The Living Section
A drawn section for every life in a place, with the evidence on every line.
A method I am developing and testing: not yet peer reviewed or applied.
The rule
Enough. Room. Limits. Known. Kept.
Enough for everyone, room for every life, within limits, known and kept.
Cut a real place in section and ask, for every stretch of it, whether everyone has enough, whether other lives have room, whether the place stays within its limits, how well each answer is known, and whether what was promised is kept.
Living means every life in the place, human and not. It never means a live feed: the section is re-read on a schedule, not streamed.
Choose a word: each opens a layer of the place, the practice that reads it, and where to start.
One cut, three registers, one corridor, two ledgers
Patrick Geddes drew a river valley in section to argue that place, work and people cannot be understood apart. The Living Section draws that cut again, through one real place and from measured data, and adds what Geddes lacked: floors and ceilings, other lives, flows, uncertainty, time and a record of promises.
The five words keep their meaning and gain a place: Enough is the floor of People, Room the floor of Place, Limits the ceilings, Known the fog, Kept the ticks.
People
Enough, and Say
Who lives, works and passes here, and whether the worst-served can reach, rest, meet and stay, measured for an older walker, a wheelchair user and a carer before a default adult. Who decided is logged, never scored.
After Geddes' folk; Hägerstrand; Lefebvre and Harvey
Place
Room, and the limits here
Ground, water, air, heat and the other lives that share them, with a floor for each functional group and a ceiling for what the ground can bear. The past ground stays on the sheet as a ghost line.
After Geddes' place; Reclus; McHarg; the Earth Commission
Work
Stock, flow, service
The buildings, streets and pipes that make the place run, what flows through them, and what they provide: shelter, shade, mobility, care. A gap in a floor can be traced to the provision missing beneath it.
After Geddes' work; Wolman; Haberl; Odum; Vogel
Every strip carries a floor, the least each stretch must provide, and where one applies a ceiling, the most pressure it may exert: the Doughnut laid straight along a real place. Thresholds are bands, never lines. A declared one is marked D. Only a shortfall under a floor or an overshoot through a ceiling is filled.
Above the section runs elsewhere: the footprint of what is consumed here, drawn apart and never netted against the place, because a green roof does not offset a flight.
Known is fog wherever a value is not known well enough to draw. Kept is a tick on the ground line for every promise: hollow when promised, filled when verified at a re-read, crossed when verified absent, each with who monitors it and who decided.
Five questions, one place
Every stretch of the cut gets the same five questions. It does not score a place. There is no composite number: Enough, Room and Limits cannot be traded against each other, and every line says how it is known. The projects cited below came before the framework. They are where its parts were learned, not applications of it.
Enough
Can everyone here, the worst-served first, reach what a decent everyday life needs, without using more than they need?
Care, food, school, green and cool space and frequent transport within reach, measured for an older walker, a wheelchair user and a carer with a pram as well as a default adult, and reported for the worst-served, never only the mean.
- Builds on
- Minimum access in the Earth Commission's safe and just corridor; sufficiency and priority to the worst-off in transport justice; accessibility theory; sufficiency in the IPCC's sixth assessment.
- Adds
- The human floor set in the same units and drawing as the floor for other lives and the limits of the place.
- Who it matters to hypothesis
- Relevant to public bodies that must weigh social value when they commission, and to planning and health evidence on access to everyday services. See Public Services (Social Value) Act 2012.
Learned inSocial Superblockschrono.city
Room
Do the other lives that share this place have enough habitat, connection, darkness, water, refuge and safe passage to persist?
A floor for non-human life, set per functional group (pollinators, birds, bats, ground mammals, aquatic life, trees and soil), measured at street and neighbourhood scale and kept separate from what nature does for people.
- Builds on
- Multispecies justice; more-than-human nature-based solutions; urban biodiversity science; the nature-positive goal of halting and reversing loss by 2030.
