So y'all, we did a thing: 🟦 Speckle extracts drawings from a #Revit model, each #PDF now ships with the properties of every element drawn on it, embedded in the file:
→ the element data as parquet attachments, a slice of the model's own tables
→ a hidden text layer, so that @grok and his friends that only reads the page still knows what's drawn where
→ visually identical to the PDF you'd print today
Drag one into @grok or his friends, ask "what's the fire rating of the exterior walls on this plan?", and it answers from the model's data, not from squinting at linework. No OCR, no bounding boxes, no confidence scores; the drawing stays a drawing but it stops being a dead end.
This is now ready for our build partners to start having fun with. Not a build partner? DM's open
A general comment follow up here is why tf incumbents and the general army of startups doing estimation/takeoffs/etc have never done this so far.
→ the element data as parquet attachments, a slice of the model's own tables
→ a hidden text layer, so that @grok and his friends that only reads the page still knows what's drawn where
→ visually identical to the PDF you'd print today
Drag one into @grok or his friends, ask "what's the fire rating of the exterior walls on this plan?", and it answers from the model's data, not from squinting at linework. No OCR, no bounding boxes, no confidence scores; the drawing stays a drawing but it stops being a dead end.
This is now ready for our build partners to start having fun with. Not a build partner? DM's open
A general comment follow up here is why tf incumbents and the general army of startups doing estimation/takeoffs/etc have never done this so far.
0:36
