# NeoBharat Knowledge City: Research Campus Architecture and 3D Production Plan

## Recommendation

NeoBharat should be modelled as a connected contemporary research campus that can grow into a city. Its credibility should come from a clear hierarchy of streets, buildings, entrances, courtyards and landscape. A collection of isolated landmark objects can communicate a programme, but cannot demonstrate how a knowledge city would work.

The recommended first design standard is a 20-hectare research quarter on a hypothetical 500 × 400 metre site. It combines two laboratory courts, a Learning Commons, project studios, an innovation building and shared amenities. This is a manageable area in which to judge architectural character at aerial and pedestrian scales before reproducing the approach across a much larger territory.

The accompanying working model establishes this spatial direction. It is an editable architectural visualisation prototype with consistent units and building identifiers. It does not establish a surveyed masterplan, approved floor area, laboratory performance, institutional capacity or a construction design. Its planting represents an indicative mature landscape rather than the first year of operation.

The architectural direction should combine compact academic neighbourhoods, active shared ground floors, shaded outdoor circulation and differentiated mid-rise buildings. The underlying principles are informed by MIT, ETH Zürich, EPFL, NUS, KAUST, Cambridge, IIT Gandhinagar and Stanford Research Park. These institutions are architectural or operational precedents; no association, endorsement or prospective partnership is implied.

## Evidence and interpretation

The precedent study gives priority to institutional publications and the architects responsible for the referenced projects. It separates completed buildings from long-term plans: an attractive future-campus rendering is evidence of a design intention, not proof of delivered performance. The sources support spatial principles; they do not supply a transferable NeoBharat building programme or permission to reproduce their buildings.

Quantitative dimensions in the proposed quarter are original modelling assumptions. They have not been reverse-engineered from campus photography. The full-city acreage schedule comes from the NeoBharat project brief and is a planning input rather than independent evidence of available or developable land.[^1]

### Precedent comparison

| Precedent | Evidence status | Relevant observation | Proposed NeoBharat application |
|---|---|---|---|
| MIT, 238 Main Street / Kendall Square | Completed building; completion identified as 2021 | Research workspace meets an atrium, ground-floor activity and public green space | A legible public-to-research interface and shared commons |
| ETH Zürich, Hönggerberg 2040 | Long-term campus development framework | Compact development, landscape continuity and a central public route | Infill around pedestrian courts before peripheral expansion |
| EPFL, Campus Piéton | Campus transformation project | Road space becomes a pedestrian environment while emergency and service access is addressed | Separate ordinary traffic from managed access to the academic core |
| NUS, campus sustainability programme | Institutional strategy and campus examples | Shade, greenery and outdoor comfort inform campus planning | Test the outdoor network as part of building design |
| KAUST, HOK campus | Built campus design described by its architect | Interconnected research neighbourhoods and shaded courtyards | Shared facilities and sheltered social circulation |
| Cambridge, West Cambridge | 2021 innovation-district vision | Academic and enterprise activity integrated with public spaces | An innovation street with visible shared functions |
| IIT Gandhinagar | Built campus, documented in a 2019 retrospective | Connected academic spaces and a response to existing terrain | Climate-aware courts and landscape-led site adaptation |
| Stanford Research Park | Operating estate guidance | Coordinated design review and management of tenant change | A common design code across multiple future operators |

### MIT: research interfaces and a usable public realm

MIT describes 238 Main Street as a completed laboratory and office project incorporating an existing historic building, a five-storey atrium and a twelve-storey addition. Its ground-floor uses and internal connection to public green space make the research district legible to people outside the laboratories.[^2] The relevant lesson is how a building connects public, shared and controlled spaces. It is not the particular height, historic façade or laboratory specification.

For NeoBharat, the Learning Commons should act as the recognisable civic entrance to the quarter. The first floor of public activity should include exhibition, orientation and informal learning space where programme permits. The model should show where a visitor arrives, where they can continue without authorisation, and where secure research circulation begins. This is a proposed access hierarchy, not an operational security design.

MIT’s description of its Kendall Square design development also stresses architectural variation within a collective setting.[^3] NeoBharat should therefore share a material family and public-realm standard while varying façade rhythm, building depth and roof form. Repeating one façade across every use would weaken the impression of a working institution.

