A Question Fascinating Me Right Now

I have seen less talk than I expect on if Large Language Models (LLMs) are a technology where little of the value will accrue to the companies and investors, but instead is reaped by consumers. Economist call this “consumer surplus” where most of the value created by a product or industry is captured by the customer. 

The most famous of these is the technology of aviation and the airlines throughout their history. Flight is as magical of a technology as AI LLMs, that dramatically changed the world, yet the airline industry has been a value destroyer for its investors and for most of its history a terrible business to be in. However, it has been amazing for its customers and mankind, reshaping our world for the last 100 years. The reason why miraculous technologies can make poor businesses is extremely high fixed costs (data centers, GPUs/TPUS, model training), and intense competition akin to a knife fight.

Warren Buffet calls these type of industries a “death trap for investors”. Regarding the airlines and aviation technology his quotes summarize perfectly what might happen with foundational LLM companies and the industry:

“Investors have poured money into a bottomless pit, attracted by growth when they should have been repelled by it.”

“The airline business has been extraordinary. It has eaten up capital over the past century like almost no other business…”

Foundational LLMs take a horrendous amount of capital and infrastructure to build, let alone stay on the cutting edge. At the time of this writing (September 2025) credible reports on OpenAI’s financials indicate an expected $13Bn in revenue in 2025 leading to an $8Bn loss with a 2026 forecasted loss of over $17Bn. The cash burn over the next 4 years is forecasted to be over $100Bn in losses.

It is one of the big questions of our time as to if these foundational LLMs have a path toward the high growth, high profitability we have come to expect from large technology companies. Or are LLM providers a “bottomless pit, attracted by growth when they should have been repelled by it”?

(Post written the old fashioned way - via human. No AI used in this writing)

AI Tags for Content

With the wider adoption of LLMs, more content is being completely produced by AI giving more weight to the "dead Internet theory" where the majority of the web will be created by Bots or AI agents. As this continues I see people valuing purely human generated content more and more.  

Going forward I hope authors, news sources, and content platforms (such as LinkedIn) start putting disclaimers on content to indicate the extent to which AI was used in the piece. This would be similar to how advertisers flag advertisements as ads even if it is native advertising. Authors will likely start doing this voluntarily (and hopefully honestly) to put more weight on a piece. Was AI used to completely generate the piece, to only edit the content, to take the bullet point ideas of the author and make it cohesive, etc.

Hopefully, this will combat the rapid decay of the Internet as more of what we consume online is completely AI generated. 

(Post written the old fashioned way - via human. No AI used in this writing)

My AI Experiments

I use AI daily and try to throw it at every problem or area in life I can. Some interesting use cases I have discovered so far off of the normal path of email (and blog post) editing, etc:

• Identifying the series, artist, and valuation of fine art pieces from a photo

• How to fight and win against my health insurance company when they do something illegal around a claim

• Creating a physical rehab plan for a knee injury

• Having it checking if I am doing prep and cooking write while doing meal prep using the real-time camera

• Acting as an advisor and negotiator for various leases

(Post written the old fashioned way - via human. No AI used in this writing)