Faster, Please! · Economics & Policy
TIER 4 Sat, 2 May 2026 12:16:49 +0000
Also: Key Up Wing and Down Wing news from the previous two weeks that were
͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏ ͏
| |
---|---|---
| | | Forwarded this email? Subscribe here for more
---
_This is a free edition of_ _**Faster, Please!**_ ,_my regular newsletter about creating a better America and world by accelerating scientific discovery, technological progress, and commercial innovation. (And creating a pro-progress culture.) If you enjoy it and find it helpful in any way, please consider buying a subscription to the twice-weekly regular issues that include in-depth essays, Q &As with smart people, and summaries of relevant news stories. _
Upgrade to paid
* * *
# 🚀 FP! Week In Review, Briefly #31
### Also: Key Up Wing and Down Wing news from the previous two weeks that were
May 2
---
---
| | |
---
| |
---
| |
---
| |
---
| | READ IN APP
---
| |
---|---|---
## **In Case You Missed It ...**
☢️ **40 years on, Chernobyl reconsidered** (Monday)
✨ **The future of work in an age of AI: My chat (+transcript) with digital tech economist Daniel Rock** (Tuesday)
Faster, Please! is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
Upgrade to paid
🚀 **The high frontiersman: Gerard K. O 'Neill** (Thursday)
* * *
## **⤵ Up Wing/Down Wing**
A selection of pro-progress and anti-progress news items from the past week.
### **⤴ Up Wing Things**
* Sectors Embracing AI Are Seeing a Surge in New Business Formation \- The Daily Spark
* AI Sends Its Best Wishes to the US Economy \- Bberg Opinion
* A.I. Spending Sets a Record, With No End in Sight \- NYT
* Behind the A.I. Boom, a Boring Business Is Soaring With Better Ads \- NYT
* Three reasons why DeepSeek's new model matters \- MIT
* DeepMind's David Silver just raised $1.1B to build an AI that learns without human data \- TechCrunch
* BMO Turns to AI and Quantum Computing to Predict Earthquakes \- Bberg
* The Podcast Where You Can Eavesdrop on the A.I. Elite \- NYT
* The Filmmaker Using AI to Beat Hollywood \- TFP
* How bots could help revive democracy \- FT
* Chip Startup Aims to Shatter AI's Dreaded Memory Wall \- WSJ
* Meta wants to power data centers from space \- Axios
* See How the Robotaxi Industry Is Taking Off Across the U.S. \- WSJ
* Welcome our new robotaxi overlords \- FT
* How AI is powering the next generation of robotaxis \- FT
* Start with the sensors, then design the rest: How Zoox built its robotaxi \- Ars Technica
* I've Seen the Future of Electric Vehicles, and Gen Z Will Love It \- Heatmap
* Air Taxis to Fly Between JFK Airport and Manhattan for 10 Days \- Bberg
* I've Covered Robots for Years. This One Is Different \- Wired
* I've Covered Robots for Years. This One Is Different \- Wired
* Humanoid robots start sorting luggage in Tokyo airport test amid labor shortage \- Ars Technica
* Quantum Computing Companies Are in a Race to Go Public \- WSJ
* This company says nuclear fusion could finally power the grid -- and soon \- CNN
* How to Build a Better Kind of Nuclear Power? This Side Hustle Might Help. \- NYT
* US research reactor first to produce electricity \- WNN
* Chernobyl Wasn't a Nuclear Disaster--It Was a Communist Disaster \- WNN
* A Radical New Engine Shows Why Internal Combustion Still Matters \- WSJ
* Blue Origin certainly has ambitious launch targets for New Glenn \- Ars Technica
* The hunt for a new Earth \- FT
* Commercial space station developers make their business case to NASA \- The Space Review
* Craig Venter raced to decode the human genome \- The Economist
* Exclusive: Zuckerberg-backed Biohub bets $500M on AI biology \- Axios
* Could At-Home Brain Stimulation Reduce Psychiatry's Reliance on S.S.R.I.s? \- NYT
* * *
### **⤵ Down Wing Things**
* AI Productivity Growth Won't Match the Computer Revolution \- PS
* San Francisco, AI capital of the world, is an economic laggard \- The Economist
* Silicon Valley Is Bracing for a Permanent Underclass \- NYT Opinion