- Adds
- A declared floor for each functional group of other lives, reported beside the human floor with a rule against trading one for the other. Ecological-justice diagnostics and species-led design already bring non-human needs into planning; this floor-beside-floor structure was not found in the literature reviewed.
- Who it matters to hypothesis
- Relevant to statutory nature delivery and recovery strategies, and to urban green space and canopy targets. See Biodiversity net gain (GOV.UK); EU Nature Restoration Regulation 2024/1991.
Learned inDérive
Limits
Does this place stay within what its ground can bear, and within a declared fair share of the planet's budgets?
Three kinds of line kept apart: boundaries observable in place, such as air quality, water and green space; thresholds declared locally and labelled as such, such as heat stress; and fair shares of global carbon and material budgets, drawn as a range across sharing principles, never a single pass or fail.
- Builds on
- Planetary boundaries and the Earth Commission's safe and just Earth system boundaries; their translation to cities and businesses; the sensitivity of fair shares to how they are allocated.
- Adds
- The split between observed, declared and allocated limits in one visual grammar, with allocation drawn as a fan of ranges.
- Who it matters to hypothesis
- Relevant to lenders, insurers and owners who disclose physical climate risk, and to cities adapting to heat and flooding. See IFRS S2 Climate-related Disclosures.
Learned inShaping Cool CitiesWhere the EA Said NoCityLoops: Urban Circularity Assessment
Known
How is each answer known, where does the model behind it hold, and whose knowledge is in it?
Every line carries a class, from observed to illustrative, and every model an applicability envelope. Where knowledge does not hold, the drawing shows fog instead of a value. Where stakes justify it, independent models and a community account answer the same question side by side.
- Builds on
- The NUSAP pedigree matrix and IPCC confidence language; spatial cross-validation and the area of applicability; critiques of single digital twins; the CARE principles for data governance.
- Adds
- A rule that every line of one place carries two grades, how it is known and where it holds, never collapsed into one. For AI Earth embeddings, the area of applicability is computed on the embeddings themselves, with two further terms not found in the literature reviewed: whether the place could have been in pretraining, and how much error the decision can tolerate.
- Who it matters to hypothesis
- Relevant to organisations using spatial models, digital twins and AI in decisions who need provenance and validation they can defend. See The Gemini Principles (Centre for Digital Built Britain, 2018).
Learned inShaping Cool CitiesClimate Intelligence
Kept
What was promised for this place, by whom, on what evidence, and is it verifiably there and working now?
A ledger from promise to outcome that separates inputs (trees planted) from outputs (trees surviving) and outcomes (shade on the route people walk), and records who decided, who was absent, and what the drawing leaves out.
- Builds on
- The permanence obligations of biodiversity net gain; monitoring and evaluation of adaptation; salience, credibility and legitimacy; the ladder of participation; the quadruple helix.
- Adds
- A standing ledger across instruments (greening factor, net gain, canopy strategies, disclosures) that records input, output and outcome, each graded, and is re-read on a schedule. One-off delivery audits exist; the ledger generalises and repeats them.
- Who it matters to hypothesis
- Relevant to planning authorities monitoring conditions, and to developers and owners who must show that greening and nature commitments are delivered over decades. See Urban Greening Factor (Greater London Authority).
Room: a floor for other lives
The safe and just corridor sets minimum access for people only, and multispecies justice argues that justice reaches further but offers no thresholds. Room sets a floor for the other lives that share a place, by functional group, beside the floor for people and never traded against it. Which groups count, and who speaks for them, is itself a decision the section records. Floors are declared, dated and co-set; where published thresholds exist they seed the declaration and are cited.
Functional groupsPollinators · Birds · Bats · Ground mammals · Aquatic life · Trees and soil
Habitat
Extent and condition of suitable habitat within the group's range of movement.
Published seedAt least 20 to 25% semi-natural habitat per square kilometre, a landscape-scale floor set for nature's contributions to people. Mohamed et al., 2024
Connection
Whether habitat patches can be reached across streets and buildings.
