The Legibility Gap
The wartime essay called its village a model, not a measurement. Here is the measurement: assessed war damage against compensation actually disbursed, and the registry that has to read you before either one can reach you.
The wedge is not the same as slowness. The essay names the load-bearing reason a large share is structural: a property registry that was 40% complete at the invasion[src] and cannot read most pre-2013 owners. You can carry a laptop to the kitchen and the query still comes back empty.
The Invisible Architecture series reads classification systems as infrastructure: the most powerful ones sink into the background and sort the world while staying invisible to the people being sorted. The wartime essay pointed that lens at Ukraine, where the category assigned is a home and the adjudicator is a state under fire. It argued the device gate is removable and the registry gate is load-bearing, and it ran an illustrative cohort to show the compounding. This page replaces the illustration with sourced figures, refreshed quarterly, across the four threads the essay could only gesture at.
01The registry vs. the damage race
The load-bearing gate. Reconstruction need keeps climbing; the registry that has to verify each claim was 40% complete at the invasion.
Direct damage rose from $176B to $195B across the same two assessments. The need is a moving target the verification layer has to keep up with.
- Digitization into the electronic State Register of Property Rights began only in 2013, so most pre-2013 ownership records are absent.[src]
- Inclusion is significantly lower for Crimea and parts of Donetsk and Luhansk, the territories where damage is heaviest.[src]
- About 200,000 land parcels already held discrepancies between the property registry and the land cadastre before the invasion.[src]
Apply the record-completeness rate to the 3M impacted households[src]: at the documented 40%, about 1.8M sit off the registry. This is the essay's second prescription as a dial. Funding digitization is not back-office hygiene; it is how many households move from unrepresentable to legible. An illustrative overlay, not a per-household count.
02The damage to compensation gap
What the throughput numbers hide. Real money moved fast, set against assessed damage.
Destroyed-housing awards are far larger per family, so 28,000 destroyed-home awards account for most of the money while 116,500 damaged-home repairs account for most of the recipients. 179,200 applications were filed for damaged-housing compensation. Set the UAH 52B total against the $90B housing need and the distance still to travel is the story the recipient count does not tell.
03The device gate, mapped
The removable gate. Who can reach the digital state, and how fast the access gap is closing.
Consecutive annual KIIS waves. Adoption has risen for a third year running. 48% of public-service users went through Diia. KIIS reports a widening divide by age, income, and urban or rural location.
The series essay cited a 2024 figure of 38% of over-70s offline. The 2025 KIIS wave reports 28% of older Ukrainians never using the internet, on a different age cutoff. The direction is the same and the gap is still the widest of any cohort. Keeping both is exactly what a refreshed tracker is for.
04The residual category, made literal
The other box. Which coded categories of loss are open for claims, and which are not yet.
Roughly nineteen of a planned set of more than forty coded categories are accepting claims after the April 2026 expansion; the rest open in stages. More than 150,000 claims are recorded against an anticipated 300,000 to 600,000. Until a category opens, the loss it would hold has no field to go in: the residual box, made literal.
05The village, made honest
The wartime essay ran an illustrative cohort and labeled it a model, not a measurement. Here is the same model with the dials exposed and every parameter sourced, plus the one number it was really about: the registry completeness at which the record gate overtakes the device gate.
400 households through the same three gates. Every parameter is sourced and ranged; move them and watch which gate becomes load-bearing.
At these settings the record gate is load-bearing: it stops 193 households, against 56 at the other. Below about 80% registry completeness, the record gate excludes more than the device gate no matter how many phones reach the village. That line is the essay's argument, made quantitative.
A model, not a measurement. The population-average registry presence equals the completeness dial by construction; older and rural households carry a documented lag. Deterministic and seeded, so a given setting always renders the same village.
06The non-registry path, abroad
The essay's first prescription was a non-registry path to proof, treated as core, not edge. Other post-conflict states have already had to answer it. Here is who rebuilt the record, who reversed the burden, and who demanded the document.
Four post-conflict regimes, placed by how much they make the claimant supply the record. Ukraine's eRecovery sits at the far right, beside Cyprus.
How it handled the missing record. An international commission built its own computerized cadastre and occupancy database, parallel to the local registries it could not rely on.
The proof it accepted. A claim could be filed with no documents at all; the commission verified ownership against the record it had reconstructed, not the one the claimant had lost.[src]
The strongest non-registry path is to rebuild the record as a public task, then adjudicate against it.
Where Ukraine sits. eRecovery verifies ownership against the State Register of Property Rights, about 40% complete. It sits at the documentary-proof end, beside Cyprus and far from Bosnia and Colombia. Prescription #1 is to move it left.[src]
Methods and provenance
The headline compares compensation disbursed to date against assessed need. It is not a claim that Ukraine has paid only a fraction of what it owes. eRecovery is one program on a multi-year horizon, a large share of assessed damage sits in occupied territory beyond any current claim path, and "need" is a decade-long reconstruction estimate. The gap measures distance still to travel. The series essay's argument is about why a large part of that distance is structural rather than merely slow: the legibility gate sits below every interface, including the AI agent now bolted on top.
Initial publication and first re-verification pass. Folded in the Register of Damage's 29 April 2026 opening of legal-entity (group C) categories and its roughly 150,000 claims by May 2026 (up from 30,000), and established the quarterly refresh cadence.
Content-as-code: every figure is a sourced constant in dataset.mjs, refreshed by pull request. USD conversions use an approximate 42 UAH to the dollar. Figures reflect public reporting as of the refresh date above.
The essay this dataset is the measurement for. It called its own cohort a model, not a measurement.
Two readings of digital public infrastructure that carries constitutional force without a ratifying moment. Unratified builds the identity layer in the reader's hands; the Wartime Edition points the same argument at property, where the category assigned is a home and the adjudicator is a wartime state.
The same essay, two countries and two stakes. The Sorting Machine watches special-education placement sort children for compliance; the Wartime Edition watches Ukraine's eRecovery sort homes under fire. Same residual category, same inherited installed base, far higher stakes.
Classification as Infrastructure names the pattern in the abstract and inside AI systems. The Wartime Edition shows it converging in one live system: an LLM agent sitting on top of a state classification machine, mediated through MCP and the Trembita bus, inheriting every category the registry already enforces.
What the State Keeps reads the state as a classified inventory evaluated from above. The Wartime Edition reads one wartime state's compensation machine from below, from inside the residual category it cannot represent.