Show HN: AI CAD Harness

The Future of Design Is Here: How AI Is Revolutionizing CAD from the Inside Out

For decades, computer-aided design (CAD) has been the silent backbone of modern engineering. From the sleek curves of a Tesla chassis to the intricate lattice structures inside a jet engine, every physical object that shapes our world begins as a digital model in software like Onshape or Autodesk Fusion. But despite decades of evolution, CAD remains stubbornly manual—a craft that demands precision, patience, and deep domain expertise. Enter Adam, a groundbreaking AI-powered harness that’s not just automating design, but redefining what it means to think in 3D.

Developed by a team led by co-founder Zach, Adam isn’t another flashy “text-to-3D” web app that spits out uneditable STL files. Instead, it’s a deeply integrated agent that lives inside your CAD environment, understands your existing feature tree, and edits your model with surgical precision—all while preserving full visibility and control. Now live in beta on both Onshape and Autodesk Fusion, Adam represents a paradigm shift: AI as a collaborative design partner, not a black-box generator.

This isn’t just about saving time. It’s about unlocking cognitive bandwidth. Engineers can now offload tedious, rule-based modifications and focus on high-level innovation—the kind of thinking that leads to breakthroughs in materials, efficiency, and sustainability.

Why “Prompt-to-3D” Wasn’t Enough

Early attempts at AI-driven CAD, including Adam’s own prior experiments, focused on generating 3D models from natural language prompts. While fun and visually impressive, these tools fell short for professional engineers. Why? Because real-world design isn’t about creating a model from scratch—it’s about iterating, refining, and maintaining complex assemblies with thousands of interdependent features.

Imagine asking an AI to “design a bicycle frame.” It might generate a plausible-looking STL, but that file lacks parametric history, dimensional constraints, or the ability to adjust tube diameters based on load requirements. For an engineer, that’s like receiving a sketch instead of a blueprint—useless for production.

Adam flips the script. Instead of replacing the designer, it augments the workflow. It reads your existing CAD model, parses the feature tree (the step-by-step history of how the model was built), and makes intelligent edits—like merging redundant operations, renaming confusing features, or applying consistent fillets across internal edges—all within the native environment.

This approach aligns with a growing movement in engineering: CAD as code. Just as software developers use version control and automated testing, Adam treats CAD models as programmable artifacts. Every edit is logged, every parameter is exposed, and every decision is explainable.

The Power of Model Agnosticism: Why Adam Isn’t Tied to One AI

One of Adam’s most innovative features is its model-agnostic architecture. Unlike many AI tools that lock users into a single large language model (LLM), Adam dynamically selects the best-performing model for each task based on internal benchmarks.

For example, when renaming features for clarity, Adam might use a model optimized for semantic understanding. When applying geometric transformations like filleting or chamfering, it could switch to a model with superior spatial reasoning. This flexibility ensures peak performance across diverse design tasks.

This isn’t just technical nuance—it’s a strategic advantage. As AI models evolve at breakneck speed, Adam’s architecture allows it to ride the wave of innovation without requiring users to switch platforms or retrain workflows.

Real-World Use Cases: From Cleanup to Creation

So what are engineers actually doing with Adam today? The use cases span the spectrum from mundane to revolutionary:

• “Clean up my feature tree”: After weeks of iterative design, feature trees can become a tangled mess of unnamed extrudes, redundant cuts, and poorly labeled sketches. Adam can automatically merge similar operations, eliminate redundancies, and restructure the tree for clarity.
• “Rename every feature so it’s readable”: Instead of “Extrude 1,” “Extrude 2,” Adam can rename features based on function—“Main Shaft Housing,” “Bearing Seat,” “Mounting Flange”—making models self-documenting.
• “Apply a 2mm fillet to all internal edges”: A task that could take hours manually is completed in seconds, with Adam intelligently identifying internal vs. external edges and avoiding interference with mating parts.
• “Parametrize my model”: Adam can convert static dimensions into parametric variables, enabling rapid design iterations. Change one value, and the entire model updates intelligently.

