Reid riffs on Trump’s $100K visa fee, 3-day work weeks, and AI trust issues

This week, Reid and Aria dive into the debate over new $100,000 fees on H-1B visas, and discuss how immigration powers U. S. innovation and how to ensure startups aren’t shut out. They also cover developers’ mixed feelings about using AI in coding, the leap from 60% to 85% accuracy in medical AI predictions, and what it will take to bring regulators and clinicians on board. Plus, Reid shares why he’s skeptical of a three-day work week and reflects on the Trust in American Institutions Challenge, highlighting some inspiring finalists. For more info on the podcast and transcripts of all the episodes, visit https://www.possible.fm/podcast/

• 0:00Speaker 00

  Some people in the tech community are saying this is going to destroy innovation. This is, we are getting the best talent from all over the world to come here.

• 0:07Speaker 01

  If you haven't found something where the tool, where AI is useful to you in a serious way, then you haven't tried hard enough because it exists for everybody right now.

• 0:17Speaker 00

  Reid, great to be here today.

• 0:19Speaker 01

  Yes, in person in New York.

• 0:21Speaker 00

  Exactly, in this beautiful setting. So the tech world is abuzz because last week, President Trump announced that they were going to institute $100,000, perhaps annual, we don't know, some people thought annual, some people thought not, on H1B visas. And as we all know, Microsoft, Amazon, Google, these big tech companies really rely on H1B visas for a lot of their workers. And so some people in the tech community are saying this is going to destroy innovation. This is, we are getting the best talent from all over the world to come here. Other people are saying, these big tech companies, they can afford it. This is fantastic. This is a way to raise revenue, but still have these employees here. What do you think about this new potential, we'll see, it's an evolving topic,

• 1:05Speaker 00

  this potential idea of $100,000 fees on H1B visas annually?

• 1:09Speaker 01

  Well, so this is a funny, and you know, here we are, so I'll share a first awkward moment, which is actually Trump's idea resembles an idea that I've been pitching for eight plus years. So I was like, okay, right. Now, I think you have to do the whole idea, otherwise it's a disaster. And the problem is it's only a part, which is roughly the idea that I've been pitching is, you should have unlimited H1Bs, you should impose an additional tax on them, whether 100K, once, a yearly, whatever is the right thing. And you should make some, you know, kind of provisions for startup companies, right? Because the startup companies obviously can't afford the, like 100K or that sort of thing.

• 1:53Speaker 00

  So for startups, would it be cheaper than for the Amazon of the world?

• 1:56Speaker 01

  Yes, right, because it's, how do we get, because you absolutely want the talent. That is one of our great superpowers from the very founding of our country is immigration. You want that talent. And by the way, does it take? American jobs like no if this person comes over here and does this this this high paid job. Cause it should be there. There's a bunch of regulation about the fact that it has to be. You know, comparable salaries and all the rest does this high paid job. Then there are also, um, you know, uh, patronizing restaurants, hiring accountants, using dry cleaners. Like, like, you know, hiring electricians, renting, you know, apartments and all the rest. And all of this stuff is actually adding to our economy. That's why you want it here.

• 2:37Speaker 01

  And frankly, one of the challenges that you have to be careful about in this is that and you have to set the pricing right is. You say, well, for the large multinationals, you know, they can hire in other countries too. So you say, don't hire them here, hire them in a different country. Then America loses all of that derivative revenue. So you actually. Want that immigration, you want the people here. But adding an extra tax is good for a couple of reasons. One. Is this seriously talent? You can't get here.

• 3:03Speaker 00

  Right.

• 3:03Speaker 01

  Because then you've actually got a an economic incentive like, and that's why unlimited, because like, well. Hey, if I'm paying extra for it, I'm only going to pay for it if I can't hire it locally. But then we're bringing that talent in, we're having all that economics on our side. So unlimited H one Bs. Yes, a tax hundred K, whatever, you know, figure it out. Is it a hundred K? Is it something else? You know, it could simply be like an additional X percent payroll tax. In which case, by the way, you might even then be able to not make a special provision for startups. Because that may be within the economic.

• 3:36Speaker 00

  If it was based on revenue, if it was based on something so that smaller companies, people who are starting out were paying less. Bigger companies were paying their fair share. And so what do you say to the people who are like, yeah, that's great innovation, but like this absolutely takes an American job. This is an American tech worker coder who would be working in that job. And you say, no, that job would be overseas anyway. What's the, why doesn't that explanation work?

