I thought the hard part was going to be the geometry.
It was hard, sure. But once the mountain shapes started becoming workable, the project revealed its real personality: molds, wax, curing, shrinkage, leakage, demolding, lettering, and all the little physical details that do not care how elegant the code looked five minutes ago.
This was the phase where the project stopped being a “cool terrain candle” idea and became a proper manufacturing problem.
The geometry pipeline was getting more useful in spring 2025. wax-cam was under version control, terrain workflows were getting cleaner, and I was starting to build reusable pieces instead of only mutating notebooks in place.
At about the same time, the notes got more blunt.
One of the cleanest turning-point lines is:
Aluminum doesn’t scale, let’s try silicone molds
That sentence is doing a lot of work.
It means the first physical strategy was no longer good enough. It also means the next phase of the project was going to be ruled by mold engineering and material behavior whether I liked it or not.
The product had to satisfy a pretty rude stack of requirements all at once:
Unfortunately, every one of those requirements could undermine the others.
A shape that looked good on screen could be fragile in wax. A mold that captured detail could become miserable to demold. A process that reduced one defect could introduce another.
Classic project energy.
This is where the Dropbox logs become gold.
CandlePours.md is basically a lab notebook of the project arguing with itself in real time:
At the same time, Molds.md shows the mold lineage and ties later molds directly to wax-cam revisions.
That combination matters because it lets me see both sides of the process:
And the notes fill in the emotional and decision layer:
2025-05-04 — soy wax in aluminum mold, test 3D print silicone2025-05-05 — silicone mold did not cure properly overnight2025-05-19 — silicone pour #4 notes in Dropbox2025-05-20 — revising molds, shorter text, proud of mountain2025-06-17 — overwhelmed; still trying to understand what happens when the candle burns and whether patterning affects curingThat last one is important. Even after a lot of progress, the system was still not fully understood.
The mold lineage tells a pretty specific story.
Early molds are more generic and material-defined:
Later molds get much more coupled to the software:
wax-camThat means the project slowly built a bridge between code lineage and object lineage.
That is a big deal.
Once the mold entries include wax-cam commit references, I am no longer just making candles and hoping I remember which version did what. I am actually tracking artifact provenance.
The pour log also changes character over time.
Early pours read like exploratory troubleshooting:
Later pours get more procedural:
That is a meaningful shift. It is the difference between:
The process still fought back, but it was at least being measured now.
A few concrete things changed in this phase:
No more pretending it was just downstream implementation.
Wax chemistry, scent load, cooling behavior, and demolding reality all started pushing directly on geometry and tooling.
There is still a lot of mountain-poetry around the edges, but the core work here is temperatures, failures, measurements, and revised procedures.
If a geometry change made the object harder to cast or release, that was not a purely aesthetic choice anymore.
I do not think this phase ends with a neat “and then molds were solved.”
What it does produce is a much better understanding of the actual adversary.
The hard part was not only pulling mountains into geometry. The hard part was getting all of these to line up at once:
That is a real product-development problem, not a novelty prototype problem.
The move away from aluminum and toward silicone / printed strategies brought new flexibility, but also new failure modes:
More flexibility up front often means more fuss later.
That sounds bad, but honestly I think it was the correct trade. The aluminum path was cleaner conceptually, but it does not appear to have been teaching me the right lessons at the right speed.
This phase made the later project possible in a few different ways:
Most importantly, it killed any illusion that the project could be solved by software alone.
Good.
That illusion needed to die.
Once the project had working geometry and a less-naive understanding of molds and wax, the next problem was repeatability.
I did not just need to make interesting objects. I needed a workflow that could survive iteration without burying me in reruns, STL clutter, and inconsistent product shape.