* Most Americans Think the Job Market Will Get Worse. Here's Why That Matters. \- WSJ
* The great American data centre divide \- FT
* Inside Josh Shapiro's Attempt to Navigate the Data Center Backlash \- Heatmap
* AI is confronting a supply-chain crunch \- The Economist
* Washington has a new Anthropic problem \- Axios
* Taxing Artificial Intelligence Would Be a Big Mistake \- Bberg Opinion
* Chatbot horror stories are inspiring an unhelpful jumble of fixes \- Wapo Opinion
* Pressure over AI regulations mounts for Florida lawmakers \- Politico
* What Happens if Trump Seizes AI Companies \- The Atlantic
* From Indiana to Idaho, a Backlash Against A.I. Gathers Momentum \- NYT
* AI and the danger of cognitive surrender \- The Economist
* AI's biggest critic has lost the plot \- The Argument
* A.I. Bots Told Scientists How to Make Biological Weapons \- NYT
* It's the Age of Electricity and America Isn't Ready \- NYT Opinion
* Senate Dems tussle with Burgum over permitting \- E&E
* Nuclear AI Startup Fermi Promised Land and Ample Power. But It Couldn't Sign a Single Client
* A Trump-branded nuclear power project thrilled investors. Then came the crash. \- Wapo
* King Charles, America and the futility of growth \- FT
* A Crude Awakening on Inflation Is Hitting Markets \- Bberg Opinion
* It's never a good idea to sack the entire National Science Board \- FT
* National Science Board eviscerated; Trump admin fires all 22 members \- Ars Technica
* The U.S. Could Lose the Space Race to China \- WSJ Opinion
* The tortoise and the hare: will China beat the US in the race back to the moon? \- The Guardian
* Gateway manufacturer finally acknowledges issue, fails to mention "corrosion" \- Ars Technica
* When the orbital layer is the kill chain \- The Space Review
* * *
### **↕ ️ Which Wing Things?**
* AI Investment Boosted Economic Growth, While Consumers Tapped the Brakes \- WSJ
* AI May Be the US Economy's Only Hope \- Bberg Opinion
* What Silicon Valley layoffs hide about the future of the job market \- Wapo
* Big Tech's earnings get ever bigger, and ever less useful \- FT
* Big Tech Strikes Gold With AI, but at a Steep Cost \- WSJ
* Sam Altman's Next High-Wire Act: Getting OpenAI to Make More Money \- NYT
* China Orders the Unwinding of Meta's Acquisition of an A.I. Start-Up \- NYT
* How OpenAI's $500bn data centre venture Stargate has shifted shape \- FT
* How Much More Power Can the U.S. Grid Provide for AI? Projections and Policy Implications for 2030 \- RAND
* Will America Finally Let Itself Build Nuclear Plants? \- TFP
* Why DeepSeek's sequel failed to impress \- The Economist
* OpenAI Codex system prompt includes explicit directive to "never talk about goblins" \- Ars Technica
* Exclusive: AI use booms in states, with mixed results \- Axios
* Beware of Government by AI \- PS
* A major new study found AI outperformed doctors in ER diagnosis -- but there's a catch \- Vox
* Is TikTok art? \- The Argument
* Put it in pencil: NASA's Artemis III mission will launch no earlier than late 2027 \- Ars Technica
* How Elon Musk Used SpaceX to Benefit Himself and His Businesses \- NYT
* This is who's developing Golden Dome's orbital interceptors--if they're ever built \- Ars Technica
* There's a lot of hype about Chinese EVs--is any of it true? \- Ars Technica
* Is Ibogaine the Next Miracle Drug? \- TFP
Upgrade to paid
* * *
## **Essays and Q &As**
**☢ ️ 40 years on, Chernobyl reconsidered**
| |
---|---|---
**A Nuclear Error:** The 2019 HBO miniseries _Chernobyl_ dramatizes the infamous nuclear event, which just recognized its 40th anniversary. The feared second explosion at the nuclear plant (a major source of tension depicted in the series) could have devastated much of Eastern Europe. Instead, that scenario was avoided, and the immediate death toll remained relatively limited. The aftermath, however, is still felt around the world--and not for the reasons one might assume.