### ETH Zürich: compact growth and landscape continuity

ETH’s Hönggerberg development material describes a compact campus within its circulation framework, improved public spaces and a green central promenade. The 2040 framework is explicitly a development horizon, not a statement that every illustrated building already exists.[^4] Its value to NeoBharat is the relationship between densification and a coherent campus landscape.

The proposed adaptation is to complete academic neighbourhoods around shared outdoor rooms. Each addition should strengthen an existing route, court or facility cluster. Growth should not leave isolated buildings in leftover grass, nor consume every open space that happens to be easy to build on. The full-city plan needs a distinction between protected landscape, ordinary development plots and spaces reserved for later intensification.

### EPFL: pedestrianisation with service obligations

EPFL’s Campus Piéton material describes a pedestrianisation project involving planting and furniture, alongside the reorganisation of service and emergency access.[^5] This matters because a car-free-looking rendering can conceal unresolved servicing. Laboratory deliveries, maintenance and emergency movement remain necessary even when the ordinary experience is pedestrian.

The NeoBharat study therefore places general circulation on a perimeter street and reserves the core for walking and shared outdoor activity. In the next planning stage, a separate access drawing must identify controlled entry points, delivery destinations and emergency routes. These routes should be coordinated with landscape and building entrances, not added after the architectural scene is complete.

### NUS: outdoor comfort as a design input

NUS describes its campus as a living laboratory and identifies shade, greenery and wind movement as considerations in campus environments such as University Town.[^6] This is useful evidence that landscape should be evaluated as infrastructure for daily use. It does not demonstrate that the same planting or building arrangement would perform identically in the Pune region.

NeoBharat should evaluate the route from arrival to learning, work, food and housing under local weather conditions. The next model should contain enough information to test sun exposure and outdoor comfort before finalising façade orientation. Covered links and planted courts are appropriate study devices, but neither establishes a quantified cooling benefit. Plant selection, irrigation and canopy development require a regional landscape strategy.

### KAUST: research neighbourhoods and shared facilities

HOK’s KAUST project account describes interconnected research buildings, common facilities, shaded courtyards and a large roof system that moderates exposure.[^7] The transferable principle is the organisation of specialist work around shared support and social circulation. The project’s scale, coastal desert conditions and engineering are not direct design parameters for NeoBharat.

NeoBharat should group research functions around common instrumentation, seminar rooms and collaboration spaces rather than assign every department a freestanding landmark. The first quarter illustrates this relationship with linked laboratory wings and courts. The specific types of laboratories remain placeholders until a research programme identifies servicing, vibration, containment, safety and environmental requirements.

### Cambridge: an innovation district with public identity

The University of Cambridge’s July 2021 West Cambridge announcement presents a future destination quarter with public spaces and a stronger interface between research and enterprise.[^8] It is evidence of a planning direction, not a comprehensive verification of subsequent construction. Its relevance is the idea that innovation activity can contribute to a recognisable place beyond individual buildings.

The NeoBharat innovation street should connect project studios, demonstration space, convening and the academic core. It should have an address, visible entrances and a useful outdoor setting. Commercial laboratories, offices and incubator space should remain distinct programme types, even where their buildings share a street. They cannot be treated as interchangeable floor area without a functional brief.

### IIT Gandhinagar: a regional reality check

IIT Gandhinagar’s 2019 campus account documents connected academic spaces, interdisciplinary movement and the importance of its terrain and drainage setting. Its discussion of the academic core and landscape is especially relevant to the organisation of shared courts.[^9] The campus is a useful Indian precedent for adapting an institutional plan to place; it is not a substitute for NeoBharat site data.

The hypothetical NeoBharat model should therefore remain replaceable at the terrain layer. When a parcel becomes available, the planning team should begin with boundary, contours, drainage and access constraints. Buildings, roads and landscape must then be repositioned together. Applying a surveyed terrain mesh beneath an unchanged architectural composition would create a misleading sense of site accuracy.

### Stanford Research Park: consistent stewardship

Stanford Research Park’s tenant handbook describes a design-review process preceding municipal review for relevant changes.[^10] This is an operational precedent for coordinated estate quality. It does not imply that NeoBharat should adopt Stanford’s legal or approval structure.