• 3:58Speaker 01

  Well, so the, the. For most of these companies, they're they are hiring. There's a quality bar they want to get to. But at that quality bar, they're more or less hire as many people as they can find. Right, so and there's all kinds of natural incentives why they would hire an American.

• 4:16Speaker 00

  Yeah.

• 4:16Speaker 01

  And the usual complaint is, oh, you're lowering the salaries because the immigrant will take the lower salary. Now, there's a bunch of different regulation about how to make that not happen, what you have to prove and all the rest of that within H1Ps. But that's part of the reason why, like, you know, I think it may even be 10 years ago when I started peddling an idea like this. It's like, look, if you just make it structurally more expensive for the companies, that will naturally play it out. And so any answer of, well, could you hire an American? It's like, well, if hiring Americans cheaper for me, I'll hire an American. Right, so then you you don't have to get the, well, does the regulation work? Am I am I actually really not lowering wages across it by bringing in people, et cetera, et cetera?

• 4:58Speaker 01

  Like, and that's part of the reason why this portion of the idea

• 5:02Speaker 00

  Yep.

• 5:02Speaker 01

  of making it more expensive for companies is, I think, a good system.

• 5:07Speaker 00

  Gotcha. So ultimately, agree with making it more expensive, but there has to be provisions for startups because we cannot just allow that ecosystem.

• 5:15Speaker 01

  By the way, you should make it unlimited. Right, like, and that's part of the dumb thing that exists right now is this lottery system and all the rest of it.

• 5:22Speaker 00

  Absolutely. Okay, so when we talk about AI, we often talk about programmers, developers, like, those are the people who especially now can get some of the very early benefits of AI. And a new survey from Stack Overflow said that 84% of software developers use AI or are going to use AI in creating their code, which is super positive. But 46% have real concerns and don't trust the code that they're deploying. So they, half of them say they waste time debugging, they have to go check the code, they go get a second opinion. And so the question is, how do we bridge this gap from software developers using AI but not trusting it? Or do we not need to bridge it? Is it good to get a second opinion, to debug, to make sure to to check the AI output?

• 6:07Speaker 00

  Or is this a problem that needs to be solved?

• 6:09Speaker 01

  Well, it's definitely not a problem that we need to focus on. So the baseline is it's good, not because there's necessarily saying AI code is buggy, but it's good to be diligent about what the outputs are and what you're doing and how to do that. And of course, we're in the early stages of both this, this, you know, in the, in the classic, the worst AI you're going to use is the AI you're using today, and so it's, and it's accelerating. But also like, look, what is our new pattern of doing software development? And how is the software development work relative to, you know, kind of like, producing all different kinds of code? And this is one of the things is like some kinds of code,

• 6:52Speaker 01

  you'd say, like, for example, if it's like, hey, I'm, I'm producing, you know, kind of the equivalent of a bunch of different scripts, I wouldn't be that wonkish about it unless there was mission critical in something. I'm producing something that is, you know, kind of infrastructure code about how the whole service is working. I'd be more wonkish about it, you know, as you should, even before AI, in terms of how this works. So, so I don't think it's a, it's a big challenge. And I do think that one of the things that is probably the most important thing for developers to keep in mind is to say, because the natural thing for every skilled professional, a developer, a lawyer, a doctor, every skilled professional, is to say, oh, I discovered a one bad output,

• 7:38Speaker 01

  tool not ready. Right. And you're like, nope, that's, that's a bad way of putting it because the tool's constantly improving. And one of the things that I tell people pretty constantly is if you haven't found something where the tool, where AI is useful to you in a serious way, not just, hey, what can I make from dinner from the ingredients in my fridge? Or can you craft me a sonnet for my friend's birthday or things, all of which are great. But like something that actually is in fact part of the way you're working and actually adds to your skill set and your capability, then you haven't tried hard enough. Because it exists for everybody right now. Now, and that's true for developers. So the, don't find an error, wave it off, but be constantly experimenting with, okay, how do I use this? Which are the things that work really well right now?

• 8:23Speaker 01

  Which are the ones that don't work well, right well now? But by the way, even if they don't work right, very well right now, then keep trying it in various ways and keep open minded because it's going to be improving.