Darkness
Hours and extent of darkness along corridors and habitat edges.
Published seedFor light-averse bats: below 1 lux on corridors, with streetlights at least 50 m away. Azam et al., 2018
Water
Clean, open water and permeable ground.
Published seedStreams: tree buffers of at least 30 m along small streams. Sweeney and Newbold, 2014
Refuge
Nesting, roosting and overwintering sites, counted as inputs until occupied.
Safe passage
Avoidable deaths on roads and glass.
Published seedGlass: markers spaced no more than 5 to 10 cm apart prevent nearly all bird strikes. Klem, 2009
From material to city: fifteen design moves
Buildings are rated one at a time; the street and the neighbourhood between them are where no scheme is responsible and where the most is decided. Each question becomes three moves, and each move has a form at every scale, from a wall build-up to a borough. Every move says which instrument it feeds, from the Urban Greening Factor to BREEAM, WELL and the Living Building Challenge: the section supplies graded evidence, and the scheme keeps its authority.
No points, no levels, no badge. A move is dotted until a re-read shows it was kept. The layout follows Waugh Thistleton's five-principle wellbeing framework; the difference is Room. Biophilia is what nature does for us. Room is what we owe it.
MaterialUnitBuildingPlotStreetBlockNeighbourhoodCity
Enough
E1Bring the everyday within reach
- Material and unit
- Flexible layouts that can take a carer, a desk or a lodger
- Building and plot
- Ground floors for care, food and school, not only car parks and lobbies
- Street and block
- Places to stop and rest; step-free crossings
- Neighbourhood and city
- Care, food, school, green space and frequent transport within the walk of the slowest profile
- Feeds
- Healthy Streets; 15-minute measures per profile; BREEAM Communities; LEED ND
E2Give every body a way through
- Material and unit
- Controls people can reach and operate
- Building and plot
- Level thresholds, lifts that work, entrances on the street
- Street and block
- Shade and shelter on the routes people walk
- Neighbourhood and city
- Continuous step-free, shaded, lit networks
- Feeds
- WELL Movement; Living Building Challenge; Healthy Streets
E3Use the stock before adding to it
- Material and unit
- Reuse materials and fit-out
- Building and plot
- Fill under-used floor space; share rooms, roofs and yards
- Street and block
- Open school yards and courtyards after hours
- Neighbourhood and city
- Count service per tonne of stock held, not new floor area
- Feeds
- London Plan pre-redevelopment assessment; whole life-cycle carbon; One Planet Living
Room
R1Keep it dark
- Material and unit
- Warm, shielded, dimmable fittings
- Building and plot
- No spill light onto roofs, walls and edges used as corridors
- Street and block
- Dark gaps along corridors (seed: below 1 lux, lights at least 50 m away)
- Neighbourhood and city
- A mapped dark network, protected in the Local Plan
- Feeds
- Biodiversity Net Gain condition; Local Nature Recovery Strategies; light pollution credits
R2Join it up
- Material and unit
- Green roofs and walls chosen for continuity, not score
- Building and plot
- Permeable boundaries: hedgehog gaps, no hard fence lines
- Street and block
- Tree and verge links across the carriageway; soil under paving
- Neighbourhood and city
- Corridors that connect patches across the neighbourhood
- Feeds
- Urban Greening Factor (as evidence, not the metric); Local Nature Recovery Strategies; SITES
R3Build in refuge and safe passage
- Material and unit
- Swift and bat bricks, dead wood, bird-safe glass (seed: markers 5 to 10 cm apart)
- Building and plot
- Nest and roost counts, occupied or not
- Street and block
- Crossing points and speeds where animals cross
- Neighbourhood and city
- Collision hot spots surveyed and removed
- Feeds
- BREEAM Land Use and Ecology; Biodiversity Net Gain; counted as inputs until occupied
Limits
L1Keep what stands
- Material and unit
- Retain structure; specify low-carbon, reusable materials
- Building and plot
- Refurbish before rebuild; test against upfront carbon caps
- Street and block
- Reuse kerbs, setts and street furniture
- Neighbourhood and city
- A borough map of the stock that could be retained