**System Failure:** Despite being the worst nuclear disaster to date, Chernobyl caused far fewer direct deaths than commonly assumed based on media depiction. The deeper failure lay less in nuclear technology itself than in the Soviet system's secrecy and poor safety culture. From the newsletter: "'Nuclear power wasn't the problem in Chernobyl,' journalist Ronald Bailey of Reason persuasively writes in a new piece. 'The problem was communism.'"
**The Fallout:** The most significant damage came after the TV-worthy events. Chernobyl solidified public opposition to nuclear power around the world, slowing plant construction and boosting fossil fuel use. The unintended result was considerably more deaths due to air pollution from energy sources erroneously perceived as safer. The authors of a 2024 study conclude that the global lost years of life following this energy pivot totals 318 million.
**Up Wing Up Shot:** Similar knee-jerk anti-nuclear reactions followed both the Three Mile Island accident and Fukushima disasters. In each case, policy responses--higher electricity prices, dirtier air, and the inherent risks of mass-evacuation--proved deadlier than the original events.
Upgrade to paid
* * *
## **Transcript**
**▶ ️ My chat (+transcript) with digital tech economist Daniel Rock**
(A _lightly edited transcript of ourpodcast conversation.)_
**In This Episode**
* **The trouble with forecasting (1:40)**
* **The economist 's evaluation (8:09)**
* **The productivity J-curve (11:49)**
* **Exposure vs. automation (18:53)**
* **Growth projection (23:04)**
### **The trouble with forecasting (1:40)**
. . . I have a tough time seeing a place where all sorts of knowledge work changes all at once and we don't take a while to figure out the best new configurations--or, as we're figuring them out, there isn't a bunch of churn where people get hired to do new types of work.
**Pethokoukis: Just the other day, I was reading the Wall Street Journal and there was an interview with the new CEO of Verizon. And in that interview, he predicted 20-30 percent unemployment within the next two to five years due to both AI causing a white-collar bloodbath, but also humanoid robots upending manual labor jobs.**
**Now, I 've heard a couple of these kinds of forecasts from AI CEOs. This is maybe the first one from a tech CEO who's not necessarily an AI CEO, but there are other ones. I believe it was the head of Microsoft's AI unit said that within 12 to 18 months, every sort of behind-the-computer-screen job will be automated. What do you make of those kinds of forecasts?**
Rock: I find those kinds of forecasts highly risky. I'm not going to say I don't think it's possible, something could happen that would make these things possible, but I think our economy is ingesting and responding to AI at a fairly gradual clip right now, even what we have right now. So setting aside that I think there are a number of advances and important research results that'll be necessary to make humanoid robots happen to that extent, I think even if they existed, if you had this enormous explosion and capabilities, it would still take a while for organizations to integrate them properly.
So do I think it's possible longer run? Yeah, potentially. But within 18 months or two to five years, that seems a bit aggressive to me. Though, I think in this world, you kind of have to have wide confidence bands as things change, right?
**When I hear these kinds of forecasts from people in the tech sector, I sometimes wonder if they mean that this is what the technology could do in the lab, or some artificial scenario. The CEO of Verizon seems to be talking about the real world, though. And even if you 're a super-optimist, it seems that, given the real-world constraints--be they legal, organizational, or what have you--even the most powerful technologies getting to those unemployment numbers seems science-fictional.**
Yeah, in a sense, I agree. And then I think to get to the point where you have unemployment and people aren't getting redeployed, while you have these remarkable technologies shifting and getting deployed at the same time, that strikes me as a little bit hard to imagine that scenario as being realistic.
The keyword you said there, the organizational component, that's the really big part: How are firms doing trainings? How are they thinking about incorporating these tools into their workflows? I talk with a number of executives who are really taking these kinds of questions super seriously, and it's going to require a ton of buy-in from different types of people. Not everybody's stoked to use the technologies. The folks who are using them aggressively, they run into guardrails, they run into places where organizational process isn't quite set up for the speed at which they're building.
