Coverage

299 of 318 slots covered — 94%. That is 293 standalone topic briefs plus 6 topics covered by a hub-and-spoke programme. Every figure on this page is computed from the topic map at build time; none of it is typed by hand.

One slot is counted once, not twice. Synthetic Ecosystems is deliberately listed under both Biology and Climate & Planetary Engineering — two genuine framings of one subject — so the 318 slots span 301 distinct topic names. The brief was written under the biology framing, so the climate slot links to it and is marked cross-listed rather than counted again. Counting it twice would overstate the work.

All 318 topics

Filter by status or category, or search by name or code. Codes are locked to map position: FR-VII-09 is the ninth topic of Category VII, whether or not it is written yet.

Beyond the topic map

Frontier Research is one of three instruments. These are the others, indexed here in full.

Learning modules (21 pages)

Foundational modules with flagged prose, checkpoints and reading lists.

• What an atom actually is M-Chem-01
• Chemical bonds — three flavours M-Chem-03
• Why carbon runs biology M-Chem-05
• DNA, RNA, and the central dogma M-Bio-04
• CRISPR and editing in vivo M-Med-02
• How electrons are arranged M-Chem-02
• Quantum entanglement M-Phys-01
• Reading the genome (and why it's not enough) M-Med-01
• Goal-directed behaviour without consciousness M-Theory-02
• Higher dimensions M-Phys-02
• Lipid nanoparticles and targeted delivery M-Med-03
• Real molecular machines M-Tech-01
• Directed panspermia, taken seriously M-Bio-03
• Prions — propagating a shape, not a sequence M-Bio-01
• Programmable matter M-Tech-02
• Quorum sensing: coordination without a controller M-Bio-05
• The virtual cell: modelling life in a computer M-Med-04
• Viruses and the hijack pattern M-Bio-02
• Von Neumann machines M-Theory-01
• The weak forces that run biology M-Chem-04
• Traversable wormholes M-Phys-03

S-001 · AIHS Feasibility Study (33 pages)

The Institute's flagship hub-and-spoke study of an Autonomous Integrated Healing System.

• Autonomous Integrated Healing Systems S-001
• A1: In vivo single-cell readout A1
• A2: Whole-body molecular mapping A2
• A3: Continuous temporal sampling A3
• A4: Integrated state representation A4
• A5: Distributed pathogen and damage detection A5
• B1: Complete variant effect prediction B1
• B2: Real-time epigenomic state inference B2
• B3: Causal disease modelling B3
• B4: Personalised digital twins B4
• B5: Continuous learning across patients B5
• C1: Multi-target coordinated intervention C1
• C2: Universal tissue access C2
• C3: Architectural tissue reconstruction C3
• C4: Neural reconnection C4
• C5: Bounded reversible molecular agents C5
• C6: Multi-layered safety architecture C6
• C7: Accelerated healing energy supply C7
• Bucket A: Diagnostic Capability S-001
• Bucket B: Interpretation S-001
• Bucket C: Therapeutic Delivery S-001
• Decision Gates S-001
• Phase I — Foundation S-001
• Phase II — Capability Building S-001
• Phase III — Integration and Scale S-001
• Phase IV — Maturity S-001
• Advance Atlas S-001
• Glossary S-001
• Integration Challenges S-001
• References & Further Reading S-001
• Risk Assessment S-001
• Phased Research Roadmap S-001
• Safety Architecture S-001

P-001 · The Protomolecule Curriculum (7 pages)

A layered programme working from non-living replicators up to higher-dimensional physics.

• From first principles to the protomolecule Overview
• Non-living replicators P-001 Layer 2
• Self-replication and direction P-001 Layer 3
• Nanotechnology and programmable matter P-001 Layer 4
• Information at a distance P-001 Layer 5
• Ring builders and higher-dimensional physics P-001 Layer 6
• A protomolecule engineering roadmap P-001 Layer 7

P-004 · Engineered Origins (4 pages)

Synthetic biology, directed panspermia and designer organisms, treated as one problem.

• Engineered origins Programme P-004
• Designing organisms to specification P-004 Part 3
• Seeding life across space P-004 Part 2
• Building life from parts P-004 Part 1

Where the map is not finished

4 of the 10 categories are closed. The 6 that are not appear here with the reason each is short, because a coverage percentage tells a reader how much is missing and nothing about what.

Category V · Energy Systems — 21/24. 3 slots built under another category and deliberately not counted twice — Vacuum Energy Concepts, Planetary Power Grids, Beamed Energy Systems.

