Bimorph

20/07/2026

Would an LLM know what to do with your Grasshopper definitions?

A designer can open a cluster, hover the inputs and trace the wires until it makes sense. An LLM needs a contract: inputs, types, defaults, outputs. So we gave it one.

Wrap a cluster in a typed boundary and it compiles itself into an MCP JSON schema, the format every major LLM API already speaks. The logic inside never changes. When the AI calls it, the runtime solves an isolated copy and hands back the results. The LLM sees a schema. Grasshopper sees a cluster.

The upshot: existing definitions become callable AI tools without rewriting a single one.

The clip shows the runtime end to end, with no LLM anywhere in it. It's the foundation our AI experiments sit on.

Full article in the first comment. Has anyone else tried wiring an LLM into Grasshopper? How far did you get?

13/07/2026

New on the Bimorph blog: an AI-orchestrated Grasshopper workflow running entirely on a laptop, using a free, open-source model. No cloud, no API bill.

The method is a stage pipeline. A plain-English request gets broken into five small steps (rewrite, tool discovery, selection, graph build, ex*****on), each checked before the next runs. Structured this way, even a tiny model produces working, verifiable geometry.

If your firm has held back on AI tools over data concerns, this one is worth a read: nothing leaves the machine.

https://www.bimorph.com/insights/local-intelligence-stage-pipeline

03/07/2026

We're excited to announce the launch of Rhino.Inside.Civil3d!

⬇️ The newest version is available to download - see the comments for the download link.

Rhino.Inside.Civil3D is built on our work with the open-source Rhino.Inside.AutoCAD developed in collaboration with McNeel Europe. The two share a single installation, and the application automatically detects whether you're running inside Civil3D.

Rhino.Inside.Civil3D is a community-driven project, and all feedback is welcome. If there's a feature you'd like to see on the roadmap, we'd love to hear your suggestions.

🔗 Join the conversation on the McNeel forum - link is in the comments.

29/06/2026

Do programming standards matter in the era of AI agents?

With AI generating an increasing percentage of modern codebases, you might assume strict formatting rules matter less. In reality, they matter more—but not just for the usual reasons.

The traditional argument for strict standards is human readability. That remains true, but AI coding assistants introduce a new variable: Context Contagion.

An agent learns your conventions from the files already in your repository. Every file it reads becomes a template. If your codebase is inconsistent, it teaches the agent bad habits. Clean architecture breeds clean generation.

Beyond that is the issue of authorship. You can hand an agent a list of features and get something that compiles. But compiling isn't understanding. The day you stop genuinely reading your own system is the day you downgrade from an author to a coordinator.

Read our CTO Tom Norris’s full breakdown on why we enforce strict coding standards at Bimorph 👇

Why strict coding standards matter more than ever in the age of coding agents — not because the agent needs them, but because they keep you the author.