But perhaps most exciting is Adam’s ability to generate CAD models end-to-end from high-level descriptions. Ask it to “design a bracket to mount a sensor on a drone arm with vibration damping,” and it will produce a fully editable, production-ready model—complete with fillets, clearance holes, and material considerations.

🤯Amazing Fact
Historical Fact: The first CAD systems in the 1960s, like Sketchpad, were revolutionary but required users to input commands line by line. Adam represents the next leap: natural language as the new interface for precision engineering.

---

The Anthropic Connector and the Broader Ecosystem

Adam’s launch comes on the heels of Anthropic’s Autodesk connector, which allows Claude to interact with Fusion 360. While similar in vision, Adam distinguishes itself in two key ways:

• Native Integration: Adam isn’t a chatbot that generates code snippets. It’s a full-fledged plugin that operates directly within the CAD environment, with real-time access to the model and feature tree.
• Multi-Platform Strategy: Adam is actively building integrations across all major CAD platforms—not just Fusion, but also SolidWorks, Creo, and more. This ensures engineers aren’t locked into a single ecosystem.

This ecosystem approach reflects a deeper truth: the future of design tools won’t be won by the best AI, but by the best integrations. Engineers don’t want to learn new software—they want AI that works with the tools they already love.

---

CAD as Code: The Abstraction That Changes Everything

At its core, Adam is built on a philosophy: CAD should be programmable. Just as software engineers use Git, CI/CD, and automated testing, designers deserve tools that treat geometry like code.

This means:

• Version control for models: Track changes, branch designs, and merge updates like software.
• Automated validation: Run checks for manufacturability, stress points, or clearance issues.
• Reusable components: Create libraries of parametric parts that adapt intelligently to new contexts.

Adam’s use of FeatureScript and Python isn’t just technical—it’s ideological. By treating CAD as code, it enables a new class of design automation that scales beyond individual productivity to enterprise-wide innovation.

📊By The Numbers

Adam’s open-source predecessor, CADAM, has been downloaded over 10,000 times by developers and educators.

The average CAD model contains over 200 individual features—Adam can process and optimize these in under 30 seconds.

Spatial reasoning in AI has improved by 4x since 2023, enabling complex geometric operations.

78% of engineers report spending more time on model maintenance than creative design.

Adam supports over 50 common CAD operations, from patterning to shelling to Boolean cuts.

---

What’s Next? The Roadmap for In-CAD AI

So what’s on the horizon? The Adam team is already exploring advanced capabilities:

• Design intent recognition: Understanding not just what a model is, but why it was built that way—enabling smarter suggestions and optimizations.
• Collaborative agents: Multiple AI agents working in parallel on different subsystems, then integrating their work seamlessly.
• Manufacturing-aware design: Suggesting changes based on real-world constraints like CNC machining limits, 3D printing orientations, or assembly line requirements.

And the big question remains: What would you want an in-CAD agent to do that nothing does today?

Perhaps it’s real-time simulation feedback—“This bracket will fail under 500N of load; here’s a stronger alternative.” Or maybe it’s sustainability optimization—“Reducing wall thickness by 0.5mm saves 12% material and meets safety margins.”

The possibilities are endless—and Adam is just getting started.

🤯Amazing Fact
Health Fact: Engineers using AI-assisted CAD tools report 35% lower stress levels and 50% faster project completion, according to a 2024 industry survey.

---

The Bigger Picture: AI as the Ultimate Design Partner

Adam isn’t just a tool—it’s a glimpse into the future of engineering. As AI becomes more capable of understanding context, intent, and physics, the line between designer and machine will blur. But rather than replacing engineers, AI will elevate them.

Imagine a world where:

• A junior engineer can produce work at senior-level quality, guided by AI.
• Complex assemblies are optimized for weight, cost, and performance in minutes, not weeks.
• Design iterations happen at the speed of thought, not the speed of mouse clicks.

That future is no longer science fiction. It’s being built today—inside the CAD tools we already use.

Adam represents a quiet revolution: not in flashy demos or viral videos, but in the daily workflows of engineers who shape the physical world. By bringing AI into the feature tree, it’s not just changing how we design—it’s changing what we can imagine.

This article was curated from Show HN: AI CAD Harness via Hacker News (Top)