The current narrative of AI eating up college graduate entry-level jobs will have interesting and lasting ramifications if true. But before we get there, let’s back up for a second. Here is where we (seemingly) are at:

George Arison, CEO of Grindr: “Companies are ‘going to need less and less people at the bottom.’” […] Matt Sigelman, President of Burning Glass Institute: “This is a more tectonic shift in the way employers are hiring. Employers are significantly more likely to be letting go of their workers at the entry level—and in many cases are stepping up their hiring of more experienced professionals.” […] Ford CEO Jim Farley: Stated he “expects AI will replace half of the white-collar workforce in the U.S.”

And it’s not just CEOs talking about this shift:

Jadin Tate, University at Albany graduate: Recounted his mentor’s warning that his chosen field is being “taken over by AI” and “may not exist in five years.” […] Arjun Dabir, student at University of California, Irvine, on intern work: “That task is no longer necessary. You don’t need to hire someone to do it.”

Which is all well and good for the hiring companies (and, of course, terrible for college grads) – but just like China’s infamous “one child per family”-policy, it will bite you in the tail down the line. You might not need the entry-level worker anymore – but how does someone progress to a mid- to high-level worker if they never had the chance to, well, start somewhere?

Chris Ernst, Chief Learning Officer at Workday: “Genuine learning, growth, adaptation—it comes from doing the hard work. It’s those moments of challenge, of hardship—that’s the crucible where people grow, they change, they learn most profoundly.”

Time will tell.

Link to article in WSJ.

Vogue, the iconic fashion magazine, used AI to generate “models” in its latest edition – and caused quite the stir:

Vogue’s August 2025 issue, starring Anne Hathaway on the cover, has ignited a heated debate because of its use of AI-generated models. […] The inclusion of AI-generated “models” has led to subscription cancellations and criticism online.

There are, of course, many angles to this critique – from concern about jobs (models, makeup artists, photographers, etc.), to issues with the uncanny valley (“Although the models wear real fashion from top labels, many say the images resemble luxury video game renders more than genuine editorials.”), to more philosophical questions around “detractors believe it sacrifices emotional depth and the artistry that human models bring.”

All of which brings up an interesting point – with AI and AI-generated “art” becoming more and more prevalent, where do we draw the line? And what’s the market size for both the AI-enabled (digital) and AI-free (analog) world? Draw a parallel to the world of music, and you see a niche market for vinyl records emerge from the depths of the streaming platforms – but it’s tiny in comparison and surely will always be tiny.

Fashion fans aren’t just reacting emotionally, they’re calling out a deeper concern about the future of representation and authenticity in the industry.

Link to article.

I recall when Beyond Meat was the “hot thing” – we fed the participants of the Singularity University Executive Program Beyond Meat meatballs and burgers, which were, at the time, quite difficult to source. It was the future. On your plate. Now it is something only shortsellers appreciate.

From a fundamental perspective, Beyond Meat is one of the worst stocks in the entire market. […] Any purely financial model here would suggest that the equity is worth zero, and that in 2027 the Beyond Meat business will wind up in the hands of its bondholders.

The markets are a harsh mistress:

Beyond Meat’s plan was to change the world; yet it almost certainly won’t be able to pay its debts.

Link to article.

Here’s an interesting one (talking about the implications of the implication – or, in other words, our Disruption Mapping tool):

GLP-1s (Ozempic) have the potential to break the life insurance industry – and maybe not for the reasons you would expect.

Life insurers can predict when you'll die with about 98% accuracy. […] Typically, underwriters- suspiciously sounds like undertakers-rely on a handful of key health metrics like HbA1c, cholesterol, blood pressure, and BMI to calculate your risk of dying earlier than expected (and thus costing them money). Those eagle-eyed readers among you have probably noticed something interesting already. Those same four metrics are exactly what GLP‑1s improve. Not just a little, but enough to entirely shift someone's risk profile within at least 6 months of using them.

