Mazer23
Joined 16 years ago
Karma
336
Submissions
47
Member since
Apr 8
Recent submissions
Comment · 1 day ago
Comment · 22 days ago
Shameless plug: I’m workin on http://partforge.ai right now it’s mostly aimed at 3d printing but it can create laser cut sheet parts with dxf file exports and will support cnc routed parts with drill holes and socket diagrams built in in the near future. Board cut lists as a export is a little further out, but you can ask it for measurements of anything you build and it would probably be a pretty useful tool.
Comment · 26 days ago
I feel like this is missing the for the trees a bit. I don’t think anyone who is saying “we don’t understand how llms work” means what this author claims. The underlying math of transformers isn’t the part we’re talking about.
What people actually mean is we don’t and often can’t understand an entire reasoning chain of a large llm and often can’t trace that back to the specific training data that caused that reasoning.
Anthropic itself had build all kinds of tools to try to understand llm internal “thoughts” https://www.anthropic.com/research/natural-language-autoenco...
I don’t think it’s unreasonable to say “no one understands how llms work.” What they mean is we don’t and can’t follow every one of its internal chains of thought, especially at the pace new models and techniques are being developed.
Comment · 27 days ago
I spent quite a bit of time using openscad which Claude code to try to build a capstan (rope driven) joint for robotics. It worked well but had some issues with agentic stuff and speed. In the end I ended up taking a detour and building my own agentic cad modeling harness based on some of the same open source libraries that openscad uses. Now I’m trying to turn that into its own standalone product. You can check it out at http://partforge.ai
Comment · 1 month ago
We tested this on our agentic CAD coding harness. This seems like a decent improvement vs 5.6. It's lower costs seems to be offset by more token use so those cancel themselves out, but the actual results for spacial reasoning and coding are an improvement.
Comment · 1 month ago
We tested this on our agentic CAD modeling harness. Opus 5.5 seems to be a pretty dramatic improvement in spacial reasoning and CAD coding. Our go-to is still Gemini Flash 3.8 for speed and cost reasons, but Opus 5.5 is a big update.
https://aireel.tv/s/hazLIx8f
The vector graphics and animations are pretty llm quality. I wonder if you could leverage some of the new image frameworks instead of an llm animating vectors to make the animation in the middle a lot more compelling.
For comparison, here is our human-made product video. https://www.partforge.ai/video
Given that our product is about 3d printing real parts having real video footage makes it a lot more compelling. Also the real VoiceOver feels much more authentic. I'm already getting a serious aversion to ai style voiceovers. Maybe there are some ways to make yours feel a bit more out of the box so that it introduces new ideas and new pitches that might stretch what a product creator thinks of?
Cool idea and nice execution. Makes a lot of setup you would normally do with llm's and skills pretty seamless and packaged nicely.