So there's still a lot to get figured out. If you want to do productive activity, it's a little bit of that stone soup story and you have to configure a lot of different components all at once.
I'm not the only one who's been thinking and writing about a bunch of this. There's a long history of the economics of technology where researchers have been considering these ideas for a long time. Part of my job as a junior researcher in the field right now is going back to stuff that was written in the late '80s and '90s and being like, "Hey, these people are still right." So I'm digging up those papers and reformatting them, in a sense, for the AI age.
**What I see, and you 're on social media as well, and we follow each other, but I get a sense there are some--I think this is certainly the case among technology people, but even there are some economists--who think that exactly what you're doing, which is looking at the history of technology, that that has created a blind spot to this powerful new technology, and it's really kind of a "this time is really different" argument. Are you concerned that you have a blind spot and are you aware of that and you're trying to make sure that you don't?**
Yeah, definitely. I think the Silicon Valley adage here is "strong opinions weakly held," where you really let evidence and what's happening guide your thinking. My friend, Kevin Bryan, who's a professor at University of Toronto, he pointed out the other day on social media that the burden of proof for people to say "this time is different" really falls on them. You can look at the new technology and say, "Yes, this is different." As Erik Brynjolfsson would say, it's the most G of all GPTs: It's the most general of all general purpose technologies. And you could very much make a case that if not all knowledge workers are going to have to change what they're doing, then this would actually create quite a profound impact on the economy.
At the same time, maybe it's a lack of imagination on my part, but I have a tough time seeing a place where all sorts of knowledge work changes all at once and we don't take a while to figure out the best new configurations--or, as we're figuring them out, there isn't a bunch of churn where people get hired to do new types of work. Some of it will look like it's automation, some of it will look like it's transformation, and it's all going to get figured out. I think that works at human-time skills, not necessarily machine-time skills. So you really have to buy that the machines are going to be responsible for much, much more. AI systems can do much, much more in verifiable, high-quality ways to think that people aren't going to be part of the loop in some way, shape, or form. I think we'll find those bottlenecks.
In a sense, I would love it if it were the case that machines could do way more. Society's got a redistribution problem there and we're very wealthy at that point. Yeah, there's power dynamics and other things--you could imagine not quite sci-fi, but deep-change scenarios where we have really important social problems to sort out, but all sorts of amazing things are here, and I think more likely, my modal expectation over the next few years is this does look like the kind of normal technology historical precedent, even if there are winners and losers from that kind of scenario.
### **The economist 's evaluation (8:09)**
The fact that I see a lot of large company executives stumped by this at a big-picture level makes me think that, yeah, indeed it will take a little while to sort out . . .
**Having spent many years and most of my career in journalism, I understand that there 's a certain bias toward novelty. You're looking for what's new, you're looking for the new take, and I see this in economics coverage all the time: Someone has a new theory and it changes everything, and rarely does it actually change everything. And I was kind of thinking that when you were recently quoted in a New York Times piece, and the headline of that piece was, "Economists Once Dismissed the A.I. Job Threat, but Not Anymore." I don't think that's right. I don't think economists have ever dismissed the AI job threat. I think what they dismissed were these very huge, very quick kinds of scenarios, and I'm not sure, generally, among economists that's changed.**
Yeah, I don't think it's changed a ton, though I'll say, given the narrative and some of the media, maybe that's changed a little bit. Ben Casselman did a great job researching that piece, really talking to a bunch of people.
**It 's a good piece. He doesn't write the headlines.**
Yeah, that's true too. What I think he was finding was a little bit less certainty amongst economists, especially the group that he spoke to. I think we have a sense that we should pay attention to the technological capabilities and be expansive in terms of how we think about what they can do, listen to the AI researchers who say, "Hey, maybe this can automate a bunch of work."
