Search

https://solidean.com/blog/2026/solidean-hello-blender/

Philip Trettner @artificialmind · 2026-05-13 · 13:45 UTC

Blender's python API is finally fast enough that interactive exact mesh booleans are feasible without forking:

(This is pretty minimal for now: single-shot booleans, no attributes preserved, but has live preview & connected output topology)

Download (drop-in): https://solidean.com/download/blender-addon/
Github (addon layer): https://github.com/solidean/solidean-blender-addon

https://solidean.com/blog/2026/solidean-hello-blender/

Philip Trettner @[email protected] · 2026-05-13 · 13:45 UTC

Blender's python API is finally fast enough that interactive exact mesh booleans are feasible without forking:

(This is pretty minimal for now: single-shot booleans, no attributes preserved, but has live preview & connected output topology)

Download (drop-in): https://solidean.com/download/blender-addon/
Github (addon layer): https://github.com/solidean/solidean-blender-addon

https://solidean.com/blog/2026/solidean-hello-blender/

Philip Trettner @[email protected] · 2026-05-13 · 13:45 UTC

Blender's python API is finally fast enough that interactive exact mesh booleans are feasible without forking:

(This is pretty minimal for now: single-shot booleans, no attributes preserved, but has live preview & connected output topology)

Download (drop-in): https://solidean.com/download/blender-addon/
Github (addon layer): https://github.com/solidean/solidean-blender-addon

https://solidean.com/blog/2026/solidean-hello-blender/

#modeling #geometry #b3d #blender

Philip Trettner @[email protected] · 2026-05-13 · 13:45 UTC

Blender's python API is finally fast enough that interactive exact mesh booleans are feasible without forking:

(This is pretty minimal for now: single-shot booleans, no attributes preserved, but has live preview & connected output topology)

Download (drop-in): https://solidean.com/download/blender-addon/
Github (addon layer): https://github.com/solidean/solidean-blender-addon

Philip Trettner @artificialmind · 2026-05-01 · 08:26 UTC

I've added some light "automated proving" capbilities to my geometric predicate generator.

(Context is still high performance mesh booleans, but this is applicable to so much more geometric computation)

With enough symbolic perturbation I was able to prove that the perturbed query point cannot lie on the plane.

This is not done via proof search but via rewriting / optimization.

Basically, "prove <expr>" is definitely true if we can simply "<expr>" to "true".

Philip Trettner @[email protected] · 2026-05-01 · 08:26 UTC

I've added some light "automated proving" capbilities to my geometric predicate generator.

(Context is still high performance mesh booleans, but this is applicable to so much more geometric computation)

With enough symbolic perturbation I was able to prove that the perturbed query point cannot lie on the plane.

This is not done via proof search but via rewriting / optimization.

Basically, "prove <expr>" is definitely true if we can simply "<expr>" to "true".

Philip Trettner @[email protected] · 2026-05-01 · 08:26 UTC

I've added some light "automated proving" capbilities to my geometric predicate generator.

(Context is still high performance mesh booleans, but this is applicable to so much more geometric computation)

With enough symbolic perturbation I was able to prove that the perturbed query point cannot lie on the plane.

This is not done via proof search but via rewriting / optimization.

Basically, "prove <expr>" is definitely true if we can simply "<expr>" to "true".

Philip Trettner @[email protected] · 2026-05-01 · 08:26 UTC

I've added some light "automated proving" capbilities to my geometric predicate generator.

(Context is still high performance mesh booleans, but this is applicable to so much more geometric computation)

With enough symbolic perturbation I was able to prove that the perturbed query point cannot lie on the plane.

This is not done via proof search but via rewriting / optimization.

Basically, "prove <expr>" is definitely true if we can simply "<expr>" to "true".

#proglang #proof #geometry

Philip Trettner @[email protected] · 2026-05-01 · 08:26 UTC

I've added some light "automated proving" capbilities to my geometric predicate generator.

(Context is still high performance mesh booleans, but this is applicable to so much more geometric computation)

With enough symbolic perturbation I was able to prove that the perturbed query point cannot lie on the plane.

This is not done via proof search but via rewriting / optimization.

Basically, "prove <expr>" is definitely true if we can simply "<expr>" to "true".

#booleans #geometry #godot #mesh

Philip Trettner @artificialmind · 2026-04-30 · 15:32 UTC

It's kinda meditative to see a wooden bunny auto-carved in AR.

I'll try to figure out how to do video captures I promise :D"

(Still fiddling with the high-perf exact mesh booleans demo on the Quest 3...)

#geometry #ar #vr #quest3 #godot

Philip Trettner @artificialmind · 2026-03-23 · 13:09 UTC

Real-time interactive milling simulation using exact mesh booleans (~1 ms per step).

I teased this on the weekend. This is super hard for various reasons.

The sphere is actually sweeped each step, so we have a unique ~300 tris mesh each time. No preprocessing possible.

Then the sweep starts at the exact location the previous ended, meaning ~100 coplanar tris each time.

The workpiece accumulates triangles away from the "action" which must not affect perf.

Philip Trettner @artificialmind · 2026-03-21 · 14:37 UTC

Interactive drilling is a lot of fun.

Sneak peak of a new in progress demo, where you can control a 3 axis drill (various shapes) via point-and-click. The drill then moves to your target location, subtracting sweeped tool meshes in ~20 hz frequency.

This is all done using exact mesh booleans and takes roughly 1ms per step in most cases.

(Rendering/UI via godot again)

Philip Trettner @[email protected] · 2026-03-21 · 14:37 UTC

Interactive drilling is a lot of fun.