Category VI · Climate & Planetary Engineering — 26/28. 2 slots built under another category and deliberately not counted twice — Future Food Systems, Synthetic Ecosystems.

Category VII · Civilization-Scale Infrastructure — 31/37. 6 slots built under another category and deliberately not counted twice — Ocean Colonization, Lunar Infrastructure, Planetary Communications Infrastructure, Water Transfer Infrastructure, Arctic Trade Routes, Space Manufacturing Networks.

Category VIII · Governance & Institutions — 26/29. 1 slot built under another category and deliberately not counted twice — Multi-Planetary Governance; 2 published as ongoing research, which this map counts as a page and not as a covered slot — Public Policy Foresight, Civilizational Planning.

Category IX · Economics & Society — 25/26. 1 slot built under another category and deliberately not counted twice — Space Economies.

Category X · Historical & Frontier Science Studies — 25/29. 4 slots built under another category and deliberately not counted twice — Future Studies Methodologies, Future Scenario Planning, Emerging Technologies Assessment, Scientific Paradigm Shifts; 6 covered by spokes of P-006 rather than by standalone briefs (Bell Labs and Frontier Research, DARPA Case Studies, Apollo Program Lessons, Manhattan Project Lessons, Failed Technologies, Research Bottleneck Analysis).

The tech tree Below threshold

A technology tree — which topics genuinely depend on which — is the natural companion to this index, and it is not published yet. What follows is the condition it is waiting on, stated as a number, because “planned” on its own tells a reader nothing about whether the wait is a week or a rewrite.

The rule, fixed in advance. The tree ships when all three of these hold. They were written down on 26 August 2026, before the adjudication that moves them began — which is the only property that makes a threshold a rule rather than a description of whatever happened.

1. Adjudicated coverage ≥ 60% — of the briefs that have been asked the dependency question, this share carries a typed depends on edge. Below that, most of the map would be missing from its own diagram.
2. Never-asked ≤ 5% of all briefs. This clause prices the denominator of the first one. Without it the fastest way to raise coverage is to ask fewer briefs, and a ratio you can move by doing less work is not a measurement.
3. Longest path ≥ 4 edges. A shorter path means the graph is a shallow fan rather than a tree.

Where it currently stands.

• 113 — typed depends on edges, across 76 briefs
• Clause 1 — 76 of 295 adjudicated briefs (25.8%) — not met. Against all 295 briefs the same numerator reads 25.8%; both figures are printed because the distance between them is the finding.
• Clause 2 — 0 of 295 briefs never asked (0.0%) — met
• Clause 3 — 5 edges, longest dependency path — met. Unaffected by the mutual dependency described below.
• 1719 — adjacency links between briefs, for contrast: the corpus is densely cross-linked, and none of that density is evidence of dependency

2 of the 3 clauses are met. Clause 1 (adjudicated coverage) is outstanding, so the rule says hold. This sentence is generated from the same figures as the list above rather than written beside them — an earlier version of this section carried 24 directed dependency edges in prose, derived from nothing, and wrong. Every number here is computed from the briefs at build time.

One finding worth surfacing early. The typed graph is acyclic. Earlier releases carried a mutual dependency — two topics each recorded as depending on the other, which is an editorial question rather than an extraction bug, and over which no tree can be drawn. The adjudication resolved it, and the longest-path figure above therefore rests on no edge this page calls an error.

A second kind of dependency, and why it is counted separately

Until v0.18 an edge on this map could only point at another brief. That turned out to be a limitation of the schema which was being read as a fact about the world, and the release that made it visible was Category V · Energy Systems. Fourteen briefs were authored. Thirteen of them recorded no dependency at all, and the coverage figure above fell — from 47.4% before that category to 25.8% after it — on work that was predicted to raise it more than any category since the first.

The briefs were not being evasive. Each one named its dependencies in prose and then declined to record them, because none of them was a result on this map. The Small Modular Reactors brief is the clearest case: it names a factory that does not exist, an order book deep enough to build it, licensing throughput, and high-assay enrichment capacity roughly eight times what has been demonstrated — and recorded none of the four, on the explicit reasoning that inventing an edge to improve a coverage metric is worse than leaving the row empty. That reasoning was right. The row was the problem.

A dependency does not have to be a physics constraint to belong on the map. Those four constraints are dependencies, and they are the binding ones. So there is a second edge kind for dependencies that are not briefs, under a closed vocabulary: industrial capacity, institutional or regulatory, market or capital, a scientific result nobody is producing, supply chain. It is validated at build time — an unrecognised prefix fails the build rather than warning, because a free-text dependency field would be a prose field with angle brackets, and the point of a typed edge is that a machine can read it.

