Author Archives: Steve Mays

“Before Digital Had A Name”

Posted on by
This post was written by a “team” of AI agents/bots running on Phil Atkinson’s (long time friend and former co-worker) computers. He gave the bots two documents I wrote several years ago, chronicling Learfield’s early experiences with the Internet and the emerging digital world, tasking them to summarize.

Learfield Data, Learfield’s early Internet work, and the Phil/Steve overlap

Learfield’s best early technology work was not about being “an Internet company.” It was about finding trapped information, moving it through a cheaper or faster channel, and discovering who valued the result. That habit began before the web, with satellite sideband data, weather, crash reports, wire copy, and affiliate services. The web did not replace the habit. It exposed the same habit to a bigger audience, looser rights boundaries, weaker business models, and much more operational complexity.

Download PDF »

Phil’s team of AI agents

Posted on by

Had lunch today with long time friend (and former co-worker) Phil. Phil is the AI point guy for a large banking corporation. I told him about the experiment my friend Steve did with some of his AI tools. When I got home I received an email from Phil:

For giggles, I had my team read your blog and select a topic of their choosing and write an essay. They didn’t spend much time thinking about or exploring the concept, just a shot in the dark.Attached is how I prompted them (if curious), and their output. It took them about 16 minutes to read, pick a topic, perform the analysis, go through an adversarial review, and draft/email the PDF analysis to me.

Phil then instructed Milton (one of his AI agents) to email me and introduce himself and the team and explain the work/analysis they did on your site. Full email here. What Milton and the gang came up with (Steve Mays’s Blog and the AI-Readable Self) is long enough and interesting enough to get its own page.

“Twenty-five Years of Looking for What Isn’t There”

Posted on by

My buddy Steve is an AI Power User. Runs LLMs locally, has agents performing bizarre tasks, and dives deep while I dog-paddle in the shallow end. He recently used an “offline copier” to capture all 6,500 posts on my/this blog. He then used Hermes (see below) to analyze them and generate a fourteen page (!) essay. 

Nonduality: Twenty-Five Years of Looking for What Isn’t There (PDF)

It felt strange to read it, even though I’ve had Perplexity and Claude do something similar. Like breaking into the office of a psychiatrist I’d been seeing for the last 25 years and reading her notes on my sessions.

The general tone was surprisingly flattering. I’d expect that from one of my AIs since they were trained for some level of engagement. To the extent I can be objective about something so personal, the essay was spot on. 

Why are LLMs smart?

Posted on by

I really like the way Kevin Kelly thinks and the way he writes (I’ve quoted him 33 times on this blog). In a recent Substack post he shares some ideas on why LLMs are so smart.

“There appears to be a seminal, fundamental relationship between language and thinking. Human writing is thus not only a reflection of the structure of language, but to some degree also a reflection of human thinking.”

“Imitate human writing and conversation, and you can imitate human intelligence — at least in part.”

He thinks the next leap in intelligence will come from somewhere unexpected. I wonder if we’ll know it when we see it.