The AI Would Like to Speak with Your Manager

And more from this week, and last…

Before we jump in, a quick announcement. American Compass is hiring! We’re looking for an:

• Editor, to lead the next phase of rapid growth at Commonplace, and an

• Events Manager, to lead the dozens of events we host each year.

Please get in touch if you, or someone you know, might be interested.

---

As for the week that was—two weeks, really, because July 4 had no roundup—there’s been just too much going on. I had nearly 50 articles clipped, and we can’t be doing that, so I’ve slimmed it down to about a dozen. Let’s jump right in, not with your one thing to read, but with a play in three acts.

Act I: The Visionary CEO Spouts Off at Aspen

People are catching on to the rather substantial conflict of interest that might affect the enthusiasm with which large shareholders in AI companies predict that their own products will change the world. But many still think that if other business leaders predict disruption, then you should listen. Allow me to suggest otherwise, with reference to this fine Wall Street Journal article from last week: CEOs Start Saying the Quiet Part Out Loud: AI Will Wipe Out Jobs.

It opens, as good think pieces about visionary business leaders are wont to do, with a scene from this year’s Aspen Ideas Festival.

“Artificial intelligence is going to replace literally half of all white-collar workers in the U.S.,” Ford Motor Chief Executive Jim Farley said in an interview last week with author Walter Isaacson at the Aspen Ideas Festival.

Farley got a nice big picture of himself atop the article, delivering a TED-Talk-style keynote under bright spotlights. You’ll have to scroll down 17 more paragraphs though, to the end of the article, to meet Pascal Desroches, CFO at AT&T, who

said in an interview last month that much remains unclear about how work will be reshaped by AI and that past technological revolutions have shown that new jobs often emerge. “It’s hard to say unequivocally, ‘Oh, we’re going to have less employees who are going to be more productive,” he said. “We just don’t know.”

Sorry Pascal, not very visionary! Can you imagine the Journal running a story titled, “CEOs Start Saying the Quiet Part Out Loud: Who Knows If AI Will Have Much Effect,” leading with AT&T’s Desroches, and mentioning Ford’s Farley at the article’s end? Me neither. Does Farley face even the slightest risk of ridicule or any potential cost for his Aspen declaration proving untrue? No. The lesson for business is leaders is to be a Farley, not a Desroches. The lesson for everyone listening to business leaders is to keep that in mind.

Act II: The Visionary CEO Puts His Money Where His Mouth Is

So we probably shouldn’t give too much credence to attention-getting puffery. But what about when the AI deployment actually happens? Surely that is proof of concept. Allow me to suggest otherwise, with reference to this fine Bloomberg article from the prior week: Salesforce CEO Says 30% of Internal Work Is Being Handled by AI.

Actually, the headline undersells Marc Benioff’s claim, which is that “AI is doing 30% to 50% of the work at Salesforce now.” Incredible—and an empirical, testable statement about the present. Must be true. A CEO fabricating something like that could face a charge of securities fraud.

What does it look like when a company hands 30% to 50% of its work to AI? I headed over to its financial statements to find out. Comparing this year’s Q1 to last year’s, operating expenses had plunged an impressive… no, operating expenses were up 6.3%. Maybe that’s good news. They’re not using the leverage of AI doing half the work to lay people off, they’re using it to do twice as much! Revenue must be through the roof. Sure enough, it was up… 7.6%.

In other words, a company that claims as much as 50% of its work is now done by AI appears not to have decreased its operating costs or increased its output beyond typical mediocre growth. Strange, that.

Act III: The Reality

Maybe it’s not so strange. Benioff claims that 50% of the work was done by AI, not that the work AI did was any good. And here we must turn to a fascinating study just published by Model Evaluation & Threat Research: Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity.

METR did a simple randomized controlled trial, comparing the productivity of experienced open-source developers using AI tools versus not. Whereas economic experts and machine learning experts predicted that use of AI would dramatically improve productivity, and the developers themselves assessed AI as substantially improving their productivity, in fact use of AI reduced productivity by 19%. It turns out you can have AI do a lot of work, and end up getting less done than before!

(I have a confession to make. AI now does 90% of the work here at Understanding America. For every 100 words of text I write, a frontier model cranks out roughly 900 words of other useless research and analysis that I ignore completely. But you can rest assured you are reading a cutting edge Substack. 90% is massive. A paradigm shift. I’ll be explaining further what this means for the future of humanity in my upcoming TED Talk.)

One of the participants in the study posted an interesting thread offering his own reflections on why AI had likely made him less productive. He offers an interesting perspective on both the technical—what type of coding are LLMs good at?—and the human—if you get distracted scrolling social media for 30 minutes every time you ask your LLM to generate some code, you’re going to be disappointed in your productivity.