NeoBharat’s development, leasing, BOT, R&D, operations and naming pathways need a common spatial code. Future participants should receive defined plot limits, access responsibilities, signage rules, public-realm obligations and a model-delivery standard. A unified master model can record changes from different designers while preserving consistent coordinates and identifiers.

## Full-city planning framework

### Scale and land accounting

The proposed 2,700 acres equal approximately 1,092.65 hectares, or 10.927 million square metres, using 4,046.8564224 square metres per international acre. The detailed study covers 200,000 square metres: approximately 49.42 acres, or 1.83% of the full programme. It is a sample quarter within the academic ecosystem, not a miniature of the entire city.

The brief’s land programme totals 2,700 acres. These should remain gross planning envelopes. Shared streets, utilities, landscape and service areas must either be allocated inside those envelopes or explicitly reconciled through an amended schedule. They cannot be added as extra land while retaining the same total.[^1]

| Gross programme envelope | Illustrative acres | Planning treatment |
|---|---:|---|
| University core | 200 | Academic neighbourhoods and shared institutional facilities |
| Student and staff housing | 100 | Residential clusters linked to daily learning and services |
| Lakes and forest conservation | 500 | Landscape and water framework; status and extent require site evidence |
| Meditation and wellness | 40 | Quiet setting with accessible connections |
| Data and research centres | 150 | Research clusters; specialised data infrastructure separately briefed |
| Concert and conference venues | 60 | Public arrival and event servicing outside ordinary academic movement |
| Sports and solar programme | 150 | Two different land uses to be subdivided and reconciled |
| Joint-venture growth zone | 700 | Staged development plots tied to infrastructure readiness |
| Amusement and entertainment | 30 | Independent public access and buffers from sensitive uses |
| Micro-industry clusters | 200 | Serviced productive areas with use-specific environmental requirements |
| Future expansion reserve | 570 | Explicitly unresolved future land programme |
| **Total** | **2,700** | **Gross concept allocations, subject to feasibility** |

The study does not validate an enrolment or residential capacity. Capacity needs net academic floor area, timetabling, teaching methods, laboratory intensity, housing occupancy and supporting services. A visual count of buildings is insufficient. The programme should distinguish people present on campus, enrolled learners, residents, visitors and staff before assigning densities.

### Spatial organisation

The academic core should be the initial destination. Research and innovation activity should meet it along a shared street or sequence of courts. Housing should connect through continuous everyday routes, with food, recreation and convenience functions distributed along those routes. Distances should be reported from the model’s actual paths once the wider network exists, rather than promised as walking times before the plan is measured.

Culture and large events require a public edge that can operate beyond teaching hours. Quiet wellness areas need buffers from event sound, traffic and service activity. Micro-industry and high-service research should connect efficiently to logistics without turning the central campus into a through route. Data infrastructure requires a dedicated operational brief and should not be represented merely as another glazed office building.

The 700-acre growth zone and 570-acre reserve together account for a large portion of the programme. They should be shown as future development envelopes with explicit uncertainty, not as densely completed neighbourhoods invented to fill the view. Landscape should establish the continuity between early occupied areas and later plots.

### Hypothetical site convention

The quarter uses local metric coordinates, with ground at Y = 0 and a provisional north direction. It has no geographic coordinate system, cadastral identity, verified contour, geological model or surveyed watercourse. The flat base is deliberate: invented slopes and lakes should not imply knowledge of a real parcel.

A future surveyed model should record the source, date, coordinate system and accuracy of its boundary and terrain. It should also separate observed features from proposed works. Parcel access, infrastructure easements, water movement, retained vegetation and development constraints must be resolved before the current quarter is placed on a real site.

## Architectural pattern book

The following dimensions are study assumptions, not statutory limits, engineering recommendations or benchmark measurements taken from the precedents.

| Component | Initial model rule | Purpose and next design test |
|---|---|---|
| Laboratory wings | 22–26 m principal depth; four to six principal floors in this study | Establish a mid-rise research scale; replan for the actual lab type |
| General upper-floor module | 4.2 m | Reserve visible depth for a research-building section; verify services and structure |
| Ground-floor module | 5 m | Distinguish public/shared entrances from ordinary upper floors |
| Façade rhythm | 3.1 m subdivisions; larger shading/infill rhythm | Provide scale and variation; coordinate with structure and daylight |
| Learning Commons | 54 × 38 m, three principal floors | A shared civic focus with a generous sheltered approach |
| Innovation Hub | 72 × 36 m, seven principal floors plus a roof pavilion | A differentiated vertical accent requiring a separate area and access review |
| Makers Hall | 65 × 46 m, two principal floors with a taller upper module | Express a different working volume; production processes remain unspecified |
| Main perimeter carriageway | 12 m within a 27 m paved envelope | Establish a legible circulation study; redesign with a transport and access plan |
| Pedestrian scale references | People at approximately 1.75 m; initial close cameras at 1.8 m | Make façade, entry and landscape proportions assessable |