The insurer sees a 'mirage' of good health and approves them as low-risk. […] If we assume about 65% of people who start GLP-1 medications quit by the end of year one, that creates a big problem. When someone stops the medication, they'll usually regain the weight they lost, and in two years, most of those key health indicators bounce back to their starting point.

Yep, it’s going to get messy.

Link to article.

As the saying goes: A picture is worth a thousand words.

The Federal Reserve Bank of St. Louis has updated the World Uncertainty Index for the second quarter of 2025 , a period characterized by US tariffs, the One Big Beautiful Bill Act, and a slew of other actions across the globe.

And you thought the COVID years were bad…

Link to report.

More food for thought:

The rise of Artificial General Intelligence (AGI) marks an existential rupture in economic and political order, dissolving the historic boundaries between labor and capital. Unlike past technological advancements, AGI is both a worker and an owner, producing economic value while concentrating power in those who control its infrastructure. Left unchecked, this shift risks exacerbating inequality, eroding democratic agency, and entrenching techno-feudalism. The classical Social Contract-rooted in human labor as the foundation of economic participation-must be renegotiated to prevent mass disenfranchisement.

Link to paper.

Can AI run your business? Anthropic (maker of the Claude AI models) wanted to find out:

"We let Claude manage an automated store in our office as a small business for about a month. We learned a lot from how close it was to success—and the curious ways that it failed—about the plausible, strange, not-too-distant future in which AI models are autonomously running things in the real economy.”

The short answer: No. But there is a whole lot more to look at and learn from the experiment:

"It's worth remembering that the AI won't have to be perfect to be adopted; it will just have to be competitive with human performance at a lower cost in some cases.”

"An AI that can improve itself *and* earn money without human intervention would be a striking new actor in economic and political life.”

And it comes with a bunch of warnings/red flags:

"We do think this illustrates something important about the unpredictability of these models in long-context settings and a call to consider *the externalities of autonomy*.”

"In a world where larger fractions of economic activity are autonomously managed by AI agents, odd scenarios like this could have cascading effects—especially if multiple agents based on similar underlying models tend to go wrong for similar reasons.”

In summary:

"Although this might seem counterintuitive based on the bottom-line results, we think this experiment suggests that AI middle-managers are plausibly on the horizon.”

The whole study is worth perusing.

It took a while, but AI is coming for the enterprise for real now…

"The Census conducts a biweekly survey of 1.2 million firms, and one question is whether a business has used AI tools such as machine learning, natural language processing, virtual agents, or voice recognition to help produce goods or services in the past two weeks, see chart below. Nine percent of firms reported using AI, and the rising trend in AI adoption increases the likelihood of a rise in productivity over the coming quarters.”

Source: Apollo Global Management 

Microsoft’s WorkLab team is out with a new report on what our workplace looks like mid-2025 – and the results are ugly:

Nearly half of employees (48%) – and more than half of leaders (52%) – say their work feels chaotic and fragmented. […]

Half (50%) of all meetings take place between 9–11 am and 1–3 pm—precisely when, as research shows, many people have a natural productivity spike in their day. […]

On average, employees using Microsoft 365 are interrupted every 2 minutes by a meeting, email, or notification.

And is goes on and on… Particularly hilarious (and sad is this):

In the final 10 minutes before a meeting, PowerPoint edits spike 122% – the digital equivalent of cramming before an exam.

The report offers some positive outlook though:

The future of work won't be defined by how much drudgery we automate, but by what we choose to fundamentally reimagine. […] The most effective organizations know this—and act on it. Frontier Firms are putting the Pareto Principle into practice, focusing on the 20% of work that delivers 80% of the outcomes.

Link to report.

Obi, an app that aggregates real-time pricing and pick-up times across multiple ride-hailing services, has just published what it’s calling the “first in-depth examination of Waymo’s pricing strategy.” The company found Waymo’s self-driving car rides to be consistently more expensive than comparative offerings from Uber and Lyft — and it doesn’t seem to matter.