I still think, at the task level for a bunch of it, what are the processes? What are the tasks people are doing? What are the new systems that you build up and how do you reconfigure them? That's hard managerial work. The fact that I see a lot of large company executives stumped by this at a big-picture level makes me think that, yeah, indeed it will take a little while to sort out, but that said, the new systems might be way more efficient, and for a short-run period, you could have reconfiguration of work that is potentially job-destroying if you have inelastic demand.
So it's all a question of what is the speed? You have to mix a few things in your head. What's the speed of capabilities advance? What's the speed of organizational structure reconfiguration? And then how big is the pie?
I am paying a little bit more attention to scientific teams now. A couple of reasons is, when you use AI in science, the returns are huge, and that's very, very exciting. But another is, scientists want to tell you how they did it and how they're doing things. So there's a little bit of what economists would call incentive compatibility there, where they want to say, "Okay, yeah, here's how I transform my workflow using AI." And if we can get some insight around how they're redesigning their workflows, maybe that can be extended to larger companies where you've got more invasive control systems or set structures.
One of the things I think people find as they start trying to do digital transformation is they do an inventory of their existing workflows and how they do stuff. And nothing to do with AI, they look at some of this and they say, "Well, that doesn't make any sense. Maybe we should do this differently." The worst thing you ever want to hear is, "We do this because that's how it's always been done," not because it makes a ton of sense. And then they go and change that, and they get some productivity gains just by stripping out some of the stuff that they don't need to do anymore. So if you are finding that that's what's happening in your work, it may be a good sign because then you're on your way to the new way, with new capabilities.
### **The productivity J-curve (11:49)**
People are not doubting that AI is valuable, they're just trying to figure out how to use it.
**Anybody who has written or been a key researcher on the notion of a productivity J-curve, there 's no way I'm not going to ask them: Where are we on that productivity J-curve? Where are we? And how do you figure out where we are?**
My best guess is that we're very, very early-stage on this.
**First, explain what that is. Explain what the J-curve is, probably quickly and more concisely than I could.**
So the J-curve is kind of the simple finance dynamic that we've attached to productivity measurement. Total factor productivity is, in effect, the magic beans of the economy. It's the growth that we can't explain by growth in measurable inputs. It's also referred to as a solo residual, and, commonly said, it's a measure of our ignorance more than anything else. When I teach my students how you compute this, they're like, "Wait, what? That's what productivity means?" I mean, labor productivity is a lot more intuitive, like GDP per hour worked, or per worker, or something. It doesn't correct for capital usage, but at least that's a little bit more directly interpretable.
But what we did, my co-authors Erik Brynjolfsson, Chad Syverson, and I, we bolted on an investment dynamic where in the beginning of a sufficiently transformative technology where you have to do this kind of work of creating intangible capital, new ways of doing things, training folks, installing software, configuring software, getting the security protocols in place, all that kind of fun stuff--in the beginning, it looks like you're putting a lot of real stuff in, real money, real investment, and you're not getting as much out because you're spending to create these intangible assets like the ones I mentioned, or culture, and so on. So effectively, we get less for more, and that's a problem. Everyone worries about a drag on productivity growth. It will potentially look like a drag on productivity growth.
In practice, when you go and measure productivity growth, you might see 1.5, two percent TFP growth, and that could be the counterfactual, where we're trying to build all these intangible assets, and were we actually using all of our money to just generate straight measurable stuff, we'd actually be up closer to 2.5, three percent. So in the beginning when we get this dynamic, we create the intangible asset.
But then, now that we've got this capital asset that's hidden on the balance sheet, it's not in your income statement, eventually you start using that in production, and that investment starts to pay off, but you're not measuring the input. So everyone says, "Hey, we're getting free money here. We're getting way more from less as an input and productivity is booming." So this dynamic where, in the initial stages you dip in, you make a huge investment and that investment return, or the asset you build, is not characterized anywhere in measurable accounts, but then later, it starts to trickle out "free money," ("free money" in air quotes here). That's the J-curve. So "where are we?" is the question you asked.
**Can I ask you just one question about the J-curve?**
Of course, yeah.
