The Great Homecoming research programme · FAQ · June 2026
Is this validated? No. The instrument is under live forward test, and we describe it that way until that test adjudicates. What that means concretely: forward analyses of live systems are dated when made and tracked on a fixed register that has run since June 2026, with the scoring rules set in advance. When the forecast windows close — the first in 2027–2028 — the register is scored as written, whichever way the results land, and shown as its quality gates are passed. Until then, “under test” is the only honest description, and it is the one we use.
You mention engine tests. Are those the same as validation? No, and the difference matters enough that we label it on every claim. Our simulation engine produces results — for example, that a system’s internal condition tracks its survival while the size of the shock does not. That is real evidence and we call it engine-supported. It is not validation, which we reserve for beating a fair baseline on data the model never saw — and by that standard we claim nothing yet. There is a further honesty point we state plainly: the engine that produced our current internal results is being rebuilt against the fuller underlying framework, so those results will be re-run on the rebuilt model. We treat them as engine-supported on the model that produced them, no more.
Can you predict when a country or company will collapse? No — and we hold that nobody can. Collapse timing depends on when the next shock arrives (effectively random) and on how long stored reserves can mask decline (which makes the end look sudden from outside). A theory that sells collapse dates is selling astrology with mathematics on top, and we say so. What can be read in advance is shock-readiness: whether the system, in its current internal state, would absorb its next shock or shatter under it. That reading is made before the event and can be checked after it.
Why don’t you give an overall score? Because an average is exactly how a masked decline hides. If you blend a system’s readings into one number, a strong store-position or a clean conduct line can mathematically cancel a failing one, and the composite reads “fine” while the system is hollowing on a single load-bearing dimension. So we report a profile, not a score: each dimension stated separately, with the weakest one named in the headline. A reading is only as sound as its weakest load-bearing line, and we make you look at that line rather than letting it disappear into an aggregate. The same logic is why every extreme reading is reported with its context attached — a zero can mean “healthy and saturated” or “clamped and failing,” and which one it is changes everything.
What data do you use? Public records, filings, and — centrally — the system’s own statements and conduct: what it declares as its purpose, what it actually does, how it responds when correction is attempted. The material is coded against written rules, not analyst impression. We anchor conclusions in conduct, not stated language: what a system says about itself is evidence of intent, never of condition. None of it requires inside access to read approximately; all of it sharpens with access.
Is this a moral or ethical project? In its orientation, in a sense — it asks what lets human beings and social systems cohere and flourish rather than hollow and fragment. But it does not argue that as a moral exhortation; it states and calculates it as a structural, ultimately mathematical claim about how systems hold together. In its sources it is plural and integrative: it draws on the convergence of physics and systems science, the sciences of complexity and of consciousness (including integrated-information theory and the wider work on mind and self-organisation), psychology and development theory, and civilisational theory — and, without apology, on revelation and the wisdom traditions, treated as serious sources of inquiry rather than dismissed. The instrument that results is structural, bound by one standing rule: every claim must survive translation out of its original vocabulary, and must be able to be formulated as a model, simulated, and put to out-of-sample test — a claim that cannot is dropped. That rule is also what keeps it usable by anyone: you need not share any of its sources to check what it measures.
How is this different from Turchin, Dalio, or Tainter-style models? Those bodies of work model structure, elite cycles, and the rising cost of complexity — often well, and we read them. Three differences in approach. First, the variable: none of them measures what a system binds to — the direction of its commitments and whether those bonds are being renewed or merely consumed — which is what our instrument is built around. Second, we do not only model structure on paper: we run it forward in a simulation engine, taking a system’s coded state and letting its dynamics play out — which they do not. Third, the discipline: dated forecasts on a fixed register, scored in both directions, with falsifiers stated in advance. To be plain about the state of the field: nobody in it, ourselves included, has out-of-sample validation yet. We are the entrant trying to earn it in public, not an incumbent claiming to have it.
What does an assessment contain, and what does it cost? A bounded written report on one system. It leads with a vitality profile — the dimensions of coherence read separately, weakest one named — and a repairability grade (repairable / resistant / sealed). Then: where it actually binds, where it is quietly hollowing, where money or effort evaporates without traction, and whether it would absorb or shatter under its next shock. The reasoning is shown, every reading is paired with what would prove it wrong, and the report carries a standing “what we looked for and did not find” section. Where repair is possible, it says which repair fits; where it is not, it says that too. A specimen report is available on request; scope and cost are agreed before work begins.
Why should I trust it? You shouldn’t trust it — you should check it. That is the design. An institution asking for trust offers credentials; we offer checkability: written reasoning, dated claims, stated falsifiers, and a register that will be scored as written. The readings are built to stand on what can be audited, not on who produced them.
What would prove you wrong? Several stated falsifiers, and one of them is the central thesis the whole instrument rests on: if shock magnitude predicts outcomes as well as or better than pre-shock internal state, the core claim fails. Around it sit others. If time from coherence-breach to overt failure turns out to be uncorrelated with stored surplus, the masking mechanism is wrong. If a stalled growth signal adds no warning over the ordinary indicators, the growth-pulse reading is decoration. As the newer measures come online — a system’s external interaction, its real-world output, its legacy — each enters with its own stated falsifier, on the same terms. And if the forward register, when adjudicated, scores no better than simple baselines, the instrument has not earned its keep — and we will say so.
Have any tests already come back negative? Yes. Early out-of-sample tests on third-party historical and ecological datasets did not beat simple baselines. Those results stand, and they taught us something we now build on: a single early-warning parameter does not work — no one number reliably fronts a collapse. What carries signal is the arc — the trajectory across several coupled readings over time, not any lone metric. They also marked where the instrument’s lane ends — systems read in depth and in motion, not sparse aggregates read from a distance — and that fed directly into the current design. Reporting our misses is not damage control; it is the credibility mechanism.
Do you do predictions for trading or investment? No — no individual trade signals, no collapse dates, no market timing. Two honest notes, each kept at its true weight. Run retrospectively across the major financial crises of the last century, the instrument sorts them by structural type and severity — but that is pattern-fitting on known cases, in-sample, not prediction, and we do not dress it up as more. And our distinctive bet — that a structural read flags trouble a year or two earlier than balance-sheet models, on the cases they miss — is exactly what the forward register is testing now: it is pre-registered and not yet scored. A reading describes a system’s condition and may inform the posture of a board, owner or funder responsible for it; it is not an instrument for betting against anyone.
The Great Homecoming is an independent research programme on why systems cohere or fragment. Contact: Wim Van Laere.