• 8:36Speaker 00

  I feel like so often we hold AI to the standard of 100% when humans aren't 100%. And if we can use AI just to get better than where we are with humans, we're going to be seeing big improvement. And one place where you and I talk about this all the time, where AI can be enormously helpful is healthcare. So, a recent study from Johns Hopkins found that when they were doing pre-surgery echocardiograms, this is what they do, they get a score so that they can see in the next 30 days, how likely is this patient to have complications, a stroke, problems from surgery. And their model right now spits out a number they're able to understand, and they have 60% accuracy as to what patients are going to have complications.

• 9:19Speaker 00

  They created a new model, they fused existing echocardiogram data with also the age, type of surgery, all of this info about the patient. And they found that AI, and this was a study of 37,000 patients, AI can get it right 85% of the time. So we're going from 60% to 85%. And still, some people are saying, you know, that's not good enough because we're only at 85%, even though there's this big delta. What do you think are sort of the ethical considerations that we're going to have to navigate, especially with something as critical as life and death? And how can we convince people that this step change improvement is worth it?

• 9:57Speaker 01

  Well, there's a couple of things where most people misunderstand, you know, medicine and a bunch of other things. So, the first is, a lot of this serious, when you're kind of like taking a drug about a serious condition or a surgery, it's a percentage game already. It's not zero or 100. You're already playing a percentage game. And even though, like a doctor might say, "Well, I'm actually prescribing this medicine to you because this is the way we understand the biology to work, and this is the kind of the question around, you know, what we're trying to do," it's actually never 100%. It's actually like, we think that this, on a high percentage, actually works in your condition, and sometimes doesn't work for people.

• 10:39Speaker 01

  And maybe once we get to really deep precision medicine and genetics and a whole bunch of other things... My guess is it won't get to 100, it's just improve the percentage. So it's a percentages game. So the fact that people misunderstand that, because they do have exactly, as you say, it's like, well, humans are infallible. It's like, no, that's not the way it works. And by the way, they're very, very good, and they're playing a percentage game on your behalf. So the short answer is, in this case, you, you need to say, no, no, we need to deploy the thing that the numbers, you know, dictate. Now, this gets to the second part of it, which is, part of the progress of how medicine and specific works, is people start trying things and they see if it works.

• 11:20Speaker 01

  And then if it works a percentage of the time, they go, oh, we've got something here. This could possibly be like, like, and then they start frequently, not always, sometimes they figure it out because of causal stuff in advance, but sometimes they figure out the causal stuff afterwards. That after they started going, oh, this works, and by the way, this is how the earliest medicine started working. It's like, oh, aspirin works.

• 11:44Speaker 00

  Tylenol works too?

• 11:45Speaker 01

  Yes, Tylenol works too. That, you know, we all should be carrying around some Tylenol as part of our, you know, we're, we're pro-science. We understand. And so, uh, that, that, that percentages game is the thing that makes it work. And so people need to understand, it's like, oh, my God, you increased the percentages? Give me that. Right? And so give me the AI, even though we go, we don't understand, like, okay, it's making a prediction. We don't understand why it's making that prediction. But if that prediction is all the more accurate, then we'll do it. Now, we don't stop there. We go, of course, you should always try to increase the prediction, of course, but also go, okay, why is that? And then you go, great, we've got this thing. Why is it make it like, why is, and because then, by the way, we, we can begin to see disorderal cases.

• 12:28Speaker 01

  Why is it identifying these people who we weren't identifying before as very important to, to keep in the hospital, treat, treat again, bring in again more soon, et cetera? Why is that? What is the thing it's seeing that we're not seeing? And then that improves our science.

• 12:46Speaker 00

  So to make changes and improvements in any field, but especially in the medical field, you obviously need better technology. You need humans to come on board. You need the clinicians and practitioners to say like, oh, this is working. And then third is you have the legal and regulatory framework. And that's going to be an impediment here too. What do you think are the ways that we can get, uh, you know, sort of the government, legal, all of those things on board? So that we can integrate AI better?

