In late 2016, I wrote a short book called The Real Internet of Things. The core argument was simple: IoT isn't about smart fridges. It's about a fundamental shift in the relationship between humans and technology.
The book made a series of predictions based not on where technology was heading, but on what humans actually want. I started from evolutionary psychology and desire rather than chip roadmaps and protocol specs.
The main predictions were:
I also laid out nine directional trends: centralized to peer-to-peer, forced to natural, obvious to invisible, manual to automatic, periodic to continuous, private to open, visual to multi-sensory, aggregated to curated, and designed to evolved.
It's been nearly a decade, and something interesting has been happening. People are going back to the book and noticing that much of it reads like a description of 2025-2026.
Security researcher Joseph Thacker wrote an entire analysis of the book last year. His assessment was blunt:
The thing Thacker kept coming back to was the timing. I wrote this six years before ChatGPT. The concept of "Digital Assistants" managing APIs on your behalf -- which I described in the book -- is essentially what we now call AI agents.
Thacker specifically called out the interface shift prediction -- the move from humans learning technology to technology learning human interfaces -- as "exactly what happened with ChatGPT." He also noted that the gig economy predictions were "spot on," and that the nine directional trends were "VERY directionally correct for where the tech space is going."
The book sits at a 3.97 out of 5 on Goodreads, with 69% of readers giving it 4 or 5 stars. What's notable is that the ratings and reviews have become more positive over time, not less. That's unusual for technology books.
Other Amazon reviewers have described it as "clear, thought provoking, and original" and noted that it "paints an interesting picture of a future that probably isn't too far away." One reviewer wrote that I "take a refreshingly pragmatic and simple approach to setting out the future role of technology in our lives" and that "the results are startling."
Not everyone loved it. Some found it too short, and fair enough -- it's a quick read with micro-chapters. One reviewer wanted deeper treatment of topics like mesh networking and protocols. That's valid criticism. I was painting with a broad brush, not writing a technical spec.
I think the reason these predictions aged well is because I didn't try to predict technology. I tried to predict desire.
Technology predictions fail when they extrapolate from current capabilities. They succeed when they extrapolate from what humans consistently want across all eras: less friction, more autonomy, better information, less time wasted on things that don't matter.
If you start from the premise that humans will always want technology to understand them rather than the other way around, you end up predicting something that looks a lot like LLMs and AI agents. Not because you're a genius, but because you're reading the demand signal correctly.
The book didn't anticipate the specific mechanism. I didn't predict transformers, attention mechanisms, or the scaling laws that made LLMs possible. I was thinking more in terms of narrow AI getting gradually better at specific tasks, not a generalized intelligence leap.
I also underestimated how fast the interface shift would happen. I expected a gradual transition over decades. ChatGPT compressed that into about 18 months.
And the AR predictions are still mostly unrealized. We're closer with Apple Vision Pro and Meta's work, but the ubiquitous AR overlay I described in the book remains further out than the DA predictions.
The book's vision of Digital Assistants managing APIs on your behalf is basically the pitch of every AI agent startup in 2026. The concept of "daemons" -- standardized APIs for every entity -- maps neatly onto the current push for tool use, function calling, and MCP (Model Context Protocol).
I'm building this exact future right now with PAI (Personal AI Infrastructure) -- a system where my own DA manages interactions with dozens of APIs, services, and data sources on my behalf. The book was the vision. PAI is the implementation.
If you haven't read it, the full text is free on my site. It's a short read. And if nothing else, it's a useful case study in what technology prediction looks like when you start from human desire instead of capability curves.