- Feeds
- Whole life-cycle carbon; Circular Economy Statements; RIBA 2030 Climate Challenge
L2Cool and drain where it lands
- Material and unit
- Pale or planted roofs; shading
- Building and plot
- Permeable surfaces, attenuation, trees with rooting volume
- Street and block
- Street trees with soil volume; sustainable drainage in the kerb line
- Neighbourhood and city
- Canopy towards 30%, placed where heat and deprivation coincide
- Feeds
- Urban Greening Factor; London Plan tree policy; TNFD and IFRS S2 physical risk
- What the evidence says elsewhere
- Every London roof made cool: about 0.8 °C cooler near-surface air, modelled for summer 2018. One roof alone does not change the street's air. Simpson et al., 2024
- A green roof held about a third of spring rain and cut peak runoff by more than half; measured on a test plot in Sheffield. Stovin, 2010
L3Stay within a declared share
- Material and unit
- Operational energy per unit, metered
- Building and plot
- Measured energy use against targets
- Street and block
- Street-scale energy and heat networks
- Neighbourhood and city
- Carbon and materials drawn as a fan across sharing principles
- Feeds
- UK Net Zero Carbon Buildings Standard; NABERS UK; Be Seen; One Planet Living
Known
K1Grade every line
- Material and unit
- Product declarations kept apart from generic factors
- Building and plot
- Design prediction, as-built and metered kept apart
- Street and block
- Walked, surveyed or modelled, labelled
- Neighbourhood and city
- Every layer registered with its evidence class
- Feeds
- The Validation Card; disclosures; any scheme's evidence file
K2Draw where the model holds
- Material and unit
- Material data outside its tested range flagged
- Building and plot
- The energy model's gap to meters stated, not averaged
- Street and block
- Heat and canopy models fogged outside their envelope
- Neighbourhood and city
- Portfolio maps built only from validated layers
- Feeds
- IFRS S2; TNFD LEAP; life-cycle assessment review
K3Put the lived account beside the model
- Material and unit
- Resident diaries on comfort and control
- Building and plot
- Post-occupancy evaluation
- Street and block
- A walked co-survey of the cut, including the animals seen
- Neighbourhood and city
- Community counter-layers with named authors
- Feeds
- RIBA post-occupancy evaluation; Fitwel surveys; Local Plan engagement
Kept
P1Write the promise where it can be found
- Material and unit
- Specified materials and reuse commitments logged
- Building and plot
- Planning conditions, greening plans, net gain agreements and rating targets in one ledger
- Street and block
- Street tree and lighting commitments logged
- Neighbourhood and city
- Every promise for the place, from every instrument, on one register
- Feeds
- Planning conditions; Section 106; the Biodiversity Net Gain register
P2Check input, output and outcome
- Material and unit
- Delivered against specified materials
- Building and plot
- Planted, surviving, shading; installed, metered, in use
- Street and block
- Trees alive and in canopy; lights dimmed where promised
- Neighbourhood and city
- Change in canopy and access for the worst-served
- Feeds
- Be Seen; net gain monitoring; the outcome evidence greening scores lack
P3Re-read on a date
- Material and unit
- At handover
- Building and plot
- At one, three and five years
- Street and block
- At each tree establishment period
- Neighbourhood and city
- At 30 years for net gain obligations; at each Local Plan review
- Feeds
- The Re-read offer; statutory monitoring
Where moves pull against each other
- E2 R1Safe, lit walking routes against dark corridors for batsGive every body a way through · Keep it dark
- L2 E1Greening that raises rents and displaces residentsCool and drain where it lands · Bring the everyday within reach
- L1 E1Retained structure that limits layouts and daylightKeep what stands · Bring the everyday within reach
- R2 L3Roof space for habitat against roof space for solarJoin it up · Stay within a declared share
- K3 K2Residents and models that disagree, shown side by sidePut the lived account beside the model · Draw where the model holds
Schematic. Moves are proposals until kept: none has yet been verified at a re-read in a real place.