**We look at these big aggregate productivity numbers when the government releases them and we 're like, "Is this it?" If there's any kind of blip up we're like, "Is that AI?" Would we actually expect to see productivity increase from wherever it is, or would we actually almost expect to see a **_**dip**_**before it goes up?**
I would be relatively surprised to see a big dip in measured productivity, more like anemic productivity growth in the presence of radically transformative stuff is what my base case would be. I think also these series, because they're such a measure of our ignorance, they're super noisy. It's going to be really hard to tell from the quarterly productivity numbers, are we in a boom or anything right now.
**I believe the good ones, I disbelieve the bad ones.**
Yeah, let's just be optimistic about it. We're going to need a do-over on Q3. But yeah, I think we might be seeing some of the gains from deep learning, or some of the technologies that were being developed five, six years ago, in the initial stages of this, some retrofitting that's happening where we take existing processes and drop AI in.
Now, do I think we're going to see a boom in productivity? Potentially, but I haven't seen it yet in the data. You kind of scrutinize the little squiggles in the graph and you're like, I really hope that's it.
**So we are not ascending, we are not in the ascent of the J-curve.**
Not yet on the J-curve, I wouldn't say so. I think we're very much in the early stages where a lot of the investment, honestly, is measurable. It's in chips, and data centers, and things like that. I don't think companies have really done the reconfiguration work yet. If I had to guess, it would probably be in low billions of dollars spent in the economy, which is potentially enough to move the needle importantly in some areas, but not enough to show up in aggregate statistics, and I think it'll start moving a lot more in the coming years. People are not doubting that AI is valuable, they're just trying to figure out how to use it.
**Yeah, I think, generally, we 're sort of past that point. And if you're looking for the green shoots and you don't want to wait until there's a final government statistic, I mean, where do you look? Do you look at company earnings calls and how are they talking about it? It seems like it'd be real detective work.**
It is. I cheat a little bit. So Erik and I, and then Andy McAfee and James Milan, the four of us co-founded a company called Workhelix where companies will give us the data of what's going on inside of their AI systems.
One interesting thing we see is a power lot every single time. It doesn't matter the tool, it doesn't matter the company. The numbers change a little bit, but it might be 10 percent of your user, or half the value that's getting generated. And when you see that, you think, "Okay, well, there are some people who are really running ahead and doing a great job, but that's a ton of low-hanging fruit to train people to do all sorts of new stuff."
And as long as that's persistent, and you haven't quite worked out how to spread these best practices, or constrain them in some way--because there's a lot of people that don't want to use it, but they don't want to use it because they really want to do a good job at work. They don't want to fail. They don't want to create headaches for other people. Those people are gold. You can have them develop all sorts of constraints and things that make these systems work better. So it's really about building a consensus inside of organizations, and we haven't seen a ton of that, and they're going to get there. There are a lot of top-down efforts that I know of where it's like, how can we celebrate the wins and also make people responsible for doing a good job?
### **Exposure vs. automation (18:53)**
. . . if you see that your job is highly exposed to AI, exposure does not mean automation. Exposure means change, and it might be the case that high levels of exposure are actually great for you.
**At least so far, the general attitude among companies seems to be that when a pilot project doesn 't produce a good result, the response isn't to blame the technology, it's to blame the implementation. They're not giving up on the technology, they're just redoubling their effort to figure out how to make it work.**
Definitely, and that's good, that's healthy. You want to learn what works and what doesn't. And if 90 percent of the stuff doesn't work out and 10 percent is a smashing success, great. Just make sure you're keeping track of it. Make sure you know how to double down on the things that work really well.
**What would you tell someone who either reads quotes about 30 percent unemployment, or they look at their job and there 's some index that says their job is deeply exposed to AI? What should they think? Should they be like, "I'm going to be out of a job soon. I'd better go learn how to do something else. I'd better make no large purchases"? What should they think?**
Well, I would have a different reply to the person who is reading about large unemployment versus the person who reads that their job is exposed to AI. The 30 percent unemployment or, like "Hey, the jobs apocalypse is coming," I would just dig into some of the research on what would be required for that to happen. Alex Imas--a friend of mine, and we're working on a project together, which is really cool--he has a great blog called _Ghosts of Electricity_. He's exploring a lot of these different scenarios. There are other papers that show you what you need to get to outcomes like that. There's a paper, Benzell, LaGarda, Kotlikoff and Sachs, _Robots Are Us_ , which thinks about an overlapping generations model. We don't have to get into all the technical details here, but basically a robot is a perfect replica for a person and all their work they can do.