The material family should combine light mineral surfaces, shaded glazing, metal framing and restrained warm accents. Academic wings should have enough opaque façade to avoid a generic glass-office appearance. Residences, makerspaces and public pavilions need distinct proportions and environmental responses, even if they share a palette.

Each building should have a readable ground, middle and roof. Entrances should reach a path; canopies should have visible support; roof equipment should sit on a coherent plant zone. Future detailed assets should include façade returns, recessed windows, parapet thickness, drainage edges and selected interior depth at public entrances. These details add more credibility than increasing polygon count indiscriminately.

Landscape should be modelled as occupied space. Trees belong in planting areas with room for roots and canopy, not on driveways or inside building footprints. Seating should address shade and activity. Water elements should remain illustrative until catchments, supply, treatment and maintenance are defined. Tree species and maturity should be specified only after a landscape brief.

## Working model and its limits

The delivered quarter contains fifteen named building groups, 252 generic trees, a perimeter street, pedestrian routes, planted courts, a shallow illustrative basin, furniture, a shuttle, vehicles and human scale references. The principal buildings are listed in the accompanying model manifest and in the website’s building schedule. Their coordinates and dimensions are editable through the source generator.

The model has approximately 751,000 rendered triangle instances and 228 mesh batches before additional rendering passes. The browser asset is approximately 2.80 MB, including three photographic base-colour textures. An optional daylight environment adds approximately 1.44 MB. These are measured asset and geometry quantities; they are not measured frame-rate results or guarantees for every device.

Two model exports serve different uses. The browser GLB uses repeated geometry through `EXT_mesh_gpu_instancing`. A separate approximately 3.80 MB editable GLB expands the repeated objects into ordinary mesh nodes for applications that do not support that extension. Both retain named building groups and embedded base-colour textures. The editable export has many more individual objects and is intended for authoring rather than efficient browser display.

The scene uses a conventional real-time lighting approximation, reflective material parameters and photographic paving, asphalt and ground textures. It is more detailed than an abstract massing diagram, but it is not a photorealistic finished architectural render. Façade interiors, sophisticated glazing, detailed plant species, road engineering, coordinated services and construction assemblies remain outside this version. The next quality increment should focus on a small set of hero assets rather than distributing equal detail across every object.

The five starting views are the quarter overview, Learning Commons, research courtyard, innovation street and plan view. Close-view cameras start at 1.8 metres. Orbit, rotation and zoom are presentation controls; they are not a collision-aware walking simulator. The 2D plan and building schedule remain usable without the 3D engine and are the initial presentation on narrow screens.

The first quarter is a mature architectural scenario. Its internal phase identifiers are placeholders for future grouping and are not linked to a delivery calendar. It should not be described as an approved Year 10 configuration.

## Production architecture

### Authoritative design and visualisation layers

The full project needs three related but distinct datasets. The first is the planning model: boundary, terrain, land-use polygons, access, infrastructure corridors and parcel areas. The second is the architectural authoring model: building dimensions, floors, uses, openings, major systems and landscape objects. The third is the website representation, optimised for loading, selection and viewing.

Every exported object should keep a stable identifier linking it to the authoring dataset. A visualisation mesh should not become the sole source of truth for floor area or capacity. The GLB format is appropriate for delivering geometry and materials to the web, while source CAD/BIM and authoring files remain necessary for professional design development. Three.js provides dedicated glTF import and export facilities.[^11]

The proposed workflow is: confirm a spatial brief; create coordinated architectural source models; prepare UVs and physically based materials; produce appropriate levels of detail; export glTF; validate; test the actual website. Blender or another established digital-content tool can form the asset-preparation stage. The current deliverable is a GLB and reproducible source model, not a native BIM or Blender project.