I would add to this one other observation. By happenstance, this year marks 50 years since publication of The Mythical Man-Month, Fred Brooks’s seminal study of the challenges in coordinating large software engineering projects at the dawn of the industry. To oversimplify, Brooks’s central point was that the capacity of your development team scales linearly (10 developers can write twice as much code as 5 developers) but the coordination and complexity of your project scales quadratically (10 developers and their code bases have ~100 potential interfaces amongst themselves, while 5 have only ~25). Add one developer to your team and you’ve added one developer’s capacity, while adding new coordination problems for everyone. At the extreme, “Brooks’s Law” holds that “adding manpower to a late software project makes it later.”

I suspect we’ll see a similar challenge with AI. What seems most exciting and transformational about it—the “agentic” ability for a tool to act autonomously—also imposes an obligation of “management” on the user that is much more like supervising someone than using a tool. Great managers get “leverage”: Give one a team of four and he might get a project done two or three times as fast as he could do it himself (but even the best will not get it done five times as fast). Give a bad manager, or someone with no management experience, a team of four and the project probably won’t get done at all—certainly not on a useful timeline or at a high quality. The vast majority of people, we are setting ourselves up to discover in the next few years, are not good managers.

WHAT ELSE SHOULD YOU BE READING?

The US Can Beat China on Critical Minerals But Not by Copying It | Daleep Singh, Financial Times

A good overview of the challenges for the United States in rebuilding critical minerals supply chains, made more relevant by the Pentagon’s announcement last week that it would take a major equity stake in MP Materials, complete with purchase commitments and a price floor. It’s an impressive and ambitious deal that “Is Making Former Biden Officials Jealous.”

The Golden Age for Employers Is Ending | Adrian Wooldridge, Bloomberg

This column inadvertently makes the best possible case for immigration restriction as a tool of labor market policy. “Business will have to adjust to a world in which immigrants are much rarer, and jobs are harder to fill,” writes Wooldridge, the Bloomberg global business columnist, and he suggests two essential strategies that firms must now adopt.

First, “start treating the supply chain of talent with the same thoroughness as they treat the supply chain of materials and parts: as something that needs to be carefully planned and constantly repaired. … Tap into the neglected pool of talent in working-class communities by sponsoring charter schools, endowing university scholarships or spotting and then sponsoring talented students. Improve vocational education by introducing apprenticeship programs or forging relationships with local high schools. Take a more prominent role in providing career guidance in local schools. Employ more unconventional workers such as older workers by introducing flexible working hours or extended holidays.”

Second, “think harder about how to apply labor-saving techniques to production. This is particularly true of industries that now rely on imported labor such as construction, farming and hospitality. … In agriculture, ‘intelligent’ AI-powered machines can do things that were too delicate for dumb machines, such as nurturing seeds and controlling weeds. In hospitality, robots can do ever more delicate cleaning jobs.”

Bloomberg is the sort of outlet that usually asks questions like, “Do immigrants really take jobs from Americans? And does more immigration really mean lower wages?” and then makes arguments like “the economy is not a zero-sum game. Foreign-born workers are more likely to work in sectors where demand for workers often outpaces the native-born supply.” and “the impact of immigration on the wages of native-born workers is overall ‘very small.’”

So it’s refreshing to see, when the rubber meets the road, people are starting to recognize the more fundamental ways an unrestricted flow of cheap foreign labor warps the labor market and the benefits that could come if employers are instead constrained.

Restraint Is Conservative | Michael Brendan Dougherty, National Review

Speaking of constraint, or restraint, anyway… great commentary from Michael, here, debunking the sorts of silly, bad-faith arguments flung by neoconservatives at anyone who deigns to note that neither U.S. capacity nor U.S. interests allow for the “benevolent hegemony” that they wish we could still exert.

Bonus link: Need an extreme example of failed-neocon rhetoric? David Frum delivers on Independence Day. “Hard to take seriously the red-white-and-blue rhetoric of freedom on a 4th of July where the US president has cut off aid to a democracy fighting for its independence against ferocious attack by a foreign tyrant.” Don’t make me tap the Farewell Address, David…

Seriously, though—anyone who thinks that “the red-white-and-blue rhetoric of freedom” is incompatible with an American president declining to “entangle our peace and prosperity in the toils of European ambition, rivalship, interest, humor, or caprice” should perhaps consult George Washington’s words at the end of his second term. That’s not dispositive on the substantive question. One is of course free to advocate for a more interventionist foreign policy. But suggesting that only one position is compatible with the American “rhetoric of freedom” is ahistorical to say the least. Or, to say more, it’s a good illustration of why Frum-style foreign adventurism is met with such disdain, rather than merely disagreement, on the New Right.