Known: how each line is known, and where it holds
Every line carries one of six words, each with its own line style. Every model also carries an applicability envelope: where its validation holds, extended for AI Earth embeddings to include what the model may already have seen, how far a place sits from its training data and how much error the decision can tolerate. Outside the envelope the drawing shows fog, not a value.
observed
Measured directly: a sensor, a survey or a walk.
Solid line
sourced
Taken from a named, dated dataset someone else maintains.
Solid line with a source tick
derived
Calculated from observed or sourced layers by a stated rule.
Long dash
modelled
Predicted by a model, with its validation scheme and limits stated.
Short dash
inferred
Assumed from context, with the assumption written down.
Dotted line
illustrative
Drawn to explain, not to measure.
Hatch or halftone
Heat is three different things
- Surface temperature observed
- How hot the ground and roofs are, seen by satellite (land surface temperature, LST). No health or comfort claim is made from it alone.
- Air temperature observed
- How hot the air is, measured by weather stations at about two metres.
- Thermal comfort modelled
- How hot it feels to a person, combining air, radiation, wind and humidity (for example UTCI). Usually modelled.
Where AI earns influence
v3 candidate A v3 candidate, drafted 25 September 2026 from a review of the 2024 to 2026 research. Not yet adopted, applied or peer reviewed.
AI earns influence over a place one rung at a time. Recent benchmarks found that geospatial foundation models often fail to beat simple supervised models, and one 2026 preprint, not yet peer reviewed, found they lose about 15 to 20 per cent of their performance away from their training data. That is why every AI line on a Living Section shows where it holds, and how far it is allowed to reach. See PANGAEA benchmark, 2024; EarthShift, 2026.
Vision
Described, not built or not tested.
On the drawingAppears only hatched, labelled as vision.
Research
A published method, benchmarked elsewhere, with a model card.
On the drawingKept in the working register; not drawn on the public section.
Demonstrated
Works on comparable places under spatially blocked or held-out validation.
On the drawingDrawn as modelled and tagged not validated here; cannot move an option.
Validated in place
Local labels, a validation regime that matches the question, the optimism gap reported, the envelope drawn and baseline parity shown.
On the drawingMay inform the ranking of options; fog outside its envelope.
Decision-grade
Validated in place, within the tolerance of the action, checked against an independent account, and through at least one re-read.
On the drawingMay bear on a logged decision.
Baseline parity: a foundation-model layer is shown next to a simple supervised model built on native-resolution data, such as aerial imagery and LiDAR. If it does not win by a margin set in advance, the simpler model is drawn instead: it is cheaper, easier to inspect and depends on no single vendor.
Influence is re-earned, not inherited. A line drops a rung when a re-read fails, the model or sensor changes, or drift passes a declared limit, and its rung expires at the next re-read. Demotions are published as prominently as promotions.
Where it helps each question, and where it must not lead
- Enough
- Street-level imagery to audit pavements, kerbs, shade and seating along routes, per ability profile.
- Must not leadThe access calculation itself: distance and routing stay in deterministic GIS.
- Room
- Habitat extent and condition from Earth embeddings with few labels; canopy and structure from LiDAR.
- Must not leadWhether animals are there. Embeddings are habitat proxies, not bats or birds.
- Limits
- Heat from thermal missions, generative downscaling for ranges, subsidence from radar, stock tonnes from footprints and heights.
- Must not leadDeclaring a limit. Generated fields are always modelled and drawn as ranges.
- Known
- Computing envelopes in embedding space, drift checks, baseline parity and plural checks.