I get why people ride Waymo’s (novelty), but this is telling:

“Colloquially, there is an idea that autonomous vehicles are something that will erode driver jobs and put drivers at risk. And I think the irony of what we’ve seen is that it’s actually quite expensive to run an AV, and that that’s not going to be happening, at least in the near term,” she said. 

Link to article and study.

Take this with a grain of salt – as it comes from the very people who would greatly benefit from this statement being true (venture capital investors) – but this sentence in the article stood out:

Tech companies, in particular, will see "significantly reduced hiring," he added.

Ironically, the very people building all this stuff might be the ones most affected by it (on a per-sector/industry level; maybe not on the individual level… AI engineers will be fine for a while).

Link to article.

New study about the use of AI in German offices shows that Germans don’t use AI in their workplaces all that much and that they are increasingly wary and mistrusting of AI compared to their international peers. That being said, even with those caveats, a solid 42% of white-collar workers are already using AI in their jobs. Given that ChatGPT is just a little over two years old, this is an impressive penetration number.

Link to article and study.

Wondering what the world of “work from home” truly looks like? Despite the rhetoric that it's over, the average (!) American spends 1.6 days per week working from home…

Source: Apollo research​

On the heels of my last post about Google co-founder Sergey Brin’s rather questionable approach to using AI to replace managers (or, at least, replace a lot of what they do), here is new research on the impact AI is having on the workplace.

From the study “The Dark Side of Artificial Intelligence Adoption: Linking Artificial Intelligence Adoption to Employee Depression via Psychological Safety and Ethical Leadership” published in Nature:

“While the adoption of AI brings numerous benefits, it also introduces negative aspects that may adversely affect employee well-being, including psychological distress and depression.”

The study goes on to say that

“With AI adoption, there is a real risk that employees will perceive threats to crucial resources like job stability, independence in their work, and professional competencies.”

In summary (and very much contrary to Mr. Brin’s approach):

“With the continuous advancement and integration of AI technology in our workplaces, it is imperative for organizations and leaders to prioritize the welfare of their employees, foster a supportive and inclusive working environment, and embrace an ethical approach that prioritizes people when incorporating AI.”

Link to study.

Google cofounder Sergey Brin recently rambled his way through a conversation on management in the age of AI:

"Management is like the easiest thing to do with AI," Brin said.

Apparently, management, for Brin, consists of summarizing meetings and assigning to-dos:

"It could suck down a whole chat space and then answer pretty complicated questions," he said. "I was like: 'OK, summarize this for me. OK, now assign something for everyone to work on.’”

So far, so bad. Where it gets really fun is when he lets AI make promotion decisions:

"It actually picked out this young woman engineer who I didn't even notice; she wasn't very vocal," he said. "I talked to the manager, actually, and he was like, 'Yeah, you know what? You're right. Like she's been working really hard, did all these things.’”

And as he clearly has outsourced his management to an AI, he doesn’t even really know if this all has happened or not:

"I think that ended up happening, actually," Brin said of the promotion.

All in all, it’s a pretty bleak vision for the future.

Link to article.

Think AIs will come for your CFO and his team? Not quite, not yet… Vals.ai’s Finance Agent Benchmark clearly shows we are some ways off from the professions caving in to our LLM-powered overlords:

“The foundation models are currently ill-suited to perform open-ended questions expected of entry-level finance analysts”

But fret not — not all is lost:

“Models on average performed best in the simple quantitative (37.57% average accuracy) and qualitative retrieval (30.79% average accuracy) tasks. These tasks are easy but time-intensive for finance analysts.”

Link to benchmark and analysis.

A new study by the Booth School of Business at the University of Chicago shows that the use of GenAI, at least so far, hasn’t had any measurable impact on jobs and wages:

“AI chatbots have had no significant impact on earnings or recorded hours in any occupation," the authors state in their paper.

One might consider this a reason for celebration (“no, AI won’t make you unemployed”), but it has a very important implication:

“The adoption of these chatbots has been remarkably fast," Humlum told The Register. "Most workers in the exposed occupations have now adopted these chatbots. Employers are also shifting gears and actively encouraging it. But then when we look at the economic outcomes, it really has not moved the needle.”

