E165: Vision Pro: use or lose? Meta vs Snap, SaaS recovery, AI investing, rolling real estate crisis

If Apple Vision Pro (or similar AR goggles) are deployed to the greenhouse/lab technician workflows Freiberg describes (scanning QR codes on flowers, image/data capture, task lists), they will ultimately increase worker productivity in that job by roughly 10x compared to the current iPhone + scanner + printer setup, once the necessary software is built and adopted.

“literally every aspect of this job will be massively improved and productivity will go up by ten x with these goggles. Will it happen in the next couple of weeks or months, I don't know, but my engineering team is looking into it.” View on YouTube

Over the coming years, negative outcomes currently observed in younger generations (e.g., depression, suicide, drug/SSRI dependence, reduced marriage and childbirth) will worsen, at least at the margin, as a byproduct of increasingly immersive consumer technologies (social media, VR/AR, etc.), rather than improve.

“I suspect on the margin, if you were going to bet all of these things that we see in these young people today will get worse as a byproduct of technology, not necessarily get better.” View on YouTube

Apple will generate at least $100 billion in cumulative revenue from Apple Vision Pro hardware (current and next-generation models) within less than five years of launch (i.e., before February 2029).

“I think they're going to sell $100 billion of Apple Vision pros, not this version, but this version plus the next version, probably over the next, I would guess for them to get to 100 billion in sales. It'll take them less than five years.” View on YouTube

Apple will become the dominant vendor in the high-end VR/AR headset category created around devices like Vision Pro, effectively capturing the majority of the market versus competitors such as Meta and others over the coming years.

“I think this is gonna they're gonna run the table on this.” View on YouTube

In the mature AI market, large foundational language models as a category will generate little to no direct economic value because powerful models will be broadly available for free (or effectively free), making it impossible to sustain high-margin, closed foundational model businesses trained primarily on open internet data.

“I think foundational models will have no economic value. I think that they will be an incredibly powerful part of the substrate, and they will be broadly available and entirely free.” View on YouTube

Over time, open-source AI models will erode the economic value of general-purpose model providers to near zero, while (a) infrastructure "picks and shovels" providers (especially those with proprietary AI hardware and tokens-per-second services) and (b) owners of valuable proprietary datasets will capture most of the sustainable economic gains from AI.

“So my refined thoughts today are sort of what my initial guess was when we started talking about AI a year ago, which is the picks and shovels. Providers can make a ton of money, and the people that own proprietary data can make a ton of money. But I think open source models will basically crush the value of models to zero economically. Even though the utility will go to infinity, the economic value will go to zero.” View on YouTube

Over the next several years, OpenAI will maintain a performance lead over open-source models and other competitors sufficient for it to remain the leading commercial AI model provider and to be a financially successful company.

“I do think there is an argument that open AI will stay in the lead and actually do quite well.” View on YouTube

If OpenAI continues to keep even a modest quality lead over open-source models, it will capture the vast majority of the consumer-facing GPT/query market (analogous to Google’s dominance in search), with most users preferring OpenAI’s service over alternatives.

“if OpenAI just maintains a little bit of a lead over open source, then it could basically when the vast, vast majority of the call it consumer search or consumer GPT market.” View on YouTube

OpenAI’s custom GPT platform and associated developer network effects will create a self-reinforcing ecosystem around ChatGPT that open-source model ecosystems will find difficult to match, leading to a durable advantage for OpenAI in capabilities and available applications.

“So you have a classic developer network effect where you've got OpenAI aggregating hundreds of millions of consumers because they perceive that ChatGPT is the best. Then you've got developers wanting to reach that audience. So they build custom gpts on the OpenAI platform. That actually gives ChatGPT more capability. Yeah. And that's something that open source can't easily catch up with.” View on YouTube

By sometime between February 2025 and August 2025, the quality of leading large language models trained primarily on the open internet—specifically OpenAI’s model, Meta’s Llama, Mistral, and xAI’s model—will have converged such that on standard third‑party benchmarks they achieve roughly the same performance level (no single model having a large, clear quality lead).

“I think they're all going to converge to the same quality in the next, probably 12 to 18 months.” View on YouTube

By sometime between August 2024 and November 2024, OpenAI’s frontier model and open‑source models like Llama and Mistral (trained on the open internet) will have reached essentially the same quality level on common evaluation benchmarks, eliminating OpenAI’s meaningful quality advantage that exists in February 2024.

“they're all getting to the same quality code point and they will be there within the next 6 to 9 months.” View on YouTube

If US office commercial real estate values fall by roughly two‑thirds as implied (from about $3T to $1.8T), resulting in large losses to pension and retirement funds, then within the following few years the US federal government and/or Congress will implement a material structured support or bailout program specifically aimed at cushioning retirees and pensioners from these office‑related losses (rather than allowing those write‑downs to fully hit beneficiaries).

“You're not going to see governments let that happen. You're going to see the federal government. There's going to be some action at some point, and it's unlikely the office market is going to suddenly rebound overnight. If this stays the way it is, who's going to fill that hole for retirees and pensioners? Because we're not going to let that all get written down. Someone is going to step in and say, we've got to do something about this, and there's going to need to be some sort of structured solution to support retirees and pensioners, because that's ultimately who ends up holding the bag in this massive write down.” View on YouTube

Over the ensuing period after February 2024, a significant portion of regional US banks holding office‑backed commercial real‑estate loan portfolios will experience material impairments and financial stress (including stock price declines and/or solvency concerns) as office values fall and loans are written down.

“there's a huge amount of equity that's been written off. But in addition to that, there's a lot of debt holders who are in trouble too. And that debt is is held by regional banks. So these commercial loan portfolios are significantly impaired.” View on YouTube

From 2024 onward, the US commercial and multifamily real estate sector will experience a prolonged, staggered ‘rolling’ crisis, with waves of distress and defaults occurring over multiple years as loans and leases sequentially come up for refinancing, rather than a single short, acute crisis concentrated in one year.

“that's why there's a rolling crisis in real estate is because the debt rolls over time. It's not like everybody hits the wall and has to refinance at the same time.” View on YouTube

In response to mounting losses in US commercial real estate (particularly office) that threaten creditors and investors, the US Treasury under Janet Yellen (or equivalent federal authorities if she is no longer in office) will ultimately implement a bailout or support mechanism that protects the main creditors/investors exposed to these real‑estate debts, while avoiding an explicit direct bailout of the banks themselves.

“Yeah, I mean Janet Yellen's just going to bail these folks out. I mean, she won't bail out the banks themselves, but she'll bail out the creditors. Obviously the people holding the bag, they'll get bailed.” View on YouTube