Take two beams printed from the same material, on the same printer, to the same geometry, and load them the same way. According to test data published by Novineer, one can bend six times more than the other. The only difference between them is the internal toolpath. That claim comes from Novineer CEO Ali Tamijani, writing in a guest column for TCT Magazine on October 12, 2026, titled "Your FDM part doesn't exist until you slice it." The column summarizes a test campaign Novineer has run with Stratasys. Because the company that sells toolpath-aware software is the one reporting the numbers, they should be read as Novineer's own findings, not as independent results. They are still worth a close look.
The Thesis: Geometry Is Not the Part
Tamijani's argument fits in his headline. In fused deposition modeling, a CAD model describes an envelope. The real part, with its stiffness, its strength, and the places where it will fail, isn't settled until a slicer decides where every bead of material goes. Two parts can share an STL file and behave like different objects.
Most experienced makers already know this in a general way. Infill percentage, wall count, and print orientation all affect strength, and anyone who has snapped a bracket along a layer line has learned about anisotropy firsthand. What Novineer is trying to do is put numbers on the effect and show that it is large enough to undermine the way FDM parts are usually analyzed and specified.
What Was Tested
Novineer says it worked with Stratasys on three materials: Nylon 12CF, Antero 800NA, and ULTEM 9085. The shear coupons were instrumented with rosette strain gauges. A rosette groups several strain elements at different angles so that strain can be resolved in more than one direction at a single point. That matters for anisotropic materials, where stiffness along the raster can differ sharply from stiffness across it or at an angle to it.
The column reports three headline results.
1. A 6x Swing in Bending From Toolpath Alone
The first is the beam comparison. With material, printer, geometry, and load all held constant, Novineer reports that deflection in bending differed by a factor of six depending on how the interior was toolpathed. For anyone who sizes a printed part by running the CAD geometry through a simulation with one set of material properties, a sixfold spread is the problem in a single number. The simulation sees one part. The printer can produce very different ones.
2. The Shear Modulus Trap
The second result is a measurement trap. Novineer reports that Nylon 12CF coupons printed with a ±45° raster were 3.5 times stiffer in shear than coupons printed with straight rasters. In printed carbon-filled nylon, the raster angle shapes the very property being measured. Tamijani writes that treating the ±45° figure as a general shear modulus would introduce a 250% error.
The arithmetic holds up. A value 3.5 times the straight-raster figure is 250% higher than it. If a test number taken from one toolpath gets entered into a materials database as if it were a property of the material, every simulation built on that number inherits the error.
3. One Extra Contour, 60% More Strength
The third result gives the article its headline. Novineer reports that adding one extra contour, a single additional outline pass around the edge, changed the toolpaths by less than 6% but raised strength by 60% in coupons tested in their weakest orientation. A small change to the slicing produced a large change in performance.
Tamijani uses this result to criticize standard test methods. He notes that ASTM D638, the common tensile test standard, was not written for FDM. A test coupon is itself a printed object with its own perimeters and infill, so its result depends on how the coupon was sliced. A tensile value with no record of the toolpath behind it is, on this reading, incomplete.
Where Novineer Is Taking This
The column describes a program that will grow past 1,500 tests across three materials and three printers. The data is meant to feed NoviPath, Novineer's toolpath-aware simulation.
That connects to the company's broader product. On its website, Novineer describes "Design-to-Toolpath Automation" that combines geometry layout with toolpath generation. It also lists stiffness- and strength-based optimization using Tsai-Wu failure criteria and anisotropic material properties. Tsai-Wu is a standard failure criterion from composites engineering, used to predict failure when a material's strength depends on direction. Applying it to FDM parts is consistent with the column's main point: printed parts behave more like laminates than like blocks of uniform plastic.
The commercial interest should be stated plainly. A company that sells toolpath-aware design and simulation benefits when readers conclude that toolpaths decide part performance. That doesn't make the data wrong, and the direction of the findings matches what practitioners see every day. But the column is the only source for these numbers. The coupon counts, the scatter, and the exact toolpath definitions behind the 6x, 3.5x, and 60% figures are not in the summary. Until a fuller dataset is published, these are well-framed claims, not results anyone has reproduced.
What It Means for Makers
Most hobbyists won't print ULTEM 9085 or run strain-gauge rosettes in the garage. The lessons still apply at the hobby bench.
Treat slicer settings as part of the design. If a part carries load, its wall count, raster angles, and contour strategy matter as much as its fillets and wall thickness. Record them with the model. A shared STL without slicer settings is an incomplete specification.
Be skeptical of single-number datasheets. A tensile or shear figure for a printed material means little unless you know how the coupon was printed. If the toolpath isn't stated, assume the number reflects one specific toolpath and may not match yours.
Try the cheap fixes first. Novineer's report that one extra perimeter added 60% strength in the weakest orientation suggests that before switching to a more expensive filament or making a part thicker, it is worth testing another wall loop. It costs a little print time and nothing else.
Watch the published data. If the 1,500-test program produces open, toolpath-tagged properties, it could give engineers working with printed parts something they lack today: material data that matches what the printer actually builds.