Despite the billions of dollars being poured into AI, the economic impact (again, so far) seems to have been minimal…

Link to article and study.

Many people, myself included, didn't try to build a product around a language model because during the time you would work on a business-specific dataset, a larger generalist model will be released that will be as good for your business tasks as your smaller specialized model.

This being said – with generalized models advances slowing down (see our post from earlier today), now might be the time to build:

If your business idea isn't in these domains, now is the time to start building your business-specific dataset. The potential increase in generalist models' skills will no longer be a threat.

Source.

This doesn’t come as a surprise, nor is it entirely new - but seeing it in such stark light makes you wonder…

OpenAI (and likely everyone else) has a serious copyright problem. It makes me wonder when the copyright holders will start to actually fight back:

It’s a reminder that LLMs of this type and size all train on copywritten material.

Click through to the article from Otakar G. Hubschmann to see some of his findings – if you are not convinced that AI is trained on oodles of copyrighted material, this will make you see it.

Bill Gates said at a Goldman Sachs-sponsored event in L.A. two years ago:

"Whoever wins the personal agent, that will be a big thing because you’ll never go to a search site again. You’ll never go to a productivity tool again. You’ll never go to Amazon again. Everything will be mediated through your agent."

The first part is becoming true now – Adobe reported that traffic from generative AI sources (e.g., the Perplexity AI search tool) to e-commerce websites jumped a whopping 1,200 percent in their latest report.

In February 2025, traffic from generative AI sources increased by 1,200 percent compared to July 2024.

No wonder Google is simultaneously freaking out and investing heavily in an AI-powered future of search. But the important bit is not just the sheer increase in volume referred to by generative AI tools, but also the quality of this traffic:

[…] consumers coming from generative AI sources show 8 percent higher engagement as they linger on the site for a longer period of time. These visitors also browse 12 percent more pages per visit, with a 23 percent lower bounce rate. This speaks to the value of conversational interfaces in online shopping, which seem to help consumers be more informed and confident in their purchases.

Link to report.

While one party seems to never end, with OpenAI raising an eye-watering $40 billion at an even more eye-watering $300 billion valuation, other parties are being called off…

Microsoft is calling off some of its data center deals:

The software company has recently halted talks for, or delayed development of, sites in Indonesia, the UK, Australia, Illinois, North Dakota and Wisconsin, according to people familiar with the situation.

Microsoft is widely seen as a leader in commercializing AI services, largely thanks to its close partnership with OpenAI. Investors closely track Microsoft’s spending plans to get a sense of long-term customer demand for cloud and AI services.

One possible reason being discussed by analysts is the increasing efficiency that some AI models exhibit:

Analysts have stepped up their scrutiny of data center spending since Chinese upstart DeepSeek announced in January that it had created a competitive AI service using fewer resources than leading US companies.

Link to article.

Ostensibly written about Nvidia’s plight of having to move to bigger and bigger silicon as Moore’s Law ("the number of transistors per square inch on integrated circuits doubles approximately every two years”) is dead – this is, of course, much bigger than a single chip manufacturer.

Advancements in process technology have slowed to a crawl in recent years. While there are still knobs to turn, they're getting exponentially harder to budge. Faced with these limitations, Nvidia's strategy is simple: scaling up the amount of silicon in each compute node as far as they can.

[…]

In any case, Nvidia's path forward is clear: its compute platforms are only going to get bigger, denser, hotter and more power hungry from here on out. As a calorie deprived Huang put it during his press Q&A last week, the practical limit for a rack is however much power you can feed it.

Link to article.

Assuming this is not just a thing DoorDash did for marketing purposes, it surely points to a worrying trend:

DoorDash users can use Klarna to pay in four, interest-free payments or defer payments and let people pick a 'date that aligns with their paycheck schedules.'

Link to article.