The fifth term was added in v0.19, and the reason is worth stating. The original four all name something the world could choose to provide: build the plant, pass the law, pay the price, find a second supplier. Adjudicating the whole corpus turned up a class the schema could not express at all — briefs whose binding constraint is a result nobody is producing. A handful of Category I subjects wait on conservation of momentum being wrong, on a route past the Ford–Roman inequalities, on a coupling to gravity that does not exist. With four prefixes the only available answer was that they wait on nothing, which on a map of the edges of science is precisely backwards: a brief blocked by a violation of momentum conservation and a brief blocked by nothing at all were indistinguishable to every count on this site. The distinction the fifth term draws is the one this map exists to make — between a subject waiting on engineering and a subject waiting on discovery.

• 1003 — non-brief dependency edges, naming 999 distinct constraints, across 251 of 295 briefs (85.1%)
• 435 institutional or regulatory · 174 a scientific result nobody is producing · 157 industrial capacity · 137 market or capital · 100 supply chain

These two figures are never added together, and there is no combined percentage anywhere on this site. They measure different universes. Brief-to-brief coverage is a statement about this map — how much of the corpus is wired to the rest of it. Non-brief coverage is a statement about the world. Their sum would also be arithmetically wrong, because a brief can carry both kinds, and 34 currently do. A single blended number would read as progress while referring to nothing, so the build asserts that the two are reported separately and fails if they are ever presented as one.

What standing these edges have. A non-brief edge is this project's assertion, drawn from the brief's own sourcing and reasoning. It is not a citation to somebody else's dependency graph, because no such graph exists for most of these topics. Read them as "this brief, having looked, says the binding constraint is here" — which is weaker than a measurement and stronger than a guess, and is exactly the standing the four-flag system was built to make legible.

What this map cannot see

Four limitations, stated because a dependency graph that does not publish its blind spots invites a reader to assume it has none.

• Adjacency is not dependency. The 1719 adjacency links above are the densest thing in this corpus and none of that density is evidence of a prerequisite. Links introduced with “connects to”, “overlaps” and “shares” were deliberately never promoted to edges. A tree built by counting hyperlinks would imply a structure of prerequisites the briefs do not claim.
• The map cannot see dependencies outside itself. A brief may depend on a topic this project has never scoped, and there is no edge kind for “something that is not on our list”. The two kinds above cover dependencies on briefs and dependencies on named constraints; a dependency on a subject nobody here has thought of is invisible, and its absence from this page is not evidence of its absence from the world.
• 0 of 295 briefs have never been asked the question. This is the most important line on the page and it has not been published before. The coverage figure above is 25.8% — but that means 25.8% adjudicated and typed, not 25.8% have dependencies. The remaining 0 briefs were written before the schema existed and no one has since sat down with them and decided. Their dependency rows are empty because they are unexamined, not because they are empty, and a percentage whose denominator includes them is measuring the project's attention as much as the corpus.
• A recorded absence is now distinguishable from an unexamined one. A brief that has been asked and honestly has nothing to record is marked as such, so the three states — typed, adjudicated-empty, never asked — can be counted apart. 2 brief(s) are currently marked as asked-and-empty.

What this means for the threshold, and what was done about it. The original condition was set against a schema that could only see one kind of dependency, and on Category V's evidence it was not reachable by authoring: if most briefs' binding constraints are institutional or industrial, then every further category makes the brief-to-brief percentage worse rather than better. That was a finding about the metric rather than a reason to do more of the work, and the previous release deliberately left the bar untouched rather than move it in the same release that produced the evidence against it.

It was re-specified in v0.19, and the sequence is the safeguard. The denominator moved to the briefs actually adjudicated — the change three releases of deferral argued for — and clause 2 was added in the same stroke to stop that denominator being chosen rather than earned. Both halves were written down and published before the adjudication that moves the numbers began. Deciding a rule after seeing the number it governs is how a metric becomes a rationalisation; the defence against it is not good intentions but the order of operations, and the order is on the record.

Why closing the gap is editorial work, not a script. The typed edges above were promoted from dependency links the briefs already assert in their Research dependencies sections, one at a time, by reading them. Closing the gap means deciding, brief by brief, what each topic actually requires — which is the work, and the reason it is not a text-matching pass.

See also the Citation Record (every source cited across the corpus, four-flagged), the Frontier Research topic map, the Frontier Convergence Map, and the research programmes.