### Content model

Each district should record its gross acreage, programme status, source of geometry, adjacent uses, access constraints and development state. Each building should record a stable ID, district ID, use, modelling status, local position, rotation, footprint, floor assumptions, bounding volume and source version. Model URLs, image fallbacks, captions, attribution and download sizes should be explicit fields.

Asset records should identify material textures, licences, authorship, units, origin, geometry budget and available levels of detail. Camera records should identify a target, starting position, field of view and descriptive name. Disclosure should be attached to the relevant object or scenario rather than placed only in a global footer.

The included programme JSON reconciles the eleven gross allocations and records proposed adjacencies. It deliberately supplies no fictional geographic polygons. The model manifest provides the actual quarter geometry schedule. A production content system should validate both before publishing any change.

### Levels of detail and loading

The website should not load a detailed 1,093-hectare city as one undifferentiated model. The whole-city view needs landform, district boundaries, street hierarchy and recognisable massing. A selected district can load a higher-detail package, while a named building or court can load a small set of foreground assets.

The following are proposed acceptance budgets, not completed benchmark results. They should be revised after profiling representative devices and the eventual hosting environment.

| Layer | Initial budget | Behaviour |
|---|---|---|
| Basic page and 2D city view | Target ≤1.5 MB transferred on first use | Core copy, navigation and interest pathway work before any 3D request |
| Whole-city 3D overview | Target ≤6 MB initial model payload | Coarse massing and shared instanced assets |
| Selected detailed quarter | Target ≤8 MB including lighting and required textures | Load only on selection; current study is approximately 4.24 MB for GLB plus environment |
| Foreground hero asset | Target ≤2 MB incremental per building | Load only when the view benefits from it |
| Desktop detailed view | Target ≤1 million visible triangles and ≤250 base mesh batches | Evaluate shadow and material passes separately |
| Narrow-screen default | 2D, no automatic detailed model download | Optional 3D with bounded pixel ratio and clear recovery |
| Interaction target | Sustained 30 fps or better on agreed mid-range hardware while moving | Profile real devices; reduce detail when needed |
| Idle behaviour | Render on demand | No continuous render loop for an unchanged scene |

Geometry compression and texture compression are separate decisions. Meshopt can reduce mesh transfer cost; KTX2/Basis Universal can reduce texture memory and transfer pressure when the target pipeline is configured correctly. The glTF Transform documentation supports this separation.[^12] These compression stages are proposed for the full-city pipeline; the current quarter uses embedded JPEG textures and instancing without claiming that either compression pipeline is already implemented.

### Lighting and material development

The first visual review should use neutral daylight so that geometry and proportions remain readable. A later cinematic mode can use warmer light, controlled atmosphere and selected interior illumination. A cinematic preset should never be the only setting: excessive darkness or bloom can hide planning problems and make the objects unreadable on ordinary screens.

For the next asset pass, prioritise a convincing Learning Commons entrance, one laboratory façade, an innovation ground floor and one landscape court. Add interior depth at public edges, more credible glazing, material roughness variation and detailed vegetation in those views. Establish material scale from known dimensions. Bake ambient occlusion or lightmaps where they improve depth at acceptable cost, while ensuring that time-of-day options do not contradict baked lighting.

Poly Haven’s asset licence permits reuse of the downloaded materials and environment under CC0.[^13] The accompanying provenance file records source URLs and checksums. Institutional photography and campus buildings have not been copied into the model. Future purchased or commissioned assets need their own licences and attribution records; an image search result is not an asset licence.

## Ten-year transformation model

The existing Years 1–10 interaction should ultimately control development states rather than stretch buildings upward from zero height. Buildings have proportions, and scaling an entire building distorts floor heights, entrances and façade details. Each object should transition between an unbuilt plot, a disclosed future volume, an optional construction representation and its completed architectural asset.

The scenario system should distinguish programme readiness from visual maturity. A building can be completed while its landscape is young; roads and utilities may precede occupation; a later district may remain a reserved envelope. Mature trees should not appear automatically with the first building. Exact tree sizes and growth intervals need species and landscape assumptions rather than an arbitrary global scale slider.

The first scenario pass should model dependencies: access and servicing before occupation; shared academic facilities before distant expansion; additional transport and community services as population grows. The Years 1–10 labels remain relative scenario markers without a confirmed start date. The current detailed quarter does not yet animate this sequence.