• 13:11Speaker 01

  Well, it's one of the, um, I think there's two parts to how government regulation affects in this area. One part is a set of, you know, very good thoughts around, you know, kind of like, okay, so does it work appropriately? Does it, is it, um, you know, are you avoiding downsides? Are you being, you know, kind of appropriately inclusive and broad-minded across, you know, many different conditions or many different, like, you know, young people, old people, multiple, you know, kind of racial characteristics, a bunch of other things, you know, men, women, um, you know, like, let's cancel a whole bunch of science studies because they have the word women in them. And it's like, that's a great

• 13:54Speaker 01

  idea. Um, that's sarcastic in case anyone misunderstands that. And so, um, and so that's good. The problem also is that we tend to be very influenced in our legal system by trial lawyers. And so the trial lawyers go, no, no, we, we, we should have this to a much higher standard. And we should do it because that way we can try to, you know, impose essentially the trial lawyer, lawyer tax on the whole system. And that one should be much more contained and regulated. And so, you know, like, like not allowed to infect what the regulations are, how it operates as much, not zero even in that case, but much less. And so, um, and so you need to say, well, we should be able to deploy,

• 14:39Speaker 01

  we should experiment, as long as we're not, um, like, we're, we, we can demonstrate good evidence that we're, we, we, like, for example, the classic problem with this is, well, we're going to save 10 new lives, but we're going to lose this other life that we didn't lose. It's like, yeah, but that's, that's, by the way, how medicine works, too. So it's like, okay. And that should not be, like, if you're saving the 10 and doing this in the right way, even if you lost a different life than you would have lost before. That's that's how we progress on medicine.

• 15:09Speaker 00

  Absolutely. I mean, medicine is trial and error, and we're just getting better and better and building upon previous evidence to see how we can make it even better.

• 15:16Speaker 02

  This podcast is supported by Google. Hey, everyone. David here, one of the product leads for Google Gemini. If you dream it and describe it, VO3 and Gemini can help you bring it to life as a video. Now with incredible sound effects, background noise, and even dialogue. Try it with a Google AI Pro Plan or get the highest access with the Ultra Plan. Sign up at Gemini dot Google to get started and show us what you create.

• 15:41Speaker 00

  So recently, Zoom CEO Eric Yuan, he joined many big name tech leaders, Bill Gates, Jensen from Nvidia, Jamie Dimon from JP Morgan Chase, in saying that because of AI and because of the increases in productivity, we might be heading to a three or four day work week. For some people, that sounds like utopia. This is amazing. They can have more free time, more time with their family, whatever it might be. Some people think that this doesn't sound good. So I think there's two parts of it. It's one is, is this gonna happen? And then also, sort of, what are the technical and cultural barriers? And I think the alternative some people are saying is that instead of a reduced work week for everyone, we actually might just have a bifurcated system where there's a lot of unemployment, especially

• 16:26Speaker 00

  entry level roles, and then other folks are doing just fine with AI. What do you think about the three day work week predictions? And like, what are the barriers to getting there?

• 16:36Speaker 01

  So the first is, if you had to choose from a society perspective between a three day work week and a bunch of people out of work, you choose the three day work week. It's a really important kind of part of our social objective to actually make sure that a bunch of people actually, in fact, have employment, have things to do, have, you know, a sense of at least some purpose in the work. I mean, obviously, people say this is a deep purpose. Great. Even better to have deep purpose. But it's like, I feel like I have a role. I have a role in the organization aside. I contribute, etc. I earn my money. And I think that's that's good to have. And I think there's various ways that people are worried about the Cognitive industrial revolution that AI is bringing.

• 17:19Speaker 01

  They go, oh shit, we're gonna have a bunch of unemployment. And by the way, we'll have a lot of job transitions, which will involve unemployment at minimum in the job transitions and so forth. And we need to do things to, to, to, to solve that. And one of the benefits, as you know, is AI is a good tool for that. Whether, you know, the AI can, you know, help you figure out other work you can do. AI can help you upskill, reskill for that. AI can help you do the work. We just wanna make sure we're deploying AI to help these job transitions too. Even as like what the work looks like, what the job looks like. Now, your average person says, I don't want the job change. I'm perfectly, like, I've been doing this for X years. I'm perfectly happy. It's like, I don't want the work. And it's like, jobs change. I understand that, you know, you gotta think about it as like,

• 18:03Speaker 01

  what you were was a horse and buggy driver and the cars are here now. And like, you, you can say, well, but we should all run with horses, horses and buggies. Like, no, actually, in fact, society is much better with cars. That's what we're gonna do. And so like, let's help you adjust, you know, as, as part of it. Now, that being said, the question, I actually don't think that, one is I don't think we're heading, I think we have a lot of job transition. But actually, in fact, I think that in a lot of cases, even though you have massive productivity increase of AI, just like you had massive productivity increases in the industrial age. I don't think that ends up with a systemic, well, that people don't need to work anymore.

