The Dependency Atlas ranks eight capital priorities with nine published factor weights. Change those weights and watch the queue reorder — the fastest way to see which conclusions are robust and which hang on a single editorial assumption.
Same scoring function the Atlas build uses, on Dependency Atlas data as of 2026-05-13
Every priority rescored under your weights with the same rounding the Atlas build uses, plus its movement against the published baseline rank.
The largest rank gainers and losers under the current weights, each with the contribution change that drove the move.
The published scoring rules this page replays, and the citations behind the priority definitions.
priority_score = weighted sum of strategic_vulnerability, population_growth, wartime_resilience, import_concentration, local_capability_gap, export_upside, labor_intensity, capital_efficiency, and time_horizon
contribution[k] = round(factor[k] * weight[k]) for every factor except capital efficiency, which uses round((100 - capitalIntensityPenalty) * weight); score = round(sum of contributions, 2); priorities are ranked by score descending.
Sliders set raw weights that are normalized before scoring: each effective weight is w / Σw, so only the proportions between weights matter. When the raw weights already sum to 1.00 — as at the baseline and in every preset — the effective weights equal the raw weights and the baseline positions reproduce the published queue exactly.
Four factors (strategic vulnerability, wartime resilience, import concentration, local capability gap) are derived as importance-weighted averages over each priority’s linked dependencies. Five (population growth, export upside, labor intensity, capital intensity penalty, time horizon) are manual editorial assessments. Re-weighting changes how much each factor counts — it never changes the factor values themselves.
29 sources are cited across the eight priority definitions; the full registry, per-priority citations, and factor derivations live in the Dependency Atlas.
The explorer imports the exact scoring module the Dependency Atlas build uses — integer-rounded contributions, the inverted capital-intensity term, and the two-decimal final score. At baseline positions the output must reproduce the published queue precisely, and the build fails if it ever drifts.
Sliders set raw weights that are divided by their sum before scoring, so the model stays defined for any combination and there is no "correct" total to maintain. Doubling every slider changes nothing; doubling one slider changes everything downstream of it.
A re-ranked queue is a prompt, not a finding. Use each priority's Atlas entry and investment-case article for the underlying citations, and the Capability Gap Tracker to see which priorities have identifiable companies in the research database.
Every composite ranking hides a decision about what matters. The Dependency Atlas makes that decision explicit — nine factors, nine published weights — and this explorer exists so readers can interrogate it rather than accept it. Slide a weight and the queue recomputes with the identical arithmetic the site's build pipeline uses: each factor multiplied by its normalized weight, rounded to an integer contribution, summed, and re-ranked. Nothing else changes. The factor values stay fixed, the citations stay fixed, and the eight priorities stay the same eight priorities. What you are testing is a single question: how much of the published ranking is driven by the data, and how much by the weighting laid over it?
The most useful output is not any particular reordering but the pattern of stability. Some priorities hold their position across nearly every defensible weighting — that robustness is itself evidence, because it means a reader does not need to agree with the editorial weights to accept the conclusion. Other priorities swap places the moment one factor gains a few points of emphasis. Those unstable rankings are not errors; they are honest reports that two theses are close, and that the published order between them rests on judgment rather than on a decisive gap in the underlying scores. For a capital thesis, that distinction is material: a plan built on a rank that survives every preset is a different kind of claim than a plan built on a rank that inverts under mild reweighting.
The presets are deliberately opinionated. Wartime endurance asks what the queue looks like if a long conflict is the design case and peacetime economics barely count. Demographic pressure treats population growth and scarce labor as the binding constraints. Export-led demands that resilience investments pay for themselves abroad, and fast deployment prizes capability that arrives cheap and soon. None of them is a forecast of which world arrives. They are stress tests: if a priority stays near the top of the queue under all five worldviews, the case for it does not depend on predicting the future correctly. Reading the "what moved and why" panel closes the loop — it names the specific contribution change behind each move, so a reordering is always traceable to an assumption rather than left as an unexplained output.
Two structural details deserve attention before quoting any result. First, normalization: raw slider values are divided by their sum before scoring, so only proportions matter and there is no way to inflate every priority at once. Second, the split between derived and manual factors. Four factors are computed from the Atlas dependency data as importance-weighted averages; five are editorial assessments that no weighting can make objective. Setting the manual factors' weights to zero is itself an interesting experiment — the queue that remains is the one implied by dependency data alone — but even that queue inherits editorial choices about which dependencies exist in the dataset and how they are scored. Sensitivity analysis exposes assumptions; it does not eliminate them.
Finally, keep the object of analysis in view. The scores rank abstract capital priorities, not companies, funds, or transactions, and a change in rank changes nothing in the world — it changes a lens. The dataset version and as-of date are printed at the top of the page because factor values are re-derived as source data arrives, and a weighting experiment run against last quarter's data is a statement about last quarter. Treat every configuration of this page as a question you are choosing to ask, and treat the answer as an input to diligence rather than a substitute for it.
The explorer stops exactly where the published data stops. It will not tell you which weighting is right, because no computation can — that judgment belongs to the reader, and the tool's job is to make the consequences of each judgment visible, cited, and reproducible.
Claw & Talon materials are for general informational and research purposes only. Priority scores, rankings, weightings, presets, and re-ranked queues do not constitute investment advice, a recommendation, suitability analysis, offer, solicitation, endorsement, or indication that an investment is available.