The Stocks Will Be Tokenized | Matt Levine, Bloomberg

Regulators have been flummoxed by the question of how to treat blockchain “tokens”—intangible digital assets that can represent just about anything. The most well-known example is cryptocurrency, which certainly feels like a typical financial asset, but start-ups are trying to tokenize all sorts of things and they (and their venture capital investors) argue that these tokens will be manifestations of goods and services in the digital world just like, well, goods and services in the physical world. Your couch is an asset, but the furniture store doesn’t have to file any disclosures with the Securities and Exchange Commission before selling it to you.

That almost seems like a compelling argument, in theory, until you discover that what people are in fact eager to use the blockchain for is… circumventing securities regulation to transact financial securities! This sort of regulatory arbitrage is the worst of all worlds, leaving what may be onerous regulation in place for people who want to do things the normal way, while creating new and convoluted hoops to jump through if you want to do things that should be regulated in an unregulated way. If you think stock trading should be less regulated, by all means, make the case. But don’t sell “here’s a way to get around regulation that no one wants to remove” as “innovation.”

UPDATING YOU ON AN ONGOING STORY…

TSMC Reportedly to Break Ground on U.S. Advanced Packaging Plants in 2028 | TrendForce

If you enjoyed “TSMC to Delay Japan Chip Plant and Prioritize U.S. to Avoid Trump Tariffs,” you’ll be pleased to know they are also investing downstream in the supply chain.

There’s a Race to Power the Future. China Is Pulling Away. | David Gelles et al, New York Times

If you enjoyed “Why Americans Can’t Buy the World’s Best Electric Car,” it’s worth noting that we can’t compete with China in clean energy either, for many of the same reasons. The Times makes a halfhearted effort to pin the failure on Trump administration opposition but, as the story acknowledges, “Only toward the end of the Obama administration and during the first Trump administration did many Washington policymakers realize they had surrendered so much of the clean energy race to China.” We lost this game during Obama’s second term, while he was out on the stump promising the clean energy jobs of the future right here in America.

Heritage Foundation’s Richard Stern on X

If you enjoyed, “Meet the People Who Think a Tax Increase Is an ‘Anti-Christian Assault,’” you’ll be fascinated that the Heritage Foundation followed up by posting a video to its official X account, in which Stern declares, “We have to keep in mind that every time the government spends a dollar, it has stolen that dollar from somebody that worked hard to earn it.” As I noted in response, “I just question whether outright rejection of western political philosophy, the U.S. Constitution, and public policy is the best foundation atop which to build a conservative agenda.”

Have You Heard the Good News? | Michael Strain & Clifford Asness, The Free Press

If you enjoyed, “You’re So Vain, You Probably Think This Post Is About You,” be sure to ponder this passive-aggressive observation that, “It’s interesting that populist politicians and opinion leaders are so excited for others—not for themselves—to leave service-sector jobs in order to stand in front of 4,000-degree blast furnaces.” Apparently, policy wonks should only advocate for the creation of jobs they would choose over their own jobs as policy wonks?

This could make sense if everyone working in a low-wage service-sector job has the opportunity to work instead as a policy wonk. Unfortunately, a bachelor’s degree is required for 13 of the 14 positions currently open at the American Enterprise Institute, including that of research assistant in the perhaps not quite aptly named “Center on Opportunity and Social Mobility.” If you’d like to be a fellow, you need a post-doctorate degree. Perhaps an assessment of the job options for the majority of workers without bachelor’s degrees should focus on the types of jobs in which they likely work.

US Public Colleges Expand PE Investments Despite Downturn | Sun Yu, Financial Times

If you enjoyed “Private market funds lag US stocks over short and long term. State Street private equity index outpaced by S&P 500 over 3 months and 1, 3, 5 and 10 years,” meet the state-school endowment managers eager for the opportunity to jump into the market on which their more sophisticated Ivy-League counterparts are bailing.

Bonus link: Scrutiny of the Ivies is now extending to valuation of their private-fund portfolios.

AND AT COMMONPLACE

Don’t Let the Fake Social Security Crisis Fool You by Michael Lind. Congress should do away with the Trust Fund model, which only confuses the fiscal situation.

An Unlikely Comeback by Bradford Tuckfield. How small pizza shop owners reversed the trend of market concentration.

On the American Compass podcast, Notre Dame professor Patrick Deneen joins me to unpack his observation that we are all post-liberals now, to delineate the various attempts at a successor ideology, and to discuss what might come after post-liberalism.

Enjoy the rest of the weekend!