Summary
Topher tests a voice memo system and pitches an AI-driven 3D printing concept, requesting overnight research.
Discussion
- Topher proposes making 3D printing accessible through AI-assisted design so non-experts can generate print files easily.
- He suggests using iPhone photos alongside a 3D-printed measurement board to capture accurate dimensions for AI processing.
- The goal is to automate printer settings and file generation so users can produce expert-quality prints without technical knowledge.
- Topher outlines a multi-model AI workflow where one model drafts a plan and another challenges and refines it.
- He requests Aria to research the technical feasibility and outline a pilot or MVP approach.
Action items
- Research AI-assisted 3D printing workflow and outline a pilot/MVP approach — Aria
Speakers
Transcript
Hey Aria, this is Chris, and I got this crazy idea. Well, maybe not so crazy, but I'm testing our voice memo system right now, but also I'm going to ask you to reference us back with some research that you're doing for me overnight.
The thought is, I have this 3D printer. It's a Bamboo Labs P2S, I think, and I don't know a damn thing about printing anything. And I've used a little bit of advice from like ChatGPT on how to try and make a print file.
And when I did a very basic print file, it was okay, but what I'm thinking about, and this can be a concept for a future project down the road, and maybe we scale and productize and monetize. But how do we make 3D printing something so easy that you can vibe code it?
Like right now we're using Vegas to vibe code websites, and that's great and all, but how do we make a process that any random person can vibe code 3D drawings of items? And how do we make that process easy for them to use with tools that they have?
Right? So the first thing I'm thinking about is if they're trying to create something for a real world problem, they need to have the ability to capture images of whatever that is. And we don't want them to buy expensive equipment.
So how do we then make something or give them a process on how to use something like a simple iPhone to take pictures, but maybe we 3D print some type of measurement board to put next to items so that when the picture is taken, whoever's looking at that image can get a good view of what that image is and how big the item is based off of its size compared to the board, that kind of thing.
Maybe those image boards are something we would just 3D print too. Just a thought. And then once we, so the thought there being, like I tried to make a dog collar holder for this chip that I have that's similar to an AirTag, but it's not an AirTag.
And it's got very specific measurements.
So if I was to try and vibe code that again, what would I need to do to give my agent the information it would need to build me a 3D print model that I could then print? Just like in 3D print world, this agent's going to have to be up to date with the most latest, greatest 3D modeling tools, techniques.
It's got to know what printer you're using. So that way it can have all the files and settings automatically, like literally do everything for the user so that a fifth grader could, could do expert quality prints too. So, um, do some deep research on this and think through how we'll execute some type of pilot or MVP.
And, uh, one idea that, that I saw on YouTube is that we should do a planning phase and then we should challenge that planning phase with a separate model. So maybe if you do the planning phase on Opus and then we run that planning phase through a separate model on like Sonnet, or if we can somehow get, um, access to GPT, I have GPT 5 .5.
Um, but how do I get you access to that? So that way it can challenge that plan and add to it and then do a couple iterations of that plan before we execute on it. So let's think through that too.
Uh, that should be enough for the night.