• 18:47Speaker 01

  I think that's way further out than most of the critics who are on that side think. And, you know, I'm not, I'm not sure exactly when we might get there. I mean, it might be many lifetimes. Yeah. Right? Not just my lifetime, your lifetime, you know, kind of our children's lifetime, etc. So, so I'm uncertain about that prediction. Now, not, not a hundred percent uncertain. It's one of the things to, to think about and prepare for and whatnot. Now you get to the, okay, well, does that mean on the productivity, we go to three to four, you know, day work weeks? Which, by the way, some societies have done in various ways. The Germans are an obvious example of that. Now, the problem that's brittle with that is that.

• 19:29Speaker 01

  You know, human beings divided into groups and we compete. And we, and part of how we competition is this group's gonna work a lot harder. So one of the things I think we see coming for European auto industry is the Chinese auto industry. And it's also coming, of course, for the American auto industry too. Because if they achieve much higher productivity, and some of that's robotics and all the rest, but they go, well, we're willing to work six days a week.

• 19:57Speaker 00

  Right.

• 19:57Speaker 01

  We're willing to do nine nine six, that's what we're doing. Then their industry can very well wipe out the other industries. And so there's a kind of competitive landscape to this. That's the underlying kind of point of view in this. And we, it's part of the progress of capitalism, everything else. That's part of, like, that competitive landscape is part of what happens. Now, part of the reason why the Germans could do that is they had a very well tuned system. Technology, high quality, like, not just technology of the end product of the cars, but also how they make it, the apprenticeship system, but how they trained, like, deeply skilled, you know, people in doing it.

• 20:37Speaker 01

  I think there's a bunch of organizational techniques that I think are good to learn from in the rest of the world, which is, you know, how do you have, you know, kind of capital and management working alongside labor in transition. And but, but what you have to understand, and this is one of the things that's gonna be very difficult for Germany in the next 10 years, difficult for the US is, more for Germany, I think, in this case, is the competitive, like, your competitive situation is suddenly increased. And all of a sudden then, three day work wakes, four day work wakes, don't work in that competitive situation. And I think this prediction of, well, we have this much work to do, and now we have this much productivity, it's like, well, actually, work always goes up,

• 21:19Speaker 01

  there's always competition, we're always generating productivity. So I don't think we're anywhere close to a three or four day work wake either.

• 21:27Speaker 00

  Okay, so, I mean, economists have got this wrong forever. You know, 100 years ago, they predicted in the US, as people's incomes went up, they would work less. And now in the US, we're in this sort of strange situation where actually, the more you make, the more likely you are to work more. And folks at the lower end are actually fighting to work more. Some of them only are working 24 hours in either a gig economy or in an entry level job. And they want to work more, but their employees actually won't give them that more time. And so they're sort of asking for more hours. And so I know you don't like predicting, especially more than two years because you think it's a fool's errand. But so if I had to say 10 years from now, would you predict that the economy that we're in now looks more like a three or four day work week? Or you don't think that's going to happen?

• 22:12Speaker 01

  I don't think it's going to happen.

• 22:13Speaker 00

  All right, so we are now going to shift from a more focus on tech to a focus on government and institutions. For our listeners who don't know, you launched the Trust in American Institutions Challenge with Lever for Change, which is all about rebuilding trust in important American institutions, our government, hospitals, libraries, the media. Like, how can we ensure that we have a cohesive American electorate, consumers, et cetera? And the good news is that yesterday we announced the five finalists for the challenge. And I'm going to get to them in a second. But I would love for you to talk about Lever for Change and why in particular you wanted to partner with them and their model for doing this.

• 22:54Speaker 01

  So unsurprising to many people who know me, we live in a networked age is my point of view. There's a lot of different things that come from that. Now, that also is how should we do philanthropy? Because when you think about like a serious problem in society and climate and anything else, you should be thinking, how do I have a potential network solution? Right? Because we have network challenges. We should have network solutions. And Lever for Change realizes this because it says, well, actually, in fact, let's layer in multiple layers of networks to help solve problems that otherwise might look at like, ah, is that solvable? You know, you know, really, we need some new ideas on this, a bunch of other things.

