From: schneich
Yesterday I played around with the new GPT Astra. I told it to model a box with some lashes in MOI.
I didnt expect anything since it probably has never used MOI before :-) but it one-shotted the box and after
telling it to only use 45 overhangs it corrected everything in one go... I have to say ham a bit baffled.
next up: reverse engineering scanned objects.. :-)
cheers from Cologne
Torsten
Image Attachments:
Bildschirmfoto 2026-09-14 um 7.44.38 AM.png
From: Michael Gibson
Hi Torsten, it's interesting but the problem seems to be how to verify that it has done things correctly?
Like what if it is correct 98% of the time, can you tolerate that 2% of the time it's going to make junk?
- Michael
From: tom_tm (TOMF)
It would have to, in virtual reality, 3D print its creation and then check if the box slotted together easily or not.
Seeing AI is a random bullish!t generator, it's likely the AI would then hallucinate that it got it right when it clearly hadn't!
Only humans make good engineers. Not AI's.
From: Frenchy Pilou (PILOU)
What was the process ?
Ai gave you a file .3dm ? (or other formats)
Or a text describing the commands to use ?
or a video how to?
or...
