activity, surging in early 0:13:46.840 --> 0:13:53.360 twenty twenty five when AI coding pilots became popular. And 0:13:53.440 --> 0:13:57.319 although vibe coding has helped small businesses and hobbyists create 0:13:57.360 --> 0:14:00.720 their own software, it's the professionals who leading the way. 0:14:01.080 --> 0:14:03.200 So what goes on in this building is we have 0:14:03.280 --> 0:14:06.800 a mix of folks working on Google Cloud. At Google, 0:14:06.880 --> 0:14:10.520 when we talk about autonomy and agentic engineering, we have 0:14:10.640 --> 0:14:13.000 systems that allow you to kind of have almost a 0:14:13.120 --> 0:14:15.160 virtual software engineer. 0:14:15.360 --> 0:14:20.640 Hold on, if all these brilliant engineers are using AI 0:14:20.800 --> 0:14:23.240 in this way, what is it that they're doing all 0:14:23.320 --> 0:14:25.440 day long in beautiful buildings like this? 0:14:25.480 --> 0:14:27.320 One excellent question. 0:14:29.840 --> 0:14:33.840 Adilsmani is the director of Google Cloud AI. He oversees 0:14:33.880 --> 0:14:37.280 teams of engineers currently building the next generation of AI 0:14:37.400 --> 0:14:38.760 tools for businesses. 0:14:39.160 --> 0:14:41.640 So, if you are vibe coding, you're pretty much just 0:14:41.640 --> 0:14:44.240 giving into the vibes. You don't necessarily have a clear, 0:14:44.320 --> 0:14:47.720 full idea of your vision. You're just working with the LLM. 0:14:47.960 --> 0:14:51.320 You're trying to get somewhere with it. If you are engineering, 0:14:51.680 --> 0:14:53.880 that's where you have to apply rigor to it. You 0:14:53.960 --> 0:14:57.760 have to have this clear set of requirements you are testing, 0:14:58.120 --> 0:15:00.480 and whether you are a startup or whether you're in 0:15:00.560 --> 0:15:03.600 a big enterprise. Right now, the role of the software 0:15:03.640 --> 0:15:06.160 engineer is going to be evolving to one where you 0:15:06.200 --> 0:15:08.840 are increasingly a little bit more of a manager. You're 0:15:08.880 --> 0:15:12.280 going to have effectively like a virtual team of agents 0:15:12.280 --> 0:15:15.640 that you're responsible for and you have to own the outcomes. 0:15:15.720 --> 0:15:19.160 Doesn't matter how many are responsible the output, You're responsible 0:15:19.200 --> 0:15:22.880 for the output exactly. And so you need to decide 0:15:22.920 --> 0:15:26.760 like how am I evaluating quality? How much time am 0:15:26.760 --> 0:15:28.840 I going to spend evaluating quality? Because there are some 0:15:28.880 --> 0:15:32.040 people who very much enjoy YOLO, like, okay, well the 0:15:32.080 --> 0:15:35.120 agents ran overnight looks good kind of runs, I'm just 0:15:35.160 --> 0:15:37.440 going to deploy it. But if you're building any kind 0:15:37.480 --> 0:15:39.960 of serious software, you still need to have some idea 0:15:40.000 --> 0:15:42.160 of like what is what is the quality bar? What 0:15:42.200 --> 0:15:44.720 are my quality gates? How am I making sure this 0:15:44.800 --> 0:15:47.520 is actually going to meet the needs of my users 0:15:47.520 --> 0:15:48.440 in a consistent way. 0:15:50.960 --> 0:15:54.240 For the engineers at Google's headquarters is here in Sunny Vale, California, 0:15:54.320 --> 0:15:57.720 Osmani says, AI isn't just making their lives easier, it's 0:15:57.760 --> 0:16:01.280 making them better at coding. The extension of that question, 0:16:01.360 --> 0:16:04.720 which we pose largely by investors, is how do we 0:16:04.840 --> 0:16:09.960 measure the productivity gains of that engineer or that team. 0:16:10.200 --> 0:16:12.800 I remember in the earlier days of AI, you know, 0:16:13.000 --> 0:16:15.240 org leaders would look at things like, oh, hey, well, 0:16:15.280 --> 0:16:18.520 how many lines of code are being generator? Which is 0:16:18.560 --> 0:16:21.960 not in any way a good proxy