Why I Started Sharing Geometry Fixes Every Month
I kept running into the same geometry problems in my workflow. Bad normals, overlapping faces, messy topology after boolean operations — stuff that should not be this hard to fix. So around three years ago, I started posting little solutions every month. The format stayed simple: one problem, one clean fix, sometimes two. What began as personal notes turned into something other people actually use, and now it goes by Monthly Geometry Hacks. The core idea is practical, not theoretical. Each post focuses on a specific geometry pain point that comes up during actual modeling, CAD work, or mesh processing. You will not find essays on computational topology here. The content is built around real files, real tolerances, and real failure modes. Take mesh cleanup after a boolean operation. This is the most common problem people face when using CSG in Blender, FreeCAD, or Fusion 360. The result is almost never clean. Faces overlap, vertices are slightly off, and normals flip depending on which side of the mesh you are looking from. The hack I posted most frequently is straightforward: run a boolean, import the result into a dedicated remeshing or retopology pass, then use a combine mode set to Merge by Distance with a threshold calculated relative to your model's bounding box size rather than a fixed value. A threshold of 0.0001 works for a model that is 1 meter across. It fails instantly if your model is in millimeters and you are working at a 200mm scale. I learned that the hard way on a precision bracket project where the merge distance collapsed thin walls entirely.
That bracket job took me four hours to debug because the normals were flipping on alternating faces after the export from the CAM software. The fix was not in the modeling software at all. It was a sign convention issue between the renderer and the toolpath generator. I ended up running the final mesh through a quick orientation repair script that checked face winding order against the average normal direction of adjacent faces. The script itself is about thirty lines in Python using the trimesh library. Anyone who has dealt with this kind of issue knows that third-party scripts like that are worth keeping around. I include the scripts in most of the Monthly Geometry Hacks posts because the actual implementation matters more than the concept.
What the Hacks Actually Cover
The posts span a range of topics. Some are about NURBS surface continuity when lofting between profiles of different complexity. Others deal with mesh decimation artifacts in hard-surface workflows. A few are purely about coordinate system confusion, which is surprisingly common when people mix local and global frames in assembly modeling. One of the posts that got the most traction dealt with UV unwrapping artifacts caused by non-planar seam edges. The fix involved snapping the seam vertices to an average plane before marking edges as seams. That sounds obvious in retrospect, but nobody warns you about it until you spend an afternoon wondering why your texture stretches weirdly near a seam. Another post covered a situation where people try to use a single subdivision surface modifier across an entire model that has both organic and mechanical regions. The subdivision smooths away the hard edges you actually need. The workaround is to use crease values weighted by edge angle, not by arbitrary numeric input. Beginners usually set crease to 0.8 across the board and then wonder why the model looks like a softened toy. The angle-based approach uses the actual geometric data to decide where sharpness belongs.
Get the Full Details
How to Use These Resources
You can find the archive at monthlygeometryhacks.com. It is free. There is no paywall or subscription tier. I host it myself because the bandwidth is low and I do not want a platform deciding what gets promoted or buried. The site is organized by software category and problem type. Each post includes the source file, the output file, and a short log of what went wrong during the initial attempt. That last part is important. Most tutorials show only the successful path. The failed attempts are where the actual learning happens. For people who work in Grasshopper or Blender, the files include the .gh definition or .blend project with the relevant steps isolated. I usually strip out everything unnecessary so you are not scrolling through a thousand nodes to find the one that matters. If you are using ZBrush, the posts focus on retopology workflows and DynaMesh limits rather than sculpting techniques. ZBrush handles geometry differently and most generic advice does not apply cleanly to its engine.
Common Mistakes People Make With These Hacks
The biggest issue I see is people applying a fix meant for one software to another without adjusting the parameters. A Merge by Distance threshold from a Blender workflow will break in Rhino if you paste it directly. Rhino uses a different precision model and a different interpretation of vertex proximity. I have included a software-specific note on every post to prevent exactly this kind of cross-contamination, but people still do it. Another mistake is skipping the backup step. Running a destructive boolean on an unversioned file and then trying to reverse engineer the original shape is not a hack. It is just regret with extra steps. There are also limits to what these posts can address. If you are working with extremely high-polygon sculpts above fifty million faces, most of the mesh processing tools become impractical regardless of the fix. The hacks assume a reasonable polygon budget. Similarly, if your geometry involves parametric dependencies that change daily, the exported static meshes will not reflect later updates unless you redo the fix. Nothing about this process is automatic once the geometry is baked. That is a constraint worth noting upfront rather than discovering after you have spent time following a tutorial. If you want the full collection, the site is straightforward and the downloads are direct. There is nothing fancy about it. The content is what matters, and the content exists because people keep sending me messages about geometry issues that should not be this difficult. I answer them the same way: with files, scripts, and honest explanations about where the process breaks.