Sneak peak of a new in progress demo, where you can control a 3 axis drill (various shapes) via point-and-click. The drill then moves to your target location, subtracting sweeped tool meshes in ~20 hz frequency.

This is all done using exact mesh booleans and takes roughly 1ms per step in most cases.

(Rendering/UI via godot again)

Philip Trettner @[email protected] · 2026-03-21 · 14:37 UTC

Interactive drilling is a lot of fun.

Sneak peak of a new in progress demo, where you can control a 3 axis drill (various shapes) via point-and-click. The drill then moves to your target location, subtracting sweeped tool meshes in ~20 hz frequency.

This is all done using exact mesh booleans and takes roughly 1ms per step in most cases.

(Rendering/UI via godot again)

Philip Trettner @[email protected] · 2026-03-21 · 14:37 UTC

Interactive drilling is a lot of fun.

Sneak peak of a new in progress demo, where you can control a 3 axis drill (various shapes) via point-and-click. The drill then moves to your target location, subtracting sweeped tool meshes in ~20 hz frequency.

This is all done using exact mesh booleans and takes roughly 1ms per step in most cases.

(Rendering/UI via godot again)

#booleans #milling #geometry #godot

Philip Trettner @[email protected] · 2026-03-21 · 14:37 UTC

Interactive drilling is a lot of fun.

Sneak peak of a new in progress demo, where you can control a 3 axis drill (various shapes) via point-and-click. The drill then moves to your target location, subtracting sweeped tool meshes in ~20 hz frequency.

This is all done using exact mesh booleans and takes roughly 1ms per step in most cases.

(Rendering/UI via godot again)

#godot #geometry #mesh #booleans

Philip Trettner @artificialmind · 2026-03-16 · 12:56 UTC

I really like how this solidean demo is coming together (interactive iterated mesh booleans). This still bottlenecks at creating the rigid bodies with custom collision shapes, but Godot + Jolt are great otherwise.

Philip Trettner @artificialmind · 2025-10-15 · 15:30 UTC

In-progress video of our Solidean <-> Godot integration. This time a few more shinies, some chunking to optimize physics integration.

As you can see, I should maybe implement some connected component detection to get rid of the floaties.

(engine is Godot, method is https://solidean.com/, statue from Karl_Williams)

#Godot #csg #booleans #geometry #solidean

#godot #csg #booleans #geometry #solidean

Philip Trettner @artificialmind · 2025-10-14 · 12:55 UTC

Wrote a new blog post: https://solidean.com/blog/2025/sfinae-concepts-static-assert-modern-cpp/

We're modernizing some of our internal C++ libraries and I looked at how we want to move SFINAE over to concepts/requires.

spoiler: it's not gonna be "requires requires"

#cpp #sfinae #templates

Philip Trettner @artificialmind · 2025-09-16 · 13:02 UTC

New blog post: https://solidean.com/blog/2025/the-vimpl-pattern-for-cpp/

Basically PImpl, but with a pure virtual header. It's a neat pattern I like. It hides dependencies but at the same time has the ergonomics of "raw c++".

(In a nutshell: pimpl without boilerplate)

#cpp #devblog

Philip Trettner @artificialmind · 2025-09-01 · 10:01 UTC

I finally have a site for my mesh booleans! It's been almost 3 years since my EMBER paper. We basically took the "imul" i64 x i64 -> i128 instruction and built a geometry kernel out of that.

So, if you like robust, exact, high-performance mesh booleans (for graphics, VFX, 3D printing, CAE, ...), you might find this interesting: https://solidean.com/

(This is my commercial project, though we want to open-source the exact predicates & constructions in the future)

Philip Trettner @[email protected] · 2025-09-01 · 10:01 UTC

I finally have a site for my mesh booleans! It's been almost 3 years since my EMBER paper. We basically took the "imul" i64 x i64 -> i128 instruction and built a geometry kernel out of that.

So, if you like robust, exact, high-performance mesh booleans (for graphics, VFX, 3D printing, CAE, ...), you might find this interesting: https://solidean.com/

(This is my commercial project, though we want to open-source the exact predicates & constructions in the future)

Philip Trettner @[email protected] · 2025-09-01 · 10:01 UTC

I finally have a site for my mesh booleans! It's been almost 3 years since my EMBER paper. We basically took the "imul" i64 x i64 -> i128 instruction and built a geometry kernel out of that.

So, if you like robust, exact, high-performance mesh booleans (for graphics, VFX, 3D printing, CAE, ...), you might find this interesting: https://solidean.com/

(This is my commercial project, though we want to open-source the exact predicates & constructions in the future)

Philip Trettner @[email protected] · 2025-09-01 · 10:01 UTC

I finally have a site for my mesh booleans! It's been almost 3 years since my EMBER paper. We basically took the "imul" i64 x i64 -> i128 instruction and built a geometry kernel out of that.

So, if you like robust, exact, high-performance mesh booleans (for graphics, VFX, 3D printing, CAE, ...), you might find this interesting: https://solidean.com/

(This is my commercial project, though we want to open-source the exact predicates & constructions in the future)

#meshing #graphics #geometry

Philip Trettner @[email protected] · 2025-09-01 · 10:01 UTC

I finally have a site for my mesh booleans! It's been almost 3 years since my EMBER paper. We basically took the "imul" i64 x i64 -> i128 instruction and built a geometry kernel out of that.

So, if you like robust, exact, high-performance mesh booleans (for graphics, VFX, 3D printing, CAE, ...), you might find this interesting: https://solidean.com/

(This is my commercial project, though we want to open-source the exact predicates & constructions in the future)