• 23:38Speaker 01

  And so it's like, okay, we use a RFP prize process to kind of go out to many, many different, you know, individuals and nonprofits and all say, here is a problem we're trying to solve. Then, because the money is serial, people then go, okay, I'm going to, you know, submit a proposal, request for a proposal. That's an RFP. And then you deploy two other networks, one network of experts who, you know, that Lever for Change has through its, you know, excellent, you know, kind of origins in MacArthur that are experts on all kinds of things. And it's not just the MacArthur genius people, it's many others in terms of how this operates, to kind of look at

• 24:19Speaker 01

  these proposals, you know, kind of comment on them, rank them, score them on different things, say which things, you know, have high probability, which things have low probability, which might be challenges that are not there. Then you have another network of judges that kind of works through it and kind of says, okay, these are the ones that we get to. And you run through this process by which you get to, you know, kind of semifinalists and finalists, and then allocation of money for doing this. And it's a, it's a call from a network of ideas, by the way, just like networks of entrepreneurs in Silicon Valley and all, networks of experts and, and kind of decision makers, which investors and venture capital kind of plays into, now they tie to the capital, then kind of the judging of

• 25:01Speaker 01

  it and making it happen. But it doesn't actually even just end there in terms of how the networks, because then what you've done is you've just built massive network databases, including of interesting kind of individuals and nonprofits with projects. And Lever for Change has a whole bunch of funders who are specifically interested. And you go, well, actually, in fact, this one won this particular challenge, but what this, this nonprofit is doing over here is of exact interest to this funder. So part of the benefit to participating in this network to the nonprofit is Lever for Change puts actually as much energy into routing every good project to any funder or any other kind of thing. They go, oh, this

• 25:42Speaker 01

  one is actually a good match, right, for this. And that's, again, a network property. And so that's why Lever for Change, which has just been doing amazing work.

• 25:51Speaker 00

  You believe in Lever for Change and the model. It's so based on networks, which is what you believe in. And we were talking about what issue should we focus on? We love this model, but what is an important issue? And we chose trust in American institutions. Why was that so critical to you?

• 26:06Speaker 01

  It's, obviously, we live in this very tumultuous time. There's a whole bunch of erosion in trust of all kinds of institutions. And, you know, in the last couple of months, you know, a sustained assault on the trust of the institution of science, among many other things, whether it's universities, corporations, government, etc. And what people don't realize is it's fundamentally, even as institutions have limitations, this is how our society works. Like, the mistake, if you look at any history, is tearing all the institutions down, is the French Revolution, is the Cultural Revolution, is Year Zero in Cambodia. It's, it's, it's a disaster to tear institutions down. You want to renovate them, right?

• 26:50Speaker 01

  And so, and trust is an important part of it. And people say, well, how do I trust that one has this? Like, well, one, you should have a trust in constant renovation anyway. But we should all be working on how do we restore trust? Because that's how we live in a healthy society. This isn't even just talking about democracy. This is talking about a healthy society. Like, why does money work? Because we trust it. Why does banking work? Because we trust it. If you don't do that, we have a very, very serious problem as a society. So, it was partially like, okay, what is this problem that I think many, many people understand? What is countervailing to the general, many people saying, burn it all down, including people who are in government, in terms of how it operates?

• 27:33Speaker 01

  And to say, no, no, what we should be focusing on is, what are the states we, the, the steps we do to renovate and build trust? And so, focusing on that building of trust is something is, like, is again, in the lever of change, is something I want everyone to, to think about how important that is and start acting in that direction.

• 27:53Speaker 00

  And I think importantly, this is a non-partisan initiative. You decided on this challenge in this area long before the presidential election. So now, no matter who's in charge, what party, trust in American institutions is critical. And so, as I said just yesterday, we announced the five finalists. And again, the great thing about Lover for Change is that each of these five finalists will get $200,000 to create their project, plan for the future, make a better plan. And then, one of them will receive $9 million, which we're very excited about. But to your point, all five of them are going in the Lover for Change network. And so, regardless of who wins, we're excited for all five of them to get the spotlight. They're all very worthy projects. So, I'm going to go through each of the projects. I would love for you, super briefly, to tell us why that is important.

• 28:37Speaker 00

  And just for those who are watching and listening, we are not putting thumbs on the scale. All five of these are amazing, and we just want everyone to know about them.

• 28:44Speaker 01

  And there's a judging panel. This is not, Ari and I are not the judging panel.

• 28:48Speaker 00

  All right. So, the first finalist is the American Journalism Project.

