What is Into Yesterday ?
A maze at Barangaroo that walks you back into a childhood summer.
The problem of building a feeling, not a picture.
How do you take someone back to a summer they can only remember?
The brief was immersive design. The thing I actually wanted to build was the feeling of a summer that has already happened and can only be returned to — not summer, but the idea of the perfect summer.
Where does it stand?
On the steps that lead down into the water at Barangaroo Reserve. Open views across the harbour, and heavy, mixed foot traffic — families, tourists, office workers on their way home.
What is a memory made of?
Not a picture. A temperature, a sound, the length of the grass.
How would I know it was working?
I picked something I had a personal connection to, so I could tell. The moodboard was the target, and I kept checking the renders against it.
“It doesn’t show you anything. It changes the air until you remember.”
The idea, in one line
How did Into Yesterday solve this?
Nothing in the maze is a display. Everything is a dial.
Light, grass, air temperature, how close the music is — and every dial turns the same way the further in you go.
What I did
Mine end to end — the concept, the site logic, the maze, the lighting, the 3D and the soundscape. UNSW coursework in Term 2 of 2025, set as a Vivid Sydney proposal for Barangaroo Reserve and not submitted. Modelled in Blender, with the first music I’ve ever made.
So how does a maze walk you back into a summer?
Let me walk you
through it.
What happens when I walk in?




Where does it come out?
At a waterside clearing where glowing orbs — memories — float gently on the surface. You step out, the lights stay behind you, and only the orbs are left, bobbing on the water.
Reflection — the orbs on the water
What changes as I go further in?
- LightDim glow → full bloom
- GrassAnkle → past the knee
- AirHarbour cold → summer warm
- SoundBarely there → resolved
Design intent across zones 1–4, not measured values
It’s the same instinct as the EMF lockup and the CSEDS shell: build the rule, not the piece. Set one thing that never moves, then let everything else be a value you change.
Here the thing that never moves is the direction. Every dial turns the same way, so the further in you walk, the further back you go.
01 Low immersion
Faint music. Short grass. Air barely warmer than the harbour.
02 Mild immersion
Lights brighter, breeze warmer, the music a little louder.
03 High immersion
Grass past the ankle. Summer air. Familiar, and none of it yours.
04 Reflection
You step out. The lights stay behind you. Only the orbs, bobbing on the water.
Where the dials let go
The finished maze, lit, beside the water
How do I know which way to go?
You follow the music. It plays faintly through the whole maze — you walk towards it without being told to — and only resolves when you come out and see the orbs.
The brief I gave it was “the feeling of rediscovery.” It is the first piece of music I have ever made. The soundtrack is the wayfinding.
At the entrance
Barely there.
Through the zones
A little closer each time, never explained.
At the water
Resolved, as the orbs come into view.
Where the music arrives
How many people get to walk it?
- Route~90 m, walked slowly
- Pace~0.6 m/s — a 2½-minute transit
- Inside~25 at a time
- An hour500–600
- A night3,000–3,500, over six hours
My estimates from the plan, not measured figures
The single entrance does two jobs. It meters the crowd, and it means nobody is ever walking towards you — you get the thing to yourself even though you queued for it.
The throughput is the number I would put in front of a festival operations team first, because it is the one that decides whether the piece is viable at all. Non-slip paths and wheelchair access throughout.
The maze on the steps at Barangaroo Reserve, one entrance in
That’s the walk.
But none of it existed until I made it badly first.
The first maze felt right and was structurally a mess — broken geometry everywhere. So I rebuilt it a different way and re-lit it with the grass in place. That was the lesson of the whole project: I stopped trying to decide on paper whether an idea would work, and started making it badly to find out.

Where the language came from
Before this I made nine mixed-media tiles pulling apart the Skyfall title sequence — one per narrative code. Paper, clay, foil and paint, all by hand.
I didn’t carry the meanings across. I carried what they taught me to notice: that warmth reads as safety, that layers can be a person, that repetition with small variation is how a thing gets a rhythm.

The target
Soft sunset skies, warm summer breeze, gentle reflections off the water — my own illustration of the feeling being aimed at.
The reference was Rain Room by Random International: an environment that responds to you being in it, rather than performing at you.
I over-plan, and this course left no room for that. I started the concept several ways, and each time the mistake showed up in the making, not the thinking.
Decide, then make — how I started
- Plan it
- Decide it works
- Model the maze
- Find out
Rejected. The first maze felt right and was structurally a mess.
Make it badly, then decide
- Make it
- Find out
- Rebuild
- Re-light, with the grass
Taken. A faster and more honest answer than any amount of thinking about it.
Where the mistake becomes obvious

Grey box

In Blender
The maze in grey, with a camera rig, and the orbs blocked out on a plane — before any light went in.

The poster
Barangaroo, with yesterday leaking into it.
What went over
It was awarded a High Distinction. The more useful outcome was the working method it forced on me: stop deciding whether it will work, start making it badly and find out.
The other thing that worked: letting an idea sit for a day before touching it again.
What I’d do differently
This is coursework, and it stops where coursework stops. Nobody has walked through it — there is no prototype of a zone, no test of the soundscape on a listener, and the throughput figure is arithmetic from my own plan rather than an observed rate. The animation is also limited by the machine I made it on.
If it went further, the first thing I’d build is not a better render. It’s one corridor at full height with the light, the grass and the heat running, and ten people walking it.