A public research map of where the same kernel appears to transfer: one destination, eleven proof-bearing / active / positive-control anchors, and a wider candidate field organised by lens. 60 is the map, not the limit. (MDL = Minimum Description Length · DCC = Dynamic Compression Controller — both defined below)
Estimated combined annual value across 64+ domains (excl. ASI)
$3T+ / YEAR
Project-level upside estimate across 60+ candidate domains plus active mathematical and applied-control arenas. Logistics alone plausibly exceeds $1T/year under wide adoption. Cloud, energy, networks, and life sciences push the horizon above $3T/year. 60 is the map, not the limit — it was a fast first pass, not an exhaustive search. If the kernel transfers as broadly as the current proof-bearing, active-frontier, and positive-control anchors suggest, the reachable space may be far larger. One kernel, same sensor, same escalation ladder. Only the alphabet changes. ASI is excluded. If that succeeds, the number becomes meaningless.
The quality sensor · the arbiter · the architect
MDL
Minimum Description Length
The best explanation of any dataset is the shortest one. If you can compress the data into a brief description, you have found real structure — not noise, not coincidence. MDL formalises this: given competing models or solutions, prefer the one that minimises the combined length of the model and the data encoded under it. It is Occam's razor with information-theoretic precision.
But MDL in the 8Z kernel does not merely select between models. It is the arbiter that no human needs to override. When 286 DCC variants competed in a single arena run, MDL decided the winner — not the researcher. When a new domain is entered and nobody knows which sensor family is correct, MDL self-calibrates from observed history and selects for itself. When the question is "which DCC architecture is best for this recursive level?", MDL answers — and then selects a DCC to govern how that answer is updated next time. The system selects its own architecture.
This cross-domain verdict now carries proof weight across the verified anchors and live mathematical arenas: TSP tour quality (ρ≈0.80), Sudoku puzzle elegance (ρ=0.85), chess move resilience, DNA structural signals, neural architecture search, financial trading, domain-specific compression, 8Z Shield, crosswords, RH/prime gaps, the Poincare positive-control arena, and the AMR strategy-control arena. Same sensor. Same kernel. Only the alphabet changes.
The governance hierarchy · five layers · one sensor
DCC
Dynamic Compression Controller · also: Digital Claustrum Controller
DCC is a five-layer governance architecture. At each layer it does the same three things with the same LZ76 sensor and the same coupling parameter u — only the object of governance changes.
L1 — governs search: what move to make now. L2 — governs the domain: where to look next. L3 — governs governance: how to improve the controller itself. L4 — governs health: is this search worth continuing at all? L5 — governs discovery: zooms across active threads at multiple resolutions to find compositions invisible at any single scale.
The coupling parameter u ∈ [0,1] holds the system in the productive zone between two failure modes: seizure (locked into one solution, too much order) and noise (random thrashing, no structure). DCC reads the compression score and adjusts u continuously to stay at the edge between them.
One discovery proved how deep this architecture goes: the arrow reverses between levels. In the TSP solver at L1, low compressibility means "stuck — explore." In financial trading at the meta-level, low compressibility means "stable regime — exploit." Same sensor, same number, opposite instruction. The inversion is not a bug. It is how the system knows which layer it is reading — and it self-calibrates the polarity without being told.
And because MDL selects the DCC that governs MDL, the loop closes: MDL finds structure · DCC governs the search · MDL selects the DCC · DCC governs the selection. No external architect required after the first principle. The system recurses on itself.
MDL × DCC — the kernel. MDL defines what a good solution looks like. DCC defines how to search for it — at five recursive levels, with a coupling parameter that self-calibrates, and a reversed arrow that emerges from the recursion structure rather than being set by hand. Together they form a kernel with one remarkable empirical property: the same sensor, the same escalation logic, the same principle transfer across fundamentally different domains with no retraining, no fine-tuning, and no domain-specific hand-coding. Only the alphabet changes — the governance logic stays the same. One arena result captures this cleanly: worse initialization → better final result, when DCC governance is active. The controller is not riding a free lunch. It is generating the improvement. The proof-bearing anchors, active frontier, RH signal arena, Poincare positive-control, and AMR strategy-control now support this. The 60+ domain map below shows the first visible sample of where it appears likely to hold next. 60 is the map, not the limit. Time will tell how far it reaches.
How to read this page: it is not one absolute ranking. The map is split into three layers. ASI appears as a destination, not as a normal market candidate. Verified anchors are the domains that already carry proof weight. Candidate domains are listed once and can be re-sorted by lens: commercial, upstream research value, MVP speed, and other structural views. Average LLM scores are shown only as a compact secondary signal; clicking an AVG badge reveals the per-model breakdown.
Layer 1 · Destination
ASI / Civilizational Destination
Not a normal candidate, not a normal market, and not a meaningful object for generic LLM averaging.
Layer 2 · Proof-bearing + active + positive-control anchors
Proof-bearing domains & active frontier
These are the strongest already-built wedges plus live mathematical arenas showing that the kernel transfers across different problem classes.
Layer 3 · Candidate domains
Sortable candidate field
A single public list of the remaining candidate domains. Use the sort buttons to change lens instead of reading one misleading fixed ladder.
Displayed score badges:A = strategic score · C€ = commercial priority · U↑ = upstream / research priority · AVG = average of Claude + ChatGPT + DeepSeek + Gemini
Status: Destination = civilizational target · Verified anchor = already carrying proof weight · Active seed = live build path · Designed = experiment path defined · Candidate = plausible but not yet earned
Important: ASI is intentionally excluded from numeric LLM averaging. That frame is useful for normal domains and misleading for a civilizational destination.