Every brief on this site ends its technology tree the same way: a row that says Requires (not on this map). Those rows are where the workback plan leaves the laboratory — the point past which what is missing is not a result but a licensing regime, a supply chain, a buyer, a referee, or a fact nobody is funding anyone to produce. Each such requirement carries a machine-readable token in a closed five-kind vocabulary, and this page is those tokens turned around: instead of asking what one topic requires, it asks what the frontier as a whole keeps requiring.

INST
institutional or regulatory — 435 constraints (43%)
IND
industrial capacity — 157 constraints (16%)
MKT
market or capital — 137 constraints (14%)
SCI
a scientific result nobody is producing — 174 constraints (17%)
SUP
supply chain — 100 constraints (10%)

What the atlas shows

The frontier is mostly blocked institutionally. Of the 1003 constraints the corpus names, 435 — 43 percent — are institutional or regulatory, against 174 (17 percent) that are a scientific result nobody is producing. For most of what the Institute covers, the workback plan does not end at a physics problem. It ends at an institution that does not exist yet. That is worth sitting with, because it inverts the usual picture of "impossible" research: the corpus says the decisive missing ingredient is more often a certification pathway, a consent framework, or an evaluation office than a discovery.

The same seven constraints recur almost everywhere. Grouping the tokens by what they ask for, rather than which topic asks:

  • supply or capacity at a scale nobody has built — named 79 times, in 10 of 10 categories.
  • an independent referee — verification, audit, adversarial review — named 63 times, in 10 of 10 categories.
  • a first buyer, or capital patient enough to wait — named 57 times, in 10 of 10 categories.
  • funding that outlasts one programme or one government — named 40 times, in 10 of 10 categories.
  • a licensing or certification regime that does not exist yet — named 47 times, in 9 of 10 categories.
  • liability, enforcement, or a statute with teeth — named 44 times, in 8 of 10 categories.
  • a standardised measurement, protocol, or convention — named 29 times, in 8 of 10 categories.

The top of that list is the finding. A first buyer and patient capital is named in every category the Institute covers — antimatter propulsion and civic technology have, on this one axis, the same missing ingredient. The independent referee recurs nearly as widely: again and again the brief's position is that the science could move, but no institution exists whose job is to check the claim that it has.

Energy is the exception that proves the pattern. Category V is the only category where industrial, supply and market constraints together (79) outweigh institutional ones (27). Energy's frontier is blocked the old-fashioned way — by factories, feedstocks and capital costs — which is what a field looks like when its institutions already exist. The governance-heavy categories are what a field looks like when they do not.

Where constraints concentrate. The heaviest single briefs are Advanced Battery Technologies (7), Advanced Fission (7), AI-Assisted Governance (7) — topics whose engineering is close enough to real that the non-research obstacles have become the whole story.

Constraints by category and kind

CategoryINSTINDMKTSCISUPTotal
I — Physics & Propulsion 12176138 56
II — Space & Astronomy 322611297 105
III — Biology & Human Enhancement 6226193421 162
IV — Consciousness & Intelligence 3466194 69
V — Energy Systems 2735231021 116
VI — Climate & Planetary Engineering 41920186 94
VII — Civilization-Scale Infrastructure 3621151213 97
VIII — Governance & Institutions 998151510 147
IX — Economics & Society 57418117 97
X — Historical & Frontier Science Studies 3554133 60

The atlas

Every constraint, grouped by kind, each linked to the brief whose workback plan names it. The wording is the token's own — the vocabulary the corpus uses to name what it is missing.

INST — institutional or regulatory (435)

IND — industrial capacity (157)

MKT — market or capital (137)

SCI — a scientific result nobody is producing (174)

SUP — supply chain (100)

Derivation: tokens read from data-requires attributes in the brief corpus at build time. The five-kind vocabulary is closed and machine-enforced; a brief-to-brief dependency is a different edge type and is deliberately not on this page. Version v0.5 · September 2026.