AEC AI

 WeeklyDynamo AEC AI collects everything on this blog about putting AI to work in AEC practice — connecting models to Dynamo and Revit, and the project record of what actually happened when it was tried at production scale.

Tunnel automation series — nine parts, in order

A single tunnel project, from a finished Dynamo graph that was never wired to its workbook through to a Revit add-in that would not build a solid. Read from Part 1; each part also links to the next.

  1. Dynamo Built the Whole Tunnel and Never Read the Excel File
    A 617-node Dynamo graph modelled an 892 m tunnel with no link to the 39-sheet Excel workbook that defined it.
  2. Dynamo Code Blocks and the Cost of 137 Hardcoded Values
    137 hardcoded numbers across 32 Dynamo versions, and nine hops to trace one radius.
  3. Dimension Lines vs Real Edges: 9 Cases an AI Got Wrong
    An AI pipeline kept rebuilding dimension lines as real concrete edges.
  4. Human-in-the-Loop AI for BIM: Make It Ask, Not Guess (publishing shortly)
    Restructuring an AI workflow so it asks instead of guessing: six grades of evidence authority, a manifest, and a one-sentence instruction rule.
  5. Validation Gates That Always Pass: Add a Negative Control (publishing shortly)
    A gate that always passes proves nothing.
  6. Revit TessellatedShapeBuilder Returns a Mesh, Not a Solid (publishing shortly)
    The viewer said the geometry was closed; TessellatedShapeBuilder still returned a mesh.
  7. Prompt Engineering for BIM: 101 Instructions, One Drawing (publishing shortly)
    101 instructions to an AI over 11 days on one tunnel drawing.
  8. AI Clarifying Questions: A Template From Seven Real Ones (publishing shortly)
    The AI wrote seven questions it could not decide, each with two or three options and one evidence image.
  9. Parsing 149 GB of JSON Without Loading It Into Memory (publishing shortly)
    Indexing 149 GB of JSON at 261 MB/s without json.load.

Connecting AI to Dynamo and Revit

Practical setups: cloud versus local models, vision APIs, and image-to-geometry workflows.

Where AI fits in AEC practice

What AI changes in a design workflow, and where it stalls.

Infrastructure notes

Environment and runtime fixes.

See also: Automation · Generative Design · Newsletter

댓글

댓글 쓰기

이 블로그의 인기 게시물

Structural Analysis Workflow with Dynamo and Robot

이미지
  Research Log [2025–09–21]: Building and Troubleshooting an Automated Structural Analysis Workflow with Dynamo and Robot 1. Objective To establish and document a seamless, automated workflow connecting the parametric design tool Dynamo with Autodesk Robot Structural Analysis . This initiative covers the entire pipeline from model generation and analysis to result verification, with a systematic documentation of key errors encountered and their resolutions. 2. Summary of Work A comprehensive Dynamo script was developed utilizing the ‘Structural Analysis for Dynamo’ package. This script generates a structural model (members, nodes) and defines its core properties, including materials, sections, support conditions, and loads. The workflow successfully transfers this model to Robot for structural analysis and then brings the calculated member force and stress results back into Dynamo for color-based visualization. Throughout this process, several critical issues were identified...
Read more »

Dynamo with the Gemini Vision API test(Nano Banana)

이미지
  [September 13, 2025] Today’s Research Log: Verifying the Integration of Dynamo with the Gemini Vision API and its Limitations 1. Objective 🎯 The goal was to build an automated workflow within the Autodesk Dynamo environment. This workflow would use Python scripts to connect with Google’s multimodal AI model, aiming to receive image files from the AI that were either modified or newly generated. 2. Summary of Work 💡 I wrote a Python script to call the Google Gemini Vision API from Dynamo and attempted to integrate the API while resolving a series of network and model errors. Through systematic debugging and testing, I discovered that the initial goal of generating images with the ‘Nano Banana’ (Gemini 2.5 Flash) model is not currently supported by the API. As a result, while I successfully created a stable workflow that takes an image as input and returns a text analysis, I concluded that the original goal of image output could not be achieved. 3. Detailed Process & ...
Read more »