• 28:53Speaker 01

  So, local journalism. Part of when you say, well, there's obviously a whole bunch of aspersions around journalism on a national level. I think, actually, much of that is misplaced. But, local journalism, also super important. You can see it, it's in your life, it's what happens, and it's one of the ways that you can begin to restore trust in, well, what is the reporting and investigative approach for things that matter in my life?

• 29:18Speaker 00

  Absolutely. And I love that, like, if you are in a rural town in Tennessee, The New York Times might not be covering stuff that's relevant to you, but you want to know what's happening in your town, and that can build that local community trust and fabric, which is so important. All right. The second finalist is Cal Matters.

• 29:37Speaker 01

  So, Cal Matters has taken this great idea of saying, hey, look, we've got these government institutions, which too often run bureaucratically, slowly, opaquely, et cetera, and trust gets lost in government because of that. And by the way, we want it to be more transparent. We want to shine a spotlight. So, like, let's do investigations, let's have dialogue, let's, let's, let's, let's state clearly what's working, what's not working. And it's not just a what's not working thing, but, but to have that kind of, of, of, of visibility, which then creates increased accountability. And that's one of the things where people say, ah, you are actually providing me services, and actually, you're providing me services even better than you were before.

• 30:20Speaker 00

  Yep. Third finalist is Recidiviz.

• 30:23Speaker 01

  For Recidivis, the area is like, how do we make much better, we apply data science to like kind of parole and other kinds of things. And you say, oh, that's just, is that just something for, you know, the, the people who've been incarcerated? No, it's for them too. But what people don't realize is if you do this well, it works much better for society. You get people out, you're not paying the public bill, the money that comes out of your pocket to pay for them to be in prison, etc. You want to factor, if you, if you could do effective paroling decisions, you want it because it's good for you, not just good for the person. And then integrating these people in communities and all else. But it's basically data science, which is our modern thing, applied to much better decision making.

• 31:08Speaker 00

  I think if most Americans knew that there was 70,000 people sitting in prisons and jails who have paid their debt to society, who should be out, and we're paying money for them to stay in jail just because some paperwork didn't go to the right place, they'd be very in favor of getting people out of prisons and jail and back into their communities. So the fourth finalist is Results for America.

• 31:29Speaker 01

  So one of the key things is, like, that really works in business is we share information, right? We, we, we say, hey, what's the best way to do this? This is actually one of the things that makes Silicon Valley work. It's an intensely learning network. So Results for America is to say, hey, let's have local governments that are doing things, experimenting things, trying things, and share them with other local governments to say, hey, here's a really good way of solving this problem. It might be, you know, sanitation or trash on the streets. It might be, you know, traffic, it might be, you know, kind of policies in schools, it might be zoning, anything. Here's things that worked for us, might work for you. And of course, sharing that kind of information means that it's a very cheap way of increasing quality of local government, quality of services.

• 32:13Speaker 00

  If we've already solved the problem, let's learn from that and solve it in our own local community. The final finalist is Transcend, working in public education.

• 32:22Speaker 01

  So the basic idea in Transcend, which is of course a very good thing, is bringing the communities much better in to try to be partnering and working with how do we have better schools? How do we have better outcomes from the schools? What are the kinds of things we might experiment on? Oh, there's this piece of information over here. Have we tried this? Maybe we should try this. Because then that sense of co-ownership is actually one of the things that is part of how you get trust in institutions, part of how you go, oh, well, since I have some participation in voice, I can help make it better. And I think that's one of the things that Transcend can do so well.

• 32:58Speaker 00

  Thank you so much for that information about the five organizations. I think what's critical here is all five of these organizations are worthy of trust and worthy of support. And so, please, everyone who's listening, go check them out. We want as much support as possible for all five organizations. And then we will be announcing a winner in the spring. Thank you so much, Reid.

• 33:17Speaker 01

  Always a pleasure, and New York!

• 33:20Speaker 03

  This podcast is supported by Google. Hi folks, Paige Bailey here from the Google DeepMind DevRel team. For our developers out there, we know there's a constant trade-off between model intelligence, speed, and cost. Gemini 2.5 Flash aims right at that challenge. It's got the speed you expect from Flash, but with upgraded reasoning power. And crucially, we've added controls, like setting thinking budgets, so you can decide how much reasoning to apply, optimizing for latency and costs. So try out Gemini 2.5 Flash at ai studio dot google dot com and let us know what you built.