for productivity. But 0:16:22.520 --> 0:16:24.440 these days, I think that people use a mix of 0:16:24.480 --> 0:16:27.320 different kinds of metrics. You try to use qualitative and 0:16:27.400 --> 0:16:30.920 quantitative generally speaking, there's a big productivity boost. In the 0:16:30.960 --> 0:16:32.880 earlier days of AI, would have said, you know, that 0:16:32.920 --> 0:16:35.280 boost is ten to fifteen percent. These days it's anywhere 0:16:35.280 --> 0:16:37.520 from thirty to fifty percent, and I see that number 0:16:37.560 --> 0:16:38.800 only continuing to go up. 0:16:41.600 --> 0:16:44.920 At MIT, Frank Nagel and a team of research has 0:16:44.960 --> 0:16:48.960 surveyed over one hundred and eighty seven thousand software developers 0:16:49.160 --> 0:16:52.560 who are using GitHub co pilot, a generative AI tool 0:16:52.640 --> 0:16:55.720 for coding, and they found that workers are more productive 0:16:55.800 --> 0:16:58.480 because what they spend time on has changed. 0:16:58.760 --> 0:17:00.440 I think one of the big things that we often 0:17:00.440 --> 0:17:02.360 think about with AI is that it's just going to 0:17:02.480 --> 0:17:05.760 enhance productivity, right, It's going to make us faster doing 0:17:05.840 --> 0:17:09.000 whatever it is we do. But that's just really just 0:17:09.040 --> 0:17:11.520 scratching the surface. You have one hundred percent of your 0:17:11.520 --> 0:17:13.960 time that you allocate to work. How did that break 0:17:14.000 --> 0:17:16.720 down along these dimensions of what we called core work, 0:17:16.920 --> 0:17:20.520 actual coding versus more project management type of work. And 0:17:20.560 --> 0:17:23.960 what we found is that when coders started using these 0:17:24.000 --> 0:17:27.040 types of tools, they massively shift the amount of their 0:17:27.119 --> 0:17:30.720 time that they allocate to coding, and they take away 0:17:30.800 --> 0:17:33.840 a whole lot of their allocated time towards from project management, 0:17:34.119 --> 0:17:35.840 and so part of the reason we think that this 0:17:35.920 --> 0:17:38.560 is happening is that if in the old days you 0:17:38.640 --> 0:17:41.359 were writing piece of code A, and that piece of 0:17:41.359 --> 0:17:43.800 code was dependent on some other piece of code B, 0:17:44.320 --> 0:17:46.320 you had to wait for that other person, You had 0:17:46.400 --> 0:17:49.280 to interact with them and make sure that everything worked together, 0:17:49.480 --> 0:17:51.680 whereas now you can just write it all yourself. 0:17:54.720 --> 0:17:57.639 This productivity boost is particularly true for those who are 0:17:57.640 --> 0:18:00.919 writing and deploying brand new code, and especially those with 0:18:01.040 --> 0:18:04.160 no knowledge of code whatsoever, including creatives like Chen. 0:18:04.640 --> 0:18:07.119 This is press Pedals, the new app that I'm working on, 0:18:07.600 --> 0:18:11.000 and here is a press Flower. So I'm going to 0:18:11.040 --> 0:18:14.000 go to the cloud code that I have running, and 0:18:14.040 --> 0:18:16.640 then I'm simply going to describe what I wanted to do. 0:18:17.119 --> 0:18:19.280 I actually tried a whole bunch of tools in the beginning, 0:18:19.280 --> 0:18:21.040 and this was about a year ago, which I think 0:18:21.119 --> 0:18:23.840 is like feels kind of like the stone ages of 0:18:23.960 --> 0:18:27.199 vibe coding. I was able to make the foundations of 0:18:27.240 --> 0:18:31.240 the app in maybe a month, and as like a 0:18:31.280 --> 0:18:34.720 totally non technical person, I was actually able to build 0:18:34.760 --> 0:18:37.080 a full stack app, so there's like front end and 0:18:37.119 --> 0:18:39.240 back end capabilities, and I was able to get it 0:18:39.280 --> 0:18:41.520 out on the App Store, just entirely myself. 