Some Guesses about AI in 2026
My current speculations
early “harvest” effects—again, where firms begin to see measurable gains after investing in AI and reorganizing workflows—may be jaggedly emerging at the frontier, even as economy-wide transformation remains incomplete.
He takes a cautious, two-handed approach to the question of how quickly AI will show in productivity statistics. That is wise, but boring. Meanwhile, I have noticed twice in recent weeks that incompetent economic articles expressed with conviction have gone viral. There seems to be a trade-off between competence and virality, in which case I pray that this essay does not go viral.
I will give my prognostications in terms of “targets” and “fades.” In fantasy sports, you “target” a player when you are more optimistic than the market about that player, and you “fade” a player when you are more pessimistic.
Target: Waymo
I think that in the self-driving car race, Waymo is miles ahead of the competition. They are moving ahead with a sensible mix of caution and determination. The cars work well. The business model strikes me as plausible. At some point, people will start to perceive human drivers as a menace. That isn’t going to happen in 2026, but I expect Waymo to convince people that its cars will play a big role in the future.
Fade: OpenAI
OpenAI created tremendous buzz with ChatGPT. Pundits are fond of comparing ChatGPT to Kleenex or Xerox as a brand name that in many people’s mind represents the entire product category. But I think that a better analogy would be Friendster or Myspace. Many of you know that those were once big in the social media space. Others of you may not even know what I am talking about—and that is my point: a brand can be famous now and forgotten a few years later.
The other reason that I am bearish on AI is that I do not trust Sam Altman as a fiduciary. Having read Keach Hagey’s biography, I worry that Altman inherited too much of his father’s floofy profit/non-profit financial notions.
Fade: Domain Expertise
Various professionals and AI start-ups are counting on domain expertise to give them a competitive advantage in the age of AI.
Doctors and lawyers are confident that what they know can never be duplicated by AI. They see AI making mistakes, and they extrapolate that AI always will make mistakes. Instead, I see AI improving rapidly. At this rate, I would bet that by the end of 2026 it will seem imprudent to rely solely on a human for legal or medical advice. (If you’re doctor or lawyer swears that he will never use AI, you will lose trust in him.)
Start-ups trying to write “AI for ___” based on domain expertise in a particular area strike me as deluded. Maybe you can use clever marketing to convince some folks who are otherwise afraid of AI to buy your application because you “know their business.” But only if they are fools.
The best “AI for ___” is probably already a frontier model, rather than a model tuned by a start-up with domain expertise. And if the frontier model is not better today, at the rate that the models are improving it will have bypassed your domain-specific model within a few months.
TARGET: The AI “Keeper-upper”
I have joked that UATX should hire an AI “keeper-upper.” That is someone whose job it is to keep up with developments in AI and decide which ones are ready to be used by students and faculty, which ones are worth watching but are not yet ready for prime time, and which are mirages—fads that will soon disappear.
Seriously, I think that any sizable organization could use a keeper-upper. But this is not someone who just enjoys playing with the latest toys. It is someone a keen eye for practical uses of technology and the ability to communicate with and train others in the organization.
FADE: incumbent organizations
I think that the potential for AI to increase productivity is very high. But I look at faculty at universities, for example, and think that the chances for realizing these productivity gains are pretty low.
To take advantage of AI, you need to be willing to completely re-think your mission and your role. My guess is that the professionals at large incumbent organizations who are most willing to do that are also the ones most likely to leave and strike out on their own. What organizations will be left with are the folks who are inclined toward denial and resistance.
I drive a Tesla and it drives me around 90 percent of the time. It gets better every few months. I don’t think people really comprehend how good it is and how close they are. They have millions of cars on the road doing it right now, sending back real life training data where the car is in shadow mode pretending to drive while you do, and making note when you do something that the car would have done differently. Waymo only has thousands of cars generating this data.
I want lots of companies to succeed in this space! But I don’t think Waymo is the clear winner.
It’s incredibly useful and stress relieving to have a car that does all the hard work in heavy traffic and I won’t go back.
Very brave, Arnold, "Never make predictions..., etc." My analogy is to a combination of the dot.com boom (and bust) and the re-engineering era but faster than both due to the lack of any need to create infrastructure (i.e., buy servers, develop web sites, etc.). Off the top of my head, this means:
1) lots of experimentation before convergence on standards along with emergence of new intermediaries due to radically lowered transaction costs
2) top 2 incumbents per sector survive (e.g., Walmart) but valuation migrates to AI-native new entrants as it did from Walmart to Amazon
3) previously immune knowledge and analytical sectors really change beyond recognition this time (e.g., education, professional services including law, finance).
But not healthcare. Some things never change and the money is in making incisions into old people, especially for orthopods. Long wait lines for specialists means no need to invest -- the chairs in the Doc office waiting room this morning were so worn, they reminded me of the third-world.
Will the AI firms become low margin utilities or the most valuable companies in the world for the next 20 years? Depends on the rate of improvement versus the rate of diffusion of the ability to develop LLMs into non-AI companies -- I'd bet on the rate of improvement and ever greater specialization.
I can't wait to see what AI-native firms actually offer. Perplexity offered itself and peers plus Glean (agents). Does anyone know of any non-obvious examples?