This is going to be mercifully short compared to last week’s 5500 essay on 1873, 2000, and 2008.
I think there are three epochs in the medium term regarding AI and so far we have largely only been thinking about one of them:
Epoch 1 - LLM Era
Epoch 2 - Agentic Era
Epoch 3 - Physical and Robotic Era
Epoch 1 - The LLM Era - NOW
I don’t need to explain the LLM era by now. You are living in it.
Epoch 2 - The Agentic Era - Early 2026 to the indefinite future
Perhaps you have some awareness of the agentic era of AI. In this era, AI functions like an employee(s) that you give a set of directives and permissions. It checks in periodically while it repeats various tasks. People have been doing this in earnest since at least January of 2026 when OpenClaw hit tech nerds like a ton of bricks. Since then there have been numerous agentic platforms springing up. The agentic era has yet to be made “Apple simple” for the common man, so this epoch has both been inaugurated but is still largely sequestered to maybe 1-3% early adopters. There are many who build not just agents but teams of agents and sometimes whole departments of agents (marketing department, legal department, graphic design… etc.).
The power users of agentic AI in my life get between 1-4 weeks of work done per day. Most of them are solopreneurs or work in smaller organizations. Agentic AI is a lot easier to deploy the smaller your company or organization is because you have far less rules for compliance, security, and red tape. Those things being said, I have observed three things, to one degree or another, about the power users of agentic AI in my life:
SLEEP DISRUPTION - They lose some sleep periodically to get up in the middle of the night to check in in on their agents and give them new permissions or tasks so they don’t waste down time while they aren’t working.
BASELINE ANXIETY - I have observed a slight uptick in baseline anxiety in these people. Whether or not they are cognizant of that is another question.
DOPAMINE ADDICTION - There is a level of dopamine addiction that comes from insane productivity and the reward cycle of building things so rapidly.
I don’t think these dynamics mean, “don’t do this.” These are just observations that I have had from early adopting power users of agentic AI. Personally, I haven’t played around with it for three reasons - time, learning curve, and I believe I lack the maturity to handle putting the ring of power on my finger. So, what do I know, and you probably better not take my word for it.
Epoch 3 - Physical and Robotic Era - Late 2027 to the indefinite future
The thing hardly anybody is talking about is the era when you see 30 cybertaxis on your commute to work (if you still have one) and people start having Sheldon Cooper’s in their home scrubbing their toilets and folding their laundry.
Do I want something approaching AGI in my home with toilet brush in hand?
Do I want something that smart walking around my home? Do I want it talking to my kids? Does this thing have good enough cybersecurity on it to keep it from the terrorist hacker that wants to hack it and give it the skill find the sharpest kitchen knife to attack me with in my sleep? I’d kinda like to know because I bought the Chinese model that was 70% less than that newfangled Tesla bot.
I’d like my tombstone to read something besides, “Mr. Cooper killed Mike with the kitchen knife in the master bedroom.”
BAZINGA!
In all seriousness, I say all this stuff to say that we aren’t remotely prepared for thinking about the ethics, parenting, or cybersecurity around epoch 3.
Consider this my personal plea for more than just John Dyer to be thinking about epoch 3.
Other stuff to worth reading or listening to:
Even Claude Is in the Dark About Dario Amodei’s Wife—and Her Influence at Anthropic - WSJ
AI and Emotion: Can Machines Feel? (Christopher Watkin interviews MIT’s Rosalind Picard)