One Engineer, One AI: Prompts That Turn a Solo Maker into a Full 3D Printing and CAD Department
A decade ago, taking a functional part from sketch to printed object meant a chain of specialists: a concept designer, a CAD modeler, a simulation engineer, a manufacturing planner, and a print operator. Today a single maker with a laptop, a slicer, and a well-prompted LLM can close most of that chain in one evening. The catch is that generic requests like "design me a bracket" produce generic answers. What works is structured, role-driven prompting that forces the model to reason like an engineer.
This collection walks the full cycle: concept generation, parametric CAD, structural sanity checks, generative/topology design, slicer parameters, and post-mortem diagnosis of failed prints. Every prompt below includes the task, the prompt itself, an example of what a good answer looks like, and the use case it fits — quick prototype, load-bearing functional part, or fine-detail miniature. Treat these as building blocks: the LLM handles reasoning and math, while your slicer and printer handle physics.
A note on standards and trust: print quality depends on real machine limits, so always verify numbers against your slicer's documentation (PrusaSlicer, OrcaSlicer, Cura) and material datasheets rather than trusting a model blindly. For structural claims, FDM parts are anisotropic — layer adhesion is typically far weaker than in-plane strength, a well-documented effect in fused-filament research, so treat any AI strength estimate as a first-pass heuristic, not a certification.
1. Concept generation from a functional brief
Task: turn a plain-language need into 2–3 distinct design directions.
Prompt:
You are a senior mechanical designer. My need: a wall-mounted holder for a cordless drill that must survive being grabbed daily in a garage.
Constraints: print on a 220x220x250 mm FDM printer, PETG, no supports if possible, single part.
Give me 3 concept directions. For each: geometry sketch in words, mounting method, failure mode to watch, and why it fits or doesn't.
Expected result: three named concepts (e.g., "cantilever cradle", "two-arm yoke", "saddle with retaining lip"), each with a described mounting strategy (countersunk screws, keyhole slots), and an honest failure-mode note (layer delamination at the cantilever root).
Fits: rapid prototyping and any first-pass functional part.
2. Parametric CAD as code (OpenSCAD / CadQuery)
Task: get a real, editable model instead of a picture.
Prompt:
Write a parametric OpenSCAD model of a rectangular enclosure: outer 80x60x30 mm, wall 2.4 mm, 3 mm fillet on vertical edges, lid with 4 M3 screw bosses.
Expose parameters as variables at the top. Add a 0.2 mm tolerance gap between lid and body.
Expected result: compilable .scad code with variables like wall = 2.4;, a difference() for the cavity, and minkowski() or offset() for fillets. You can then tweak numbers and re-render.
Fits: functional parts where dimensions will change.
3. Back-of-envelope strength check
Task: sanity-check a load before you waste filament.
Prompt:
A PETG cantilever arm is 60 mm long, 10 mm wide, 5 mm thick, fixed at one end, with a 30 N downward tip load.
Treat it as a rectangular beam. Compute max bending stress. Compare to typical PETG yield (~50 MPa).
State assumptions and tell me if layer orientation matters here.
Expected result: a worked calculation (σ = M·c/I), a stress figure, a margin-of-safety comment, and a warning to orient layers so bending stress is in-plane, not across layer lines.
Fits: functional parts; never skip this for load-bearing geometry.
4. Topology / generative design guidance
Task: use AI to reason about material removal where it matters.
Prompt:
I want to reduce mass of a 90 g bracket by ~30% using generative design.
Explain, step by step, how to set up a topology optimization in a free tool (e.g., nTop or Fusion's generative study):
load cases, constraints, preserve regions, and how to interpret the organic result for FDM printing.
Expected result: a workflow with load/constraint definition, a target mass, and practical advice on smoothing organic surfaces and orienting them so they print without brutal supports.
Fits: weight-critical functional parts, drones, robotics.
5. Choosing material and orientation
Task: match resin/filament and print orientation to the job.
Prompt:
I'm printing a phone stand used outdoors in summer. Compare PLA, PETG, ASA, and nylon for:
UV resistance, heat deflection, layer adhesion, and print difficulty.
Then recommend an orientation that puts the highest tensile stress in-plane.
Expected result: a comparison table, an ASA recommendation for outdoor heat/UV, and an orientation sketch reasoning about layer lines.
Fits: outdoor and load-bearing functional parts.
6. Slicer parameter translation
Task: convert intent into concrete slicer settings.
Prompt:
For a 0.4 mm nozzle on PETG, I want strong layer adhesion and minimal stringing.