0:18:45.080 --> 0:18:48.359 For business owners like Grove, his Vibe coded solution is 0:18:48.400 --> 0:18:50.440 helping his warehouse save on costs. 0:18:50.800 --> 0:18:53.600 We have three main coding solutions here that we've used 0:18:53.600 --> 0:18:57.840 AI for. One is to create batching, which is the 0:18:57.880 --> 0:19:01.439 important part in terms of taking work that are different 0:19:01.840 --> 0:19:04.480 and getting them all together into groups that make it 0:19:04.520 --> 0:19:07.480 easy to pick. And then the other solution that we 0:19:07.560 --> 0:19:11.800 have is inventory tracking. So that's a standard warehouse feature, 0:19:11.840 --> 0:19:14.600 but our clients are all so very different, and so 0:19:14.800 --> 0:19:18.120 trying to force a client into a single inventory tracking 0:19:18.160 --> 0:19:21.560 system is really difficult. So we actually we've used AI 0:19:21.640 --> 0:19:25.199 to build an inventory tracking solution that allows them to 0:19:25.480 --> 0:19:29.320 be themselves basically. So I have a pretty varied background. 0:19:29.440 --> 0:19:32.280 I started out as a programmer a long long time ago. 0:19:32.640 --> 0:19:37.560 But even someone who is a very fast coder could 0:19:37.560 --> 0:19:41.360 not have built all these solutions impossible. If I were 0:19:41.400 --> 0:19:43.120 to do it with a team of programmers, I could 0:19:43.119 --> 0:19:45.760 have five programmers working on this full time and still 0:19:45.800 --> 0:19:49.720 not deliver as many results. If I were to install 0:19:49.760 --> 0:19:52.480 a software system that does everything that we're doing now, 0:19:53.160 --> 0:19:56.760 let's just say without all the customizations and all the flexibility, 0:19:57.119 --> 0:19:59.439 we might be talking about an annual license of anywhere 0:19:59.440 --> 0:20:02.320 between six and ten thousand dollars a year, scaling all 0:20:02.359 --> 0:20:04.680 the way up to maybe thirty to fifty thousand dollars 0:20:04.720 --> 0:20:06.800 a year depending on how much volume we push through 0:20:06.800 --> 0:20:10.840 our warehouse. This doesn't have the flexibility that we would want, 0:20:10.880 --> 0:20:14.280 and it's expensive for a small boutique warehouse. 0:20:14.320 --> 0:20:15.200 That's a big. 0:20:15.000 --> 0:20:21.119 Expense now for about twenty dollars a month. Business owners 0:20:21.280 --> 0:20:24.600 like Growth are building their own software solutions, and whilst 0:20:24.640 --> 0:20:28.840 that's opened up new possibilities, there is a downside. Since 0:20:28.880 --> 0:20:31.879 twenty twenty two, employment for software engineers right out of 0:20:31.880 --> 0:20:36.560 college has fallen by nearly twenty percent. Nagel thinks companies 0:20:36.640 --> 0:20:37.960 are making a big mistake. 0:20:38.400 --> 0:20:40.240 I do think that one of the biggest risks of 0:20:40.280 --> 0:20:42.359 the whole thing is that people are going to get 0:20:42.400 --> 0:20:45.000 too focused on the short term and not think about 0:20:45.040 --> 0:20:47.600 the long term enough. First of all, you don't hire 0:20:47.640 --> 0:20:49.399 any new people who's going to run the company in 0:20:49.400 --> 0:20:52.440 ten to fifteen years. But second of all, our research 0:20:52.680 --> 0:20:55.479 and others has shown that these junior people are actually 0:20:55.480 --> 0:20:57.919 the ones who are able to change their job and 0:20:58.040 --> 0:21:00.439 get the most out of using these tools. And so 0:21:00.480 --> 0:21:02.600 if we're not hiring them at the same rate we 0:21:02.600 --> 0:21:04.680 were before, then we're not going to be able to 0:21:04.720 --> 0:21:05.560 take advantage of that. 0:21:06.200 --> 0:21:07.440 And Chan agrees. 0:21:08.000 --> 0:21:13.720 I feel very strongly that this is not replacing engineers. 0:21:14.400 --> 0:21:16.520 I think the more you use AI to build, the 0:21:16.520 --> 0:21:18.840