And then you can think like, "All right, the technology I'm seeing, that I'm using on a day-to-day basis, does that line up with what I think someone could do in their job? Is it something that could be fully automated?" There's a lot of interstitial stuff that people do to make things work. If you have folks, I think the Marxists call it "radical compliance" as a protest strategy: You do precisely what your job entails and nothing more, and things break when that happens. So people are kind of smoothers and they figure out how to work with each other and to work within systems, and that stuff shouldn't be underrated.
Now, since I've done some of this research on exposure, if you see that your job is highly exposed to AI, exposure does not mean automation. Exposure means change, and it might be the case that high levels of exposure are actually great for you. It could be that it's not good. I would say, be concerned if you are a customer service agent and it's remote. That kind of stuff tends to be highly visible to machines. There are clear output measures that are good and bad. I've already seen companies starting to automate aspects of that.
On the other hand, the famous proclamations that radiologists were going away, they have some really interesting mixtures of tasks. That is a highly expert group of people, and if you think of them as data-enabled doctors, the idea that they would go away starts to seem a little bit more and more foreign.
So Luis Garicano and his co-authors, they just had a paper come out on weak versus strong task-bundling. I think that's a good model for thinking about what your job might do. And if it's a strong bundle of tasks, it's a bunch of things that are really inextricable from each other, and using AI in one doesn't mean that you don't still need to be there, then you're in good shape. But if it's stuff where a bunch can be separated out, now you're thinking about what's left for you. If it's something that you're going to spend a ton of time on yourself, then great, and it's probably complimentary to what the machines are doing. If it's something where machines can handle the whole bundle, then we're looking at a different story. OpenAI just had a report come out, too, where they're starting to handle some of this. Shout out to Alex and Ronnie, great work there.
### **Growth projection (23:04)**
My optimistic case right now for the next few years, I would love to see somewhere in the neighborhood of two to three percent productivity growth.
**There was an interesting survey put out by the Forecasting Research Institute, and they interviewed technologists, they interviewed economists, and what was interesting is that, for all the talk that these groups see things differently, their general median forecasts really weren 't that different, their forecasts about growth. You're aware of it, so I'll cut to the chase. Since you view things academically and you're seeing what companies are actually doing, what do you think is the realistic upside of AI boosting economic growth, boosting productivity growth?**
I think it's going to be fairly large, but as we know, since we're looking at growth, you can grow very quickly. Maybe we're lucky if we get three to five percent growth in a short period of time, but then your base is getting better and it's hard to maintain the top-level growth rates.
My optimistic case right now for the next few years, I would love to see somewhere in the neighborhood of two to three percent productivity growth. I think that'd be cool. Three to four is potentially reasonable. I don't know, maybe we do get a miraculous technology and it goes faster, but when I start to see people thinking it's going to be a little bit higher than the median or even the 90th percentile respondents to those surveys where people say it's 15, 20 percent. Now you're talking about a sci-fi economy that's hard for me to conceptualize properly.
That's really radical, transformative, AI-driven change in the economy. I'm kind of curious. I know there's that curse, "may we live in interesting times." That would be a very interesting time to live through. I'm kind of curious to see what it would be like.
**There 's sort of a flip side to that question--not like, "What's the upside of this technology," but what is **_**too much**_**of an upside. Is there productivity growth? I often act like there 's no downsides, but could an economy grow **_**too fast**_**and that it would be so disruptive to society that you 'd get pushback and it would just be too much?**
Yeah, I think there's a lot of folks who like the way things are, generally, and are okay with marginal growth. There are people who are willing to break things to see what would happen in a better situation. I think you start to get into the realm of politics, and sociology, and things that I don't understand nearly as well as folks in that area for what that looks like when society changes that quickly.