Give starting values for: nozzle temp, bed temp, layer height, perimeters, infill type and density, retraction distance and speed, and fan speed.
Explain which settings trade strength against surface finish.
Expected result: a settings table with values you can type into PrusaSlicer or OrcaSlicer, plus reasoning about why gyroid infill and higher perimeters improve strength.
Fits: every part; tune once, reuse often.
7. Support reduction strategy
Task: redesign geometry to print support-free.
Prompt:
This overhang is 55 degrees and I want to avoid supports. Suggest 3 geometry changes (chamfers, teardrop holes, split parts) and estimate how each affects print time.
Expected result: concrete geometry edits like 45° chamfers and teardrop-shaped horizontal holes, with trade-off notes.
Fits: miniatures and clean-surface functional parts.
8. Print-defect diagnosis from a photo description
Task: turn a failed print into a fix.
Prompt:
My print shows stringing between towers and the top surface is rough with gaps.
Material: PETG, 0.4 mm nozzle. List the 5 most likely causes ranked by probability, and the single setting to change for each.
Expected result: ranked causes (wet filament, too-high temp, insufficient retraction, too-fast travel, cooling), each with one concrete fix.
Fits: all categories; fastest debugging loop.
9. Tolerance and fit for assemblies
Task: make parts that actually assemble.
Prompt:
I need a press-fit and a sliding fit for 6 mm shafts on FDM.
Recommend clearance values for each, explain how printer calibration shifts them, and give a test-print coupon design to dial it in.
Expected result: clearance ranges (roughly 0.1–0.2 mm press, 0.3–0.4 mm sliding), plus a coupon STL concept.
Fits: functional assemblies and mechanisms.
10. Miniature detail optimization
Task: keep fine detail on small, high-resolution prints.
Prompt:
For a 32 mm miniature on a resin printer, advise on: layer height, exposure, anti-aliasing, and how to orient for minimal support marks on the face.
Expected result: a resin-focused settings list and orientation advice favoring the back of the model.
Fits: miniatures and display models.
11. Cost and time estimation
Task: quote a job before committing.
Prompt:
Estimate material cost and print time for a 40 g PETG part at 0.2 mm layer height.
Material is $25/kg. Give a formula I can reuse.
Expected result: a simple formula (mass × price/kg) and a time estimate that you verify in your slicer.
Fits: quoting work, batch planning.
12. Design-for-manufacturing review
Task: get a second pair of eyes.
Prompt:
Review this part description for FDM DFM issues: sharp internal corners, unsupported bridges over 20 mm, thin 0.8 mm walls, and a horizontal hole.
List each problem and the fix.
Expected result: a checklist mapping each issue to a remedy (fillets, chamfers, wall thickening, teardrop holes).
Fits: every functional part before slicing.
13. Documentation and revision notes
Task: keep a printable history.
Prompt:
Write a short changelog entry for v1.3 of my bracket: added 2 mm fillet at the root, moved screw holes 3 mm outward, switched to PETG.
Format for a README.
Expected result: clean, versioned notes you can paste into a repo or project folder.
Fits: any repeatable, iterated project.
14. Failure post-mortem and next iteration
Task: close the loop after a break.
Prompt:
My PETG part cracked exactly at the layer line under the cantilever root after two weeks of use.
Propose 3 redesign changes and 2 print-setting changes to prevent recurrence.
Expected result: root-cause reasoning (stress concentration + weak interlayer bond) and fixes like a fillet, reorientation, and higher nozzle temp.
Fits: load-bearing functional parts.
15. Turning a photo into a modeling plan
Task: reverse-engineer a reference.
Prompt:
I have a photo of a broken plastic clip. Describe how to measure and reconstruct it in CAD: which dimensions to capture first, how to infer radii, and how to add a reinforcement rib.
Expected result: a measuring checklist and a modeling order (base profile → features → rib → fillet).
Fits: repair parts and one-off replacements.
16. Batch and print-farm planning
Task: scale from one part to many.
Prompt:
I need 20 identical PETG parts. Advise on gcode batching, plate packing, and whether sequential printing is worth it.
Give a simple decision rule.
Expected result: a packing strategy and a rule of thumb comparing one big plate vs. sequential printing.
Fits: small production runs.
The thread running through all sixteen prompts is simple: the AI supplies structured engineering reasoning — calculations, checklists, trade-offs — while your slicer and printer supply the ground truth. Used this way, a solo maker genuinely operates like a compact design bureau. Pick two or three prompts that match your current bottleneck, run them tonight, and iterate. The fastest way to learn is to print, break, and let a well-framed question tell you why.
Comments