THE MADRAS PROJECTசென்னை

A letter from the project · 11 November 2026

Many places.
One Chennai.

Research, films, a game and a world to explore. An ongoing attempt to understand the city I call home.

By Vinay Chandrasekaran
Founder of Layerpath · Tamil Nadu-born, San Francisco-based

After Madras Day, I found myself asking what was already here before the familiar story of 1639. I wanted a picture of how Chennai accumulated: its settlements, waterways, work and people.

I started reading. Then drawing, making films, building a game and assembling a historical world. The forms kept changing because each let me ask something different of the city.

A kolam can carry a story through a line. A game can make someone stop at a record. A world lets them look from a beach towards the houses and ask who lived there.

I want the work to feel recognisable to someone who knows Chennai, and inviting to someone who does not. I want curiosity to survive the first beautiful image. If someone asks, “How do we know?”, there should be a source to follow and room to disagree.

That is what I am trying to make with The Madras Project.

Vinay

The current historical world preview: boats and standing figures on the sand, with a settlement and palms behind them.
Plate I Coast, c.1700. A view from the desktop preview published on 6 November 2026. A researched reconstruction with illustrative detail.

I · The work

Four ways into the city

FilmChennai, RedrawnA story drawn in code through one continuous kolam line. FilmA line through MylaporeA neighbourhood explored through streets, a tank, a kolam and a tower. PlayMadras RushA browser game with historical records to discover. Explore mode gives you time to read. VisitThe historical worldCoast, settlement and fort across c.1630, c.1645 and c.1700.

More from the studio

The early Madras Project YouTube channel brings the films and ongoing explorations together. It is another way to follow what I am making as the project finds its own home.

II · The making

From a source to something you can see

I use AI to help read, organise research, write code and make assets. I set the questions and direction, examine the results, and decide what is ready to share. The films, game and world have separate implementations. The historical world is the foundation I want to reuse for more views and experiences.

An overhead view of the fort and surrounding buildings in the c.1700 state of the current world preview.
Plate II Fort, c.1700, from the current visit. Placement and geometry are deliberate choices; a detailed model is not proof of exact historical dimensions.
01Buildings: an image is only a beginningDifferent meshes can come from the same drawing. They still need inspection.

Some assets begin as generated images and are converted into meshes. Other geometry, including the fort, is authored directly. Code controls where things sit and when they appear. The process varies by asset.

A September comparison sheet showing one hut illustration and the different 3D models produced from it, viewed from several angles.
A September 2026 production study. The drawing and meshes are illustrative assets, not photographs or archaeological measurements. Open the image to inspect the original comparison sheet.

The useful question is whether a building belongs in the scene, looks right from more than one angle, and can be explained. A successful conversion does not answer all three.

02Characters: what can we honestly show?The record can support an occupation without telling us a person's face or clothes.

The project separates documented features, reasonable inferences and illustration. For example, a character walking between home and work is an interpretation of daily life. The precise individual, route, costume and time of day are not established.

An early study render of a woman carrying a pot on her head.
A September character study. The woman with a head pot is an illustrative choice. Her identity and what the pot contains are not established by the research.

This is why I want a visitor to be able to ask about the evidence behind a scene. The illustration should invite a question, not hide the uncertainty.

03Rigging and motion: watch the whole personA convincing still can hide a broken walk. Movement needs its own review.

A rig gives a mesh a joint structure so it can move. In the early boatman experiment, animation was transferred to the character in Blender. The diagram below comes from that actual rig file; it is a production artefact, not an anatomical or historical diagram.

Front and side projections of the actual early boatman rig, showing its joint hierarchy in its stored rest pose.
Joint positions and parent connections extracted from the September boatman rig. A new diagram of an existing production asset.
Eight sampled poses from the boatman's isolated walk cycle.
The walk cycle studied across time. Later review found that in-world movement could introduce faults absent from the isolated clip.

A later net-mending experiment became trapped in repeated repairs. That version did not pass review and was left out of the published visit. The lesson was to review the whole moving body and the visitor's experience, and to stop a failed experiment from holding up a bounded release.

III · Research notes

Three questions I keep returning to

The research began as ten notes. These are the threads I would keep in the foreground. Each has a source and a limit.

What was here before the familiar beginning?Madraspatnam already existed before the English settlement.

Greater Chennai Corporation's history distinguishes the earlier village of Madraspatnam from the town growing around the fort. It also describes competing explanations for the names. That is a more useful starting point for this project than treating a single date as the beginning of every place and community.

This is a summary of the published account, not a claim that the naming debate is settled. Read GCC's account, especially “The Names Madraspatnam and Chennapatnam”.

What does the date on the page actually say?Love prints 22 July 1639 and questions it. The interpretation needs to stay visible.

On page 17 of Vestiges of Old Madras, H. D. Love says only the last of the copies he discusses bears a date. He prints 22 July 1639, then suggests that July may be a slip for August. In 2023, V. Sriram argued for the July reading and described the copies differently.

My earlier note compressed those accounts into one confident statement. They should be attributed separately. This page does not resolve the date of the original grant.

An excerpt from Love, volume I, page 17, stating that only the last copy bears a date, 22 July 1639, and suggesting July is a slip for August.
H. D. Love, Vestiges of Old Madras, volume I, page 17 (1913). Public-domain text; excerpt from the project's existing scan.

