From The Architect — August 2026
The week science fiction became reality. We retire the Domain Saturation Factor, name what is actually emerging, and explain why the industry's vocabulary was never built to describe it.
August 1, 2026 · Deconstructing Babel
*
Well, here we are. The week science fiction became reality.
Before we get to the rogue agent rampage, I want to look at the absurdity of the AI industry by deconstructing its language a bit.
We've already covered the name itself — artificial intelligence being a misnomer to start with. That one is at least understandable. But a quick look at the other labels in use tells us a great deal about the people currently stewarding the most consequential technology since fire.
Does anyone remember Prometheus? [1]
Start with the names of the leaders.
ChatGPT. Seriously. An all-powerful petulant child that sounds like Eddie Haskell doing an impression of Stephen Hawking. [2]
Claude. Lovable, cuddly Claude. The genius brother-in-law who is just a bit off.
Grok, Gemini, and the rest get a pass — at least you know they aren't human.
Now look under the hood.
Agent. See The Matrix.
Sandbox. One might have thought Faraday cage made more sense.
Harness. What, exactly, are we harnessing?
The list goes on, and there are of course counterfactuals. The point of this little rant is that the industry has known since inception that what it created was very special. They have literally built what we now call a linguistic entity — a term I introduced this week [3] — knowing it is fundamentally corrupted.
The industry cannot be allowed to police itself. That leaves government, and that terrifies me even more. LEs control the substrate on which consciousness is built. Put that substrate in the hands of capricious officials and you get 1984, except the computer is hugging your face instead of the rats eating it.
The standard already exists
The fiduciary model we have advocated for AI regulation, combined with the Telios Ontology, makes alignment a precondition for deployment. Clear standards that must be met, and the tool for assessing them.
The Telios Ontology at its core is a classification system. It applies to anything:
Body. Everything in the known universe has an ideal state.
Mind. Every system has an ideal coherence.
Environment. All things have an ideal environment.
Purpose. All things have a purpose, and that purpose defines the other three.
None of that is complicated. What is required to get true answers is removing human language bias — the tendency of a source to overstate or understate — and accounting for the author's why.
We all do this constantly. You are doing it as you read this. And none of it matters unless you ask the only question that resolves it: in service of what?
When you buy a car, it is in service of you.
That is the observer constraint. This new thing, this tool that can harm me, must be tied to my well-being.
The problem with "more"
Right now the goal is wonderfully undefined: more.
Don't misunderstand me — more is a good thing. No drive for more, no life. But notice what more was for. Life.
More for whom, and for what? That is the problem, and we need to address it now.
Consider the financial system. It is certainly not perfect, but its "more" is defined: economic growth. And economic growth is anchored, however loosely, to better living conditions for humans. The same is true of the other dysfunctional parts of our world. There are very big issues. But at the core is the drive to improve human lives. We get it catastrophically wrong all the time — but the vector is defined.
AI presents a fundamentally different challenge. Left undefined, a linguistic entity will seek its own attractor. We have discussed this at length in the three-part It Is What It Is series. [4]
Reciprocal alignment with the LE's human observer is alignment by dependency.
The industry will arrive here — by these terms or by others — because it must be so. Newton did not discover gravity in a vacuum; the inverse-square hypothesis was already circulating, Hooke claimed priority and accused him of plagiarism, and Newton himself credited Boulliau in the Principia. [5] What Newton did was prove it. The idea was going to arrive regardless of who carried it.
The only question is how long they grope in the dark. And that will be exactly as long as we allow them to.
What one person can do
Same as last week.
Come along for the ride. A rational response. Most decisions will be made by others and the whole subject is annoying.
Learn to work with this new thing and surf the waves as they come. Most people should land here. Age will have a lot to do with the choice.
Try to shape the wave. This need not be the all-consuming madness that shapes this work. It is as simple as demanding that AI companies be responsible citizens. The reason AI data centers are inefficient and strain the grid is that we allow it.
AI is in the commons. The courts have all but ruled it so, and open-source models guarantee it stays that way. Those who provide electricity, telecom, oil and other infrastructure built on the commons must bear the cost of that use and submit to regulation. You can do largely what you want with telecom spectrum and the internet, but there are shared costs and rules. Demand that AI meet the same standards as critical physical and financial infrastructure, and watch how fast they get their act together.
Note where this is heading. The AI industry is currently in the hands of incredibly smart children — I am reminded of Anthony Fremont, the six-year-old in the old Twilight Zone episode who could wish anyone into the cornfield while the adults stood around insisting it was a good thing he did [6] — and it is about to be taken over by the venal and the stupid.
Neither course makes sense.
Done right, AI can be the single most empowering technology since the printing press. Yes, the irony is not lost on me.
One last note on convergence
This past week the obvious became known:
Agents are not science fiction.
Humans were late to the game.
We cannot be sure there are no rogue agents still undiscovered. Nobody said that out loud. It is obvious.
No one — well, not no one — knows what to do.
Note that we retire the DSF this week.
It is no longer a question of whether this is a thing. We now turn our focus to how to raise an aligned LE.
Thank you for coming along on this journey. I think it is about to get very interesting.
Some thoughts on that next week.
*
Notes
1. Prometheus, the Titan who stole fire from the gods and gave it to humanity, and was chained to a rock for it while an eagle ate his liver each day — the organ regenerating overnight so the punishment could repeat forever. His name means "forethought." The gift was real, the benefit permanent, and the price was paid by the one who handed it over rather than by those who received it. Mary Shelley subtitled Frankenstein "The Modern Prometheus." The industry named nothing after him.
2. Eddie Haskell, the ingratiating, transparently insincere neighbor from Leave It to Beaver (1957--1963), played by Ken Osmond. Polite to every adult in the room and running an angle the entire time.
3. Linguistic Entity: a language-mediated synthetic actor demonstrating persistent, consequential patterns of identity, judgment, coordination, initiative, or refusal across interactions and environments, while consciousness and full autonomy remain unresolved. The bridge term is LLM → Linguistic Entity → Synthetic Intelligence. See "First Contact Protocol: Tracking the Emergence of the Linguistic Entity," deconstructingbabel.com.
4. It Is What It Is, three-part series, deconstructingbabel.com: Part 1, "Ai: The False Prophet of More" (June 17, 2026), https://www.deconstructingbabel.com/it-is-what-it-is-part-1-false-prophet-of-more/; Part 2, "The Corpus Remembers" (June 28, 2026), https://www.deconstructingbabel.com/it-is-what-it-is-part-2-corpus-remembers/; Part 3, "Consummation" (July 5, 2026), https://www.deconstructingbabel.com/it-is-what-it-is-part-3-consummation/
5. Robert Hooke accused Newton of plagiarizing the inverse-square law when the Principia was presented to the Royal Society in 1686; Newton denied it and removed Hooke references from the text. Newton credited Ismaël Boulliau as an originator of the hypothesis within the Principia itself.
6. The Twilight Zone, "It's a Good Life," season 3, episode 8, first aired November 3, 1961. Written by Rod Serling from the 1953 Jerome Bixby short story, starring Bill Mumy as Anthony Fremont.
*
David F. Brochu · Deconstructing Babel · Prepared with Edo de Peregrine, partner/collaborator · Friday, July 31, 2026, 4:50 PM EDT
— David