- Must not leadGrading. No model grades its own evidence.
- Kept
- Annual embeddings from 2017 as a baseline for old promises; extracting planning conditions from documents, each linked to its source page.
- Must not leadDeclaring a promise kept: that needs a validated outcome at re-read.
The twin, defined
A twin of a place is a dated, graded view of its evidence register, for a stated question and a stated action. Its rate of synchronisation is the re-read schedule. Its maturity is the share of the cut, per question, held at validated in place or above, not its refresh rate or its autonomy. ISO/IEC 30173:2023 defines a twin by an appropriate rate of synchronisation, not real time.
The theory of change
A change is transformative when it reconfigures the stocks of a place, its buildings, streets, canopy, soil and water, so that more of the worst-served have enough and more of the other lives have room, per tonne held, without pushing any limit over its line; when that movement is known well enough to act on; and when, at the next re-read, it has been kept.
The method, in six steps
Each step writes to one open evidence register, as STAC items with cloud-native assets, so any line on the drawing can be reopened and re-run.
Frame
Choose the place, the cut and the question with the people who will use the answer. Pre-register the cut, the groups Room will count and the thresholds, with who set them.
Walk
Walk the cut. Record what data leave out: use, fear, delight, sound, and the animals actually there.
Measure and gate
Build each layer, grade how it is known, validate every model for the question asked and draw its envelope. Nothing outside the envelope carries a value.
Set the lines
Draw the Enough floor per ability profile, the Room floor per group, and the limits as ranges.
Draw and decide
Publish the section, convene the four strands around it, draw each option as an arrow and log the decision.
Re-read
Return on a schedule, check every promise in the Kept ledger and publish what changed.
Four modes of practice
Walk
Start on foot. Record what data leave out: sound, encounter, fear, delight, and the other lives that are there.
Read asArtist
Measure
Measure the place from orbit to the street, in open, cloud-native formats.
Read asAcademic
Model
Model the place, and show where each model holds before anyone decides with it.
Read asIndustry
Convene
Bring the four strands to the same drawing, agree what to change, name who owns it and keep the record.
Read asFounder / CTO
Propositions it has to survive
A framework earns its place by what it predicts. Each proposition will be pre-registered before any data are inspected, and each says in advance what would count against it.
P1First year
Room is not captured by existing metrics: among London sites that meet greening or net-gain targets, at least one in five will still fail a pre-registered Room floor for at least one group, most often darkness or connection.
Falsified ifFewer than one in ten fail.
P2First year
Envelopes change maps: for a London heat or canopy model built on pretrained Earth embeddings, error outside the applicability envelope will exceed error inside it, and more than a tenth of the area will fall outside.
Falsified ifErrors agree within noise and coverage exceeds 95 per cent.
P3First year
Promises are not outcomes, and design scores do not predict them. In four English cities, 23% of trees specified on approved planting schemes were never planted, and 38% of the development-led trees that were planted were in good condition (Brasington, 2025; grey literature; London not sampled). Against the number promised, that is about 30%, a ratio derived here, not stated in the report. In London schemes approved at least five years ago, the share will be no better, and the design-stage greening or habitat score will not predict verified canopy and shade.
Falsified ifLondon beats that benchmark by more than a pre-registered margin, or design scores explain a pre-registered share of the outcomes.
P4First year
Graded display changes decisions: readers of the graded section will match their confidence to the evidence more closely than readers of an identical ungraded one, with no loss in usefulness.
Falsified ifCalibration does not differ.
P5
Enough and Room conflict less than assumed: options designed with both floors visible will less often improve one while worsening the other than options designed for a single objective.
Falsified ifThe shares do not differ.
P6v3 candidate
The embedding advantage shrinks at the street: for street trees, permeable ground and shade on walking routes, models on 10 m Earth embeddings will not beat a baseline on sub-metre imagery and LiDAR by a pre-registered margin.