## Implementation work packages

| Stage | Deliverable | Acceptance gate |
|---|---|---|
| 1. Reference quarter | The delivered detailed quarter, manifest and editable model | Review architectural character, scale and public-space organisation |
| 2. Spatial framework | Reconciled 2,700-acre programme, access network and landscape structure | No double-counted land; provisional geography clearly identified |
| 3. Hero asset set | Detailed Learning Commons, laboratory wing, innovation ground floor and landscape court | Pass close-view review in neutral daylight |
| 4. District asset library | Differentiated academic, residential, civic, productive and recreational typologies | Each use has credible scale, entries, roof form and service assumptions |
| 5. Integrated city | District streaming, camera hierarchy, object selection and development states | Whole-city and local views agree spatially and semantically |
| 6. Release qualification | Device, accessibility, failure-recovery and model validation records | No unresolved broken routes, missing assets or blank-view failure states |

A campus architect or masterplanner should own spatial and programme decisions. A landscape designer should own planting, outdoor comfort and water strategy. Transport, infrastructure and specialist laboratory advisers should resolve their respective requirements. A visualisation artist should own asset quality; the web team should own loading, performance, interaction and fallback behaviour.

These are work packages rather than committed dates. A reliable production schedule needs an agreed asset inventory, source-model availability, number of detailed views and review cadence. Completing the first detailed quarter does not establish the effort required for the entire city.

## Quality and release requirements

### Geometry and model integrity

Verify metres, axis conventions and local origin on import. Check that building footprints do not unintentionally overlap, entrances meet paths, roofs have support, and major objects sit on the intended ground. Inspect normals, missing materials and texture paths. Check both the glTF structure and the rendered result: structural validation alone cannot detect a visually poor campus.

Khronos’s glTF Validator checks supported glTF structures and reports unsupported extensions explicitly.[^14] The browser export’s instancing extension requires an additional runtime check because the validator used here does not validate that extension. The ordinary-mesh export provides a second interoperability path. The separate QA record identifies test results and their limits without treating a validator pass as architectural approval.

### Interaction and failure recovery

Every named view must display its intended subject. Re-selecting a preset should reset its camera. Zoom and rotation must remain available through named buttons, including on touch devices. Page scrolling must remain usable. Loading state should end only after a visible frame, and failures must leave the 2D plan and building information available.

Asset loading should be cancellable when leaving the page or switching modes. The viewer must release geometry, textures, controls, rendering targets and event listeners when unmounted. Missing optional lighting should not remove the primary model. A missing model should produce a readable recovery state rather than an indefinite spinner or blank canvas.

### Accessibility, content and analytics

Keep the building schedule and disclosures in ordinary HTML. Do not place essential information solely in a canvas or tooltip. Use labelled controls, visible keyboard focus, a navigable heading structure and adequate contrast. Honour reduced-motion preferences and provide an equivalent 2D route through the programme.

The wider website retains “Register Interest” as its primary marketing conversion. Model views and downloads support evaluation of the concept. Public copy must continue to exclude financial assumptions, unsupported partner associations and claims that a hypothetical design is approved. The phrase “Envisioned as India’s largest integrated education-led knowledge city” remains an ambition, not a independently verified ranking.

If analytics are introduced, record coarse events such as model start, model ready, fallback, named view and interest-pathway entry. Do not store raw form content in event properties. The current study does not introduce an external analytics service. Avoid third-party runtime asset calls by hosting the required model resources with the site.

### Remaining design uncertainties

The parcel, topography, boundaries, developable land, final academic programme and occupancy assumptions remain unresolved. The quarter does not establish laboratory fit-out, structure, fire strategy, universal-access compliance, transport capacity, water supply or utility capacity. These should become explicit design inputs and review responsibilities as the project matures, not implied achievements in the rendering.

The immediate review decision is whether the proposed spatial character is right: a compact mid-rise research quarter, contemporary mineral-and-glass buildings, shared courts and a strong civic commons. After that direction is accepted, the next most useful investment in model quality is the four-part hero asset set and the reconciled full-city spatial framework.