You asked about blind spots earlier, I don't think [economists] really have a good sense of what this looks like for a developed, wealthy economy like the US to grow that quickly. So I don't know, I'd be fascinated to see what happens--and I hope for the best thing, I hope for an explosion in high quality drugs for things like cancer. I saw that news about the pancreatic cancer trials with mRNA vaccines, and I was like, "This is the best, we need more of this, please--_Faster, please!_ " as someone might say.
**Thank you. I appreciate that.**
So yeah, I get excited about those potential things, but . . . we rely on economic theory a little bit more, like where's the bottleneck going to show up? What sorts of things are going to slow things down?
**If we have this conversation a year from now, what do you think will be the state of play? What will we be saying? Will we still be saying, "Well, we're still kind of prowling through the business earnings calls and the micro data," or will things be a little clearer?**
All right, so you're asking me to make predictions in a very clever way.
**With all cautions, and caveats, and . . .**
The cop-out answer that I have fairly high confidence in is I think my confidence bands are going to be wider in a year, so in a Bayesian way, I suppose I should be updating backwards to say my confidence ban should be wider now for all the new information that gets released. But if I had to make a concrete prediction, we went from coding assistance about a year ago to agentic coding being mind-blowingly good, and now people are using it to build all sorts of things.
I would say let's take seriously the idea that anybody can build software for anything and that it'll be better and better and more capability-enhancing; in which case, I think a thing people are going to be discussing is how do I manage agentic teams well and how do I integrate people into those processes because they need to.
And then we might have some conversations of what it's like to be managed by an AI system. So as organizations blow up, I think there's been a lot of discussion of what does it look like for a one-person unicorn to show up. I think, okay, if that's possible, then the scope and scale of an organization with AI-enabled reduction of coordination costs and increased modularity and design, I think the so-called boundary of firms can expand quite quickly, at which point you might have an existing large firm that figures it out become much, much larger as a result. It's just we can sustain a bigger thing. So if that's the case, though, you're going to have machines managing folks, that would be my guess. So that might be an early conversation in a year.
* * *
🚀 **The high frontiersman: Gerard K. O'Neill**
| |
---|---|---
**Space to Imagine:** In the 1970s, Princeton physicist Gerard K. O'Neill imagined up massive space habitats: rotating cylinders miles long that could house millions, simulate gravity, and run on limitless solar power. Detailed in his 1976 book, _The High Frontier_ , he argued humanity could begin building these habitats imminently, using lunar and asteroid materials.
**Depicting the Dream:** In his 1981 book, _2081_ , he goes beyond engineering to paint a picture of everyday life in an extraterrestrial future. O'Neill envisioned a very attainable world of routine space travel, AI-managed homes, shorter workweeks, and cities on Earth enhanced by advanced infrastructure. His aspirations feel familiar, comfortable and livable--a stark contrast from much of today's dystopian futuristic storytelling.
**Overcoming Pessimism:** Still, O'Neill was a man of his time. He wrote in the shadow of popular, fear-inducing narratives of the 1970s like _The Limits to Growth_ and _The Population Bomb_ --both of which pointed to Earth's dwindling capacity to support all of humanity. O'Neill accepted these ideas as fact, but his proposed remedy was nothing short of inspiring.
**Up Wing Up Shot:** To O'Neill, concerns about scarcity and overpopulation should be met not with strictly capping growth, but with space-ward expansion. Move heavy industry off Earth, tap into space resources, and allow for continued human development. Scarcity, in his view, was merely a constraint of geography, not physics.
_*This issue written and edited by Julia Cataneo. Organized by your friendly neighborhood AI._
* * *
**On sale everywhere** _**The Conservative Futurist: How To Create the Sci-Fi World We Were Promised**_
| |
---|---|---
Faster, Please! is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.
Upgrade to paid
You're currently a free subscriber to Faster, Please!. For the full experience, upgrade your subscription.
Upgrade to paid
---
| | | Like
---
| | Comment
---
| | Restack
---
(C) 2026 James Pethokoukis
548 Market Street PMB 72296, San Francisco, CA 94104
Unsubscribe