Project status — July 2026: Fathom is a completed research project. Its pre-registered test has been run and its results are documented across these pages. For the findings, see the Results and Conclusion.
Fathom is a research project that set out to measure something most people assume cannot be measured: how much of a system’s own past is at work in its present.
It began with an essay. In Consciousness and Complexity, Roger Smolski sets out a deliberately simple view of what consciousness is.
A copy of the complete essay is not published here, but is available by request, via the email address on the About page.
He begins:
“Consciousness is what processing complexity, of a particular kind, already is. It is not produced by that complexity as a separate thing; it is that complexity, seen from the inside. The more of it a system has, the more conscious it is.”
The essay is careful about what it means by complexity. Not size, not the number of parts, not the sheer quantity of activity, but something narrower:
“Processing complexity is how far a system draws its own past together into its present. What counts is integration, not accumulation: how much of a system’s history is brought together to bear on what it does next, rather than how much of it is simply stored.”
From this the essay makes a prediction, and stakes the most on it:
“There will be no sharp break in nature between the systems we are sure are conscious and those we are sure are not. If neuroscience found a single neural feature that split the animal kingdom cleanly into the conscious and the rest, with nothing in between, that would count against a continuum and for a threshold, and I would have to give up the account.”
The essay is equally honest that the tool needed to settle this is missing:
“The measurement problem is real and unsolved. But an unsolved measurement problem is not an untestable theory.”
Fathom is an attempt to build that missing tool, and to point it at the hardest case the essay names.
The quantity, and how to measure it
The essay’s definition, history drawn together into the present, has a candidate measure already waiting in the study of complex systems. Fathom’s instrument, called ΔAIS, asks of any system a single question: how much does its recent past predict its present, once you subtract what you would expect from a system with the same surface statistics but no genuine memory?
That subtraction is the point. Plenty of systems look structured without truly using their past. The essay gives the example of a river, which “carries its whole history in the silt it moves and integrates none of it.” A naive measure would be taken in by every such case. ΔAIS is built to see past them. It compares a system against careful imitations of itself, versions that share its statistics but not its genuine memory, and reports only what survives the comparison.
What the instrument had to prove first
An instrument nobody has calibrated is worth nothing, so Fathom began with systems whose answers are already settled. A pattern-generating rule known to perform real computation scored high. Random noise scored zero. A chaotic, weather-like signal, intricate to look at but never actually using its past, also scored zero, which is correct. Of the several candidate measures tested against these known cases, only ΔAIS drew the lines in the right places. That is what earned it the right to be aimed at harder questions.
From the known to the unknown
With the instrument calibrated, Fathom worked towards the case the essay cares about most. The essay predicts a continuum with no sharp break, and expects artificial systems, as they grow more complex, to sit somewhere real on it. In its ordering a large language model sits above a chess program and below a mouse.
So the decisive test took a system that genuinely learns, a language model watched across its whole training, from a random starting state to a finished and competent one, and asked the instrument to measure processing complexity at every stage. Does it switch on sharply, or rise smoothly, as the essay predicts?
The answer, and what it does and does not settle, is on the Results and Conclusion page. In short, the instrument found a limit. Not a limit in the theory, but in what this kind of measurement can currently reach, and where that limit falls is the most interesting thing Fathom found.
What Fathom is, and is not
Fathom is a personal research project by Roger Smolski, not a commercial product. It makes no claim to have measured consciousness directly. It measured a proxy for the essay’s quantity, and that quantity is itself, in the essay’s telling, the outward face of something with “an inside, a point of view, which sets it apart from any ordinary physical property.” Fathom measures the outward face. The inside stays, as the essay freely admits, beyond the reach of any third-person description we currently have.
The project is complete. Its central question was tested and answered, and the answer is set out plainly across these pages.