Benn Stancil’s commentary about the changing nature of software development (vibe coding is all the rage now, kids!) draws an important conclusion: If developing software (aka writing code) becomes more and more democratized, what stronghold do places like Silicon Valley have on innovation?

Taking this thought a step further – if the value is less and less in the software development process and rather in domain expertise in the problem space, will we see a geographic shift of innovation ecosystems toward their respective client spaces?

Just as it's becoming harder to out-write an LLM, it's becoming harder to out-develop one too. And if experts can prompt their way to a product just as easily as those of us in Silicon Valley can, what winning talent are we left with?

Link to article.

Fascinating conversation with Young Cho, head of research at the financial analysis firm Jane Street, on the challenges of using machine learning and LLMs in the context of financial data.

Machine learning in a financial context is just really, really hard... you can think of machine learning in finance is similar to building an LLM or text modeling except that instead of having, let's say one unit of data, you have 100 units of data. That sounds great. However, you have one unit of useful data and 99 units of garbage and you do not know what the useful data is and you do not know what the garbage or noise is.

Link to podcast and transcript.

Here is a fascinating paper examining the impact that future assumptions about Artificial General Intelligence (AGI) (or "Transformative AI (TAI)” as outlined in the paper) becoming real will have on people’s behavior today.

The main takeaway: Just the belief that transformative AI is coming could push interest rates much higher, even before the technology actually exists. This, in turn, could affect how central banks manage the economy and overall financial stability.

The train of thought works like this:

1. The key idea is that when advanced AI arrives, it will replace human workers, and the money that used to go to workers will instead go to people who own AI systems

2. The more wealth you have when AI arrives, the more control you'll have over AI systems and their earnings

3. The researchers used economic models based on current predictions about when this powerful AI might arrive

4. They found that even moderate predictions about this future scenario are causing some interesting effects today:

- Interest rates could rise much higher (to 10-16%) compared to normal rates (around 3%)

- People are willing to accept lower returns on investments now because they're focused on building wealth to control future AI systems.

Link to the research paper.

When you follow the news headlines these days, the pervasive trail would lead you to think that RTO (Return to Office) is the order-of-the-day and we, collectively, are returning to a world of people in offices. Alas, when you look at the data, we are far from it:

The average occupancy in the top 10 office markets in the latest week was still only at 54% of where it had been before Covid, so still down by 46% from pre-Covid, and only a few percentage points higher of where it had been at the same time in 2023, and just a hair higher than at the same time in 2024.

When you look at the data from the opposite side – people working from home – you see the same picture emerge:

But as the economy reopened, those service workers were called back, and the share of WFH as a percent of overall full paid days plunged. But for office workers, WFH has remained a big factor. As a result, the share of full paid days worked from home remains at far higher levels than before Covid and has not come down further in 2023, 2024, and in January 2025.

Full analysis on Wolf Street.

A good reminder from Laurie Voss about the implications of AI on the job market – it’s the same argument Andrew Ng has been making for years now:

Jobs are more than collections of tasks. Jobs require prioritization, judgement of exceptional situations, the ability to communicate ad-hoc with other sources of information like colleagues or regulations, the ability to react to entirely unforseen circumstances, and a whole lot of experience. As I said, LLMs can deal with a certain amount of ambiguity and complexity, but the less the better. Giving them a whole, human-sized job is way more ambiguity and complexity than they can handle. It's asking them to turn text into more text. It's not going to work.

Source.

Ever wondered why Apple is special? Their AirPods business (you know, those little white earbuds you see everywhere) is a whopping $18,000,000,000 per year revenue driver.

> Bloomberg estimates that AirPods sales have exceeded $18 billion yearly since 2021. To put that number in perspective, it surpasses Nintendo's reported total net sales for 2023 (roughly $10 billion). Furthermore, PCMag recently calculated that AirPods generated more revenue than total annual earnings for companies like Spotify, eBay, Airbnb, DoorDash, and OpenAI.

Nearly twice as much as all of Nintendo, more revenue than a lot of Silicon Valley’s darlings. Chapeau, Tim Cook.

Link to article.