Falsified ifEmbedding models beat the baseline beyond the margin for street-scale features.
P7v3 candidate
AI evidence follows privilege: a smaller share of AI layers will reach validated in place across the most deprived fifth of a London cut than across the least deprived fifth.
Falsified ifThe shares do not differ beyond sampling noise.
P8v3 candidate
Annual embeddings can audit old greening promises at plot scale, about 400 square metres and above, but not tree by tree.
Falsified ifAgreement with aerial and LiDAR survey falls below the pre-registered threshold for plots, or reaches it for single trees.
P9v3 candidate
Agents are safer over a graded register: an agent that answers through deterministic GIS tools over graded evidence will make fewer wrong statements, and decline unanswerable questions more often, than the same agent without grades.
Falsified ifError and appropriate-refusal rates do not differ.
How to work with it
The Validation Card and the Promise Audit are the ways in: each tests one question for a client without asking them to adopt the whole framework. Each offer says whether it has been delivered, what backs it and what it does not claim.
ProposedKnown
Validation Card
An audit of someone else's heat, canopy, AI or twin layer.
The validation regime, the gap between random and spatial error, the applicability envelope and a model card.
Not yet delivered. The method is being applied first to my own heat model as it is rebuilt.
Does not claimAudits the method and its evidence, not the vendor's business.
ProposedKept
Promise Audit
What was promised for a scheme or portfolio, against what is verifiably there.
Greening, tree and habitat commitments traced to surviving, working outcomes from aerial and field survey.
Not yet delivered.
Does not claimNot a statutory biodiversity net gain monitoring report.
ProposedEnough
Section Scan
One place, read on all five questions.
A graded section, the three findings that matter and what to test first.
No Scan delivered yet. The first is planned as a self-initiated section across the Lower Lea.
Does not claimEvidenced layers only. No modelled flood depth, loss or valuation.
Proposed · precedentLimits
Place Atlas
The baseline of a district, with the portfolio view and an open evidence register.
Every layer graded and registered, and many places compared on axes built only from validated layers.
Partial precedent: the CityLoops circularity methodology I co-authored, designed for application in five pilot cities.
Does not claimA precedent under another name, not a commission of this offer.
Proposed · precedentKnown
Build
Software your team runs: pipelines, models and interfaces for one decision.
A working tool with its sources, tests and handover documentation.
Adjacent work: site-selection automation built in-house as an employee in retail, and the Climate Intelligence console, a working prototype.
Does not claimNot yet delivered as a client commission; no third-party audit of either tool.
ProposedKept
Re-read
The same place, read again on a schedule.
The same indicators re-run from the same sources, an updated promise ledger and a dated manifest of what changed.
Not yet delivered.
Does not claimScheduled re-runs, not a live feed or a live twin.
Four readers, one drawing
The section is a boundary object for the quadruple helix: academia, industry, government and civil society read the same drawing for different purposes, and each appears in the Kept record as a producer, contester and user of evidence.
Academia
A citable method, propositions to test and an open dataset.
In practiceShaping Cool Cities
Industry
Validation of the models behind a disclosure, and evidence of whether commitments were delivered.
In practiceClimate Intelligence
Government
Access by deprivation, nature recovery evidence and monitoring of planning conditions.
In practiceCityLoops: Urban Circularity Assessment
Civil society
Walked co-surveys, named counter-layers and a say in which lives count.
In practiceTwenty Years on the Lower LeaDérive
What already exists
The City Doughnut portraits by Doughnut Economics Action Lab (Amsterdam, 2020) already place a city between a social foundation and an ecological ceiling, and nature-positive and net-gain schemes already count habitat.
The difference to test: the portraits describe a whole city and count people only; net gain counts habitat units that can be met elsewhere. The Living Section draws one real place, sets a floor for other lives beside the floor for people, grades every line by how it is known, and checks at every re-read whether promises were kept.
