Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Antigravity 2.0 Tops the OpenSCAD Architectural 3D LLM Benchmark

modelrift.com

Post reply on HN
Print view View original

151–160 of 171 posts

Re: Antigravity 2.0 Tops the OpenSCAD Architectural 3D LLM Benchmark

#151

Post by simplyluke »

0x696C6961 wrote: post #52
simplyluke wrote: post #39

Earlier quoted context omitted.

Yeah, CAD has been my personal example of "oh the barrier to entry for this skill was high enough that I didn't do it and now I can be passably bad at it enough to get some simple things done" I've had similar experiences with making simple functional parts off a 3d printer with OpenSCAD + LLMs. I'm very aware that the models are worse at it than say, generating react code, and I'm also the antithesis of a skilled pi…

Learning to make simple parts in onshape is pretty darn easy (and fun).

I mean, like any other skill that has pretty much been my experience (though I tried fusion + openscad), but there is something about being able to ask a computer all the dumb noob questions that makes that first phase easier.

Re: Antigravity 2.0 Tops the OpenSCAD Architectural 3D LLM Benchmark

#152

Post by bitpush »

mchusma wrote: post #107
mellosouls wrote: post #2

Antigravity may well Top the whatever benchmark but: My Antigravity (forced) replacement for Gemini CLI requires me to log on via browser every time I use it, and my Antigravity IDE won't update at all, so: If it's ok I'd prefer they just work on reaching a baseline acceptable rollout before worrying about being Top in anything. Ps actual title: OpenSCAD LLM Benchmark: Building the Pantheon

I just left the google I/O feeling less confident about google's execution here. - Gemini 3.5 flash is strange. Old cutoff, basically better than 3.1 pro at soem things worse at others, sometimes cheaper, sometimes more expensive than 3.1 pro. - Antigravity had seemed abandoned, and people speculated them cutting it off, and they kind of did migrating everyone to a new antigravity - Google "shipped the org chart" and…

The cut off doesn't matter since all of them use tools.

Re: Antigravity 2.0 Tops the OpenSCAD Architectural 3D LLM Benchmark

#153

Post by mvkel »

I'm working on a parenting tech device and the enclosure for it is completely AI generated. I hadn't a clue where to even start with 3D modeling, and an LLM taught me that it's code like anything else.

Weirdly, Opus 4.5 one-shotted it perfectly, but this was right before the nerfing controversy, and it's been very difficult to make even minor tweaks to the enclosure ever since.

It's like Opus went from an expert shape rotator to not having any idea what it's working on.

Re: Antigravity 2.0 Tops the OpenSCAD Architectural 3D LLM Benchmark

#154

Post by ljlolel »

mvkel wrote: post #153

I'm working on a parenting tech device and the enclosure for it is completely AI generated. I hadn't a clue where to even start with 3D modeling, and an LLM taught me that it's code like anything else. Weirdly, Opus 4.5 one-shotted it perfectly, but this was right before the nerfing controversy, and it's been very difficult to make even minor tweaks to the enclosure ever since. It's like Opus went from an expert shap…

Ditto for my enclosure for https://quill.lorehex.co/feather

4.7 has been fine for making edits though

Re: Antigravity 2.0 Tops the OpenSCAD Architectural 3D LLM Benchmark

#157

Post by illiac786 »

pabs3 wrote: post #148
illiac786 wrote: post #125

Earlier quoted context omitted.

“The reproducibility is cryptographically verifiable with hashes“ would be the full sentence, but it’s a mouthful.

Build reproducibility checks usually use bitwise comparison, not hash comparison. The Reproducible Builds project also wrote diffoscope, which goes quite far with helping identify where differences occur and how to fix them. https://reproducible-builds.org/ https://diffoscope.org/ https://try.diffoscope.org/

Let’s say, for the positive case, hash comparison is significantly faster.

Re: Antigravity 2.0 Tops the OpenSCAD Architectural 3D LLM Benchmark

#158

Post by pabs3 »

illiac786 wrote: post #157
pabs3 wrote: post #148

Earlier quoted context omitted.

Build reproducibility checks usually use bitwise comparison, not hash comparison. The Reproducible Builds project also wrote diffoscope, which goes quite far with helping identify where differences