Build a UAV CAD, CFD and DFM optimization app with coding models
I've tried smaller models and the best open models for a specialized UAV focused CAD + CFD + DFM optimization application I'm building.The best open weights models still had a lot of defects and very egregious mistakes to fix even after I provided a very well detailed Requirement/Design/ Technical Implementation plans.Opus and Gemini 3.8 (and now Argon) did a significantly better job. Task: I've tried smaller models and the best open models for a specialized UAV focused CAD + CFD + DFM optimization application I'm building.The best open weights models still had a lot of defects and very egregious mistakes to fix even after I provided a very well detailed Requirement/Design/ Technical Implementation plans.Opus and Gemini 3.8 (and now Argon) did a significantly better job. So I've baseline using these two models to build this application. I now use Opus for design/review and Flash 3.8 for the code gen/implementation work.I'm sure small models and even open big models are probably fine for many common small apps and such, but they aren't really able to perform the same across all use cases.; tools: Claude. Human role: The original post does not say how a person stayed involved.; limitations: The source does not specify explicit operational limits, costs, or integration constraints for the chosen models. It notes that smaller models may suffice for common small applications but fail to perform consistently across all use cases, particularly for specialized engineering domains. Human involvement details are not specified in the original post..
Tools
Published on Agent Work Atlas:
Overview
I've tried smaller models and the best open models for a specialized UAV focused CAD + CFD + DFM optimization application I'm building.The best open weights models still had a lot of defects and very egregious mistakes to fix even after I provided a very well detailed Requirement/Design/ Technical Implementation plans.Opus and Gemini 3.8 (and now Argon) did a significantly better job.
Workflow
I've tried smaller models and the best open models for a specialized UAV focused CAD + CFD + DFM optimization application I'm building.The best open weights models still had a lot of defects and very egregious mistakes to fix even after I provided a very well detailed Requirement/Design/ Technical Implementation plans.Opus and Gemini 3.8 (and now Argon) did a significantly better job. So I've baseline using these two models to build this application. I now use Opus for design/review and Flash 3.8 for the code gen/implementation work.I'm sure small models and even open big models are probably fine for many common small apps and such, but they aren't really able to perform the same across all use cases.
Reported result
The author reports that Opus and Gemini 3.8 performed significantly better than smaller or open-weight models for this complex engineering task. No specific quantitative metrics, such as error reduction rates or development time savings, were provided in the source. The outcome is described qualitatively as a successful baseline selection for the application's development.
Human role
The original post does not say how a person stayed involved.
Limitations
The source does not specify explicit operational limits, costs, or integration constraints for the chosen models. It notes that smaller models may suffice for common small applications but fail to perform consistently across all use cases, particularly for specialized engineering domains. Human involvement details are not specified in the original post.
Try it
Act as a senior software architect helping me build a specialized UAV CAD, CFD, and DFM optimization application. I have detailed Requirement, Design, and Technical Implementation plans ready. First, use [Claude Opus] to review these plans and generate a high-level architecture design. Then, use [Gemini Flash] to generate the initial code modules based on the approved design. Before proceeding to full implementation, perform a dry run on one specific module: generate the code, review it for potential CFD or DFM logic errors, and present the output to me. Do not commit any changes or spend API credits on the full project until I explicitly approve the sample output. If the sample contains egregious mistakes, ask me for clarification on the specific technical constraints before retrying.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.