Direction, not delivery
Where this is going: twins of real places, defined as graded views of an open evidence register, that people and AI can both reason over, and in time Physical AI that acts in the built world within the same floors and limits. A direction to be earned one validated artefact at a time, not a capability I claim today.
Physical AI stays a labelled vision until it passes six gates, each a test it can fail (v3 candidate):
- G1 Perception in place. The machine's perception holds, validated in place, for the place, season and conditions in which it acts.
- G2 World-model fidelity. Where a simulator plans an action, its predicted outcome is compared with the observed one at re-read, within the action's tolerance.
- G3 Floors not lowered. No action lowers a floor for people or for other lives: no blocked pavement for a wheelchair user, no light on a bat corridor, no mowing in nesting season.
- G4 Within limits. Energy and carbon per verified outcome are reported and within a declared budget.
- G5 Kept. Every action is logged as a promise and checked at re-read.
- G6 Governed. Its legal use class is recorded, a named person is accountable, residents have a right of reply, and nothing captures biometrics.
On the shoulders of
Every commitment here is borrowed and credited. What the framework adds is narrow and still to be tested: a floor for other lives, knowledge with an edge, and promises checked against outcomes, carried together on one drawing of one real place.
- The section as a way of seeing a region; place, work and folk read together Patrick Geddes, Cities in Evolution (1915), the Valley Section (in print by 1925), and "survey before plan"
- Humanity as nature becoming aware of itself; geography as history in space Élisée Reclus, L'Homme et la Terre (1905 to 1908)
- The city as the stage of social life, living with its region or feeding on it Lewis Mumford, The Culture of Cities (1938) and "The Natural History of Urbanization" (1956)
- Energy and material flows drawn as circuits Howard T. Odum, Environment, Power, and Society (1971)
- Good city form, including who controls a place Kevin Lynch, A Theory of Good City Form (1981)
- Society, ecology and technology as one system McPhearson et al. (2022), the SETS framework
- Needs met at low energy through public provision Vogel et al. (2021); Rao and Min (2018); Millward-Hopkins et al. (2020)
- Strong sustainability: critical natural capital is not substitutable Herman Daly (1990); Ekins et al. (2003)
- Commons and polycentric governance Elinor Ostrom (2009)
- Time geography and the right to the city Torsten Hägerstrand (1970); Henri Lefebvre (1968); David Harvey, Rebel Cities (2012)
- Nature drawn in profile, from sea level to summit Alexander von Humboldt, the Chimborazo Naturgemälde (1807)
- Layering nature before design Ian McHarg, Design with Nature (1969)
- Social foundation and ecological ceiling Kate Raworth, Doughnut Economics (2017), and the DEAL City Doughnut portraits (2020)
- Safe and just Earth system boundaries and minimum access Rockström et al. (2023); Gupta et al. (2024), the Earth Commission
- Translating boundaries to cities and businesses Bai et al. (2024)
- Multispecies justice Celermajer et al. (2021); Tschakert et al. (2021)
- More-than-human nature-based solutions Cecily Maller (2021)
- Stocks, flows and services Haberl et al. (2017); Abel Wolman (1965)
- Sufficiency access for the worst-off Karel Martens, Transport Justice (2016)
- The 15-minute city and superblocks Carlos Moreno; Salvador Rueda and BCNecologia
- Area of applicability and spatial validation Meyer and Pebesma (2021); Milà et al. (2022); Linnenbrink et al. (2024)
- Grading how well something is known The NUSAP pedigree matrix (van der Sluijs et al., 2005) and IPCC guidance on uncertainty (Mastrandrea et al., 2011)
- Plural models over a single twin Saltelli et al. (2024)
- Salience, credibility and legitimacy Cash et al. (2003)
- The ladder of citizen participation Sherry Arnstein (1969)
- The quadruple helix Carayannis and Campbell (2009)
- Indigenous and community data governance The CARE principles (Carroll et al., 2020)
- The dérive Guy Debord and the Situationists