## Source notes

[^1]: NeoBharat Knowledge City project brief, *NeoBharat_Knowledge_City_Codex_Website_Brief.md*, supplied project reference. Land-use schedule and early-stage positioning; date not stated. The figures are concept inputs and have not been independently verified as available land.
[^2]: MIT Capital Projects, [“238 Main Street / Kendall Square”](https://capitalprojects.mit.edu/projects/238-main-street-kendall-square), undated project page; completion listed as 2021.
[^3]: MIT News, [“MIT presents updated Kendall Square Initiative plan to City of Cambridge”](https://news.mit.edu/2016/mit-presents-updated-kendall-square-initiative-plan-city-cambridge-0107), 7 January 2016.
[^4]: ETH Zürich, [“Campus Hönggerberg” development material](https://ethz.ch/en/campus/development/hoenggerberg.html), undated living page; 2040 development horizon.
[^5]: EPFL Sustainability, [“Campus Piéton project”](https://www.epfl.ch/about/sustainability/resilient-campuses/campus-pieton-project/), undated project page.
[^6]: National University of Singapore, [“Campus” sustainability report](https://sustainability.nus.edu.sg/report/campus/), undated online report chapter.
[^7]: HOK, [“King Abdullah University of Science and Technology”](https://www.hok.com/projects/view/king-abdullah-university-of-science-and-technology-2/), undated architect project account.
[^8]: University of Cambridge, [“West Cambridge innovation district will create new destination quarter and put the science on show”](https://www.cam.ac.uk/news/west-cambridge-innovation-district-will-create-new-destination-quarter-and-put-the-science-on-show), 29 July 2021. Used as a dated planning announcement, not evidence of current completion.
[^9]: IIT Gandhinagar, [*Campus on the Sabarmati: Innovations and Challenges—Construction of a New Campus*](https://campus.iitgn.ac.in/pdf/Overview_Report.pdf), 2019; especially printed pages 6–7 on terrain and academic planning.
[^10]: Stanford Research Park, [“Tenant Handbook”](https://stanfordresearchpark.com/handbook/), undated operating guidance; design-review procedures.
[^11]: Three.js, [GLTFExporter](https://threejs.org/docs/pages/GLTFExporter.html) and [GLTFLoader](https://threejs.org/docs/pages/GLTFLoader.html), living technical documentation.
[^12]: Don McCurdy and contributors, [glTF Transform](https://gltf-transform.dev/), living technical documentation; geometry and texture optimisation capabilities.
[^13]: Poly Haven, [“Asset License”](https://polyhaven.com/license), CC0 asset terms. Exact assets are listed in the delivered provenance JSON.
[^14]: Khronos Group, [glTF Validator](https://github.com/KhronosGroup/glTF-Validator) and [Node usage](https://github.com/KhronosGroup/glTF-Validator/blob/main/node/README.md), living implementation documentation.

## Sources

The sources above were consulted on 9 September 2026. Undated living pages should be revisited before relying on their operational details in a later project stage.

| Reference | Publisher / document | Date or status | Scope used |
|---|---|---|---|
| 1 | NeoBharat project brief | Supplied, undated | Programme and positioning inputs |
| 2 | MIT Capital Projects: 238 Main Street | Completed 2021 | Built public/research interface |
| 3 | MIT News: updated Kendall Square plan | 7 January 2016 | Architectural coordination and variation |
| 4 | ETH Zürich: Hönggerberg development | Living page; 2040 horizon | Compact growth and public-space framework |
| 5 | EPFL: Campus Piéton | Living project page | Pedestrianisation and service access |
| 6 | NUS: sustainability report, Campus | Online chapter | Shade, landscape and outdoor comfort |
| 7 | HOK: KAUST | Architect project account | Research neighbourhoods and shared facilities |
| 8 | Cambridge: West Cambridge announcement | 29 July 2021 | Proposed innovation-district identity |
| 9 | IIT Gandhinagar: Campus on the Sabarmati | 2019 | Terrain response and academic connections |
| 10 | Stanford Research Park: Tenant Handbook | Living guidance | Estate design review |
| 11 | Three.js loader/exporter documentation | Living documentation | Web model interchange |
| 12 | glTF Transform documentation | Living documentation | Proposed optimisation pipeline |
| 13 | Poly Haven licence | Current asset terms | Material and environment reuse |
| 14 | Khronos glTF Validator documentation | Living implementation | Structural validation and its limits |