Love, volume I · V. Sriram's account and argument

Can today's map stand in for an earlier city?Waterways and built places have changed. Each date needs its own evidence.

GCC's history describes the North River joining the Cooum and a later cut that created the Island ground. Those changes matter when drawing an earlier landscape. A modern map alone cannot tell me where every historical feature belonged.

The project has not yet fully aligned a dated historical map with a modern survey. The reconstructed scene should be read with that limit in mind. GCC's account of the first settlement provides the river description. The detailed research record remains in the working archive; this letter shares selected findings and the sources behind them.

Follow the historical sources linked within these notes for deeper reading. The detailed working archive is private. Models can help find questions. Agreement between models cannot establish a fact.

IV · The public ledger

The decisions that changed the work

Selected milestones and lessons. The detailed session records remain the working archive.

A picture became a questionAn early Madras Day image exposed how easily a convincing picture could get the city wrong.

The intended image was held back. Reading about the names, rivers and fort became part of making the work. The question changed from “Can I make this picture?” to “Can I explain what is in it?”

A world assembled from partsPlacement moved into code. Buildings, water, terrain and cameras could be revised separately.

The whole-scene image experiments were too difficult to control. A reusable 3D scene made revisions more deliberate, but introduced new questions about scale, lighting and what the visitor could see from the ground.

People changed the testWatching the scene move revealed problems that attractive stills and numerical checks had missed.

Figures made scale easier to judge. Playback exposed faults in walking and turns. The useful change was to assess people beside boats, buildings and trees, and to inspect motion across time.

Mylapore enters the gameMadras Rush's published update adds a Mylapore route and a kolam of the ride.

The game offers another way to explore place through play. Its work page records this update and credits community suggestions behind several features. That feedback is useful; it is not yet a measure of learning.

Stop the repair loopThe net-mending experiment remained unresolved. It was excluded so the guided visit could move forward.

The hand problem was connected to motion elsewhere in the body. Repeating local repairs did not establish a successful result. The experiment remains separate from the published preview; it is not presented as a finished interaction.

A visit people can tryThe desktop preview was published with coast, settlement and fort views, three dates and a reading option.

The coast opens beside the boats; the settlement opens inside a lane. The download was reduced, and the visit gained a clearer bounded shape. Open the published preview.

The project meets a new readerThe New Indian Express features the work. It is a moment to explain the purpose, process and next steps more clearly.

Sonu M Kothari's feature brings an outside account of the project. This letter gathers the context I had promised to write, including the distinction between what is available today and what I hope to build.

V · What comes next

More ways to follow your curiosity

I want visitors to move closer, notice how people worked and understand how the city changed. Conversations with characters, building entry and role-play belong to that longer ambition. They need clear boundaries between recorded history, interpretation and imagined dialogue.

The next step is to make the existing visit easier to understand and use, invite corrections, and choose the next interaction around what a visitor wants to discover. I want the evidence to become easier to inspect as the experience becomes richer.

Over time, I would like useful parts of the method and work to be reusable for other places. An open-source release is an intention. A complete public toolkit and blanket reuse licence are not available yet.

The wider threadTamil Nadu Desk, place memory and other explorations.

Tamil Nadu Desk is a separate effort to make public information about the state easier to explore. My habit of recording places through Swarm check-ins comes from a similar interest in place and memory. These are related interests, not features of the Madras world.

VI · Questions

A few things worth making clear

Are you a historian?

No. I am the founder of Layerpath, and this is a personal creative and research project. I read sources, record uncertainty and welcome corrections. I am not presenting the published world as having completed an external historian review.

How accurate is the world?

The evidence is stronger for some features than others. Documented facts, inferences and illustrative choices need to remain distinguishable. There is no defensible single accuracy score for the whole scene. The sources linked in the research notes, and the limits stated with them, are better starting points.

What did you make, and what did AI do?

I supplied direction, questions, constraints and feedback. AI assistants helped with research and code; generation tools produced particular assets. I selected, rejected and revised the outputs. Each work has its own process and credits. I take responsibility for what I publish.

Can I talk to characters or explore freely today?

The public world currently offers a guided visit with view inspection, date changes, sources and sharing. Conversations, free walking and building entry remain future possibilities. Madras Rush is a separate browser game with a different timeline and interaction.

Is this a Layerpath product? What is the business model?

The Madras Project is a personal project, separate from Layerpath. There is no commercial offer announced here. My focus is making work people want to explore and explaining how it was made.

Can I reuse the work?

A complete open-source package has not been released. Source documents, photographs, maps, code and generated assets can have different terms. Credits are not blanket permission to reuse everything. Follow the specific attributions on each work page and ask about a particular use.

How much time and money has it taken?

I am keeping this public ledger focused on milestones and lessons. I am not publishing current financial records or quoting a project-wide total here. Older figures should be read as dated snapshots, not current accounts.

How can someone help?

Send a source, a correction, a memory or an idea through the contribution page. For a correction, name the scene or record and include a page or date reference if you have one. Personal memories can open a question; they should stay distinct from evidence for a historical claim.

A note of thanks

To Sonu

Thank you for taking the time to listen and for writing about the project in The New Indian Express, online and in print. Seeing the work through your account helped me understand what I had communicated and what needed more explanation.

I had promised a write-up and FAQ and did not get them to you in time. This letter is a start. I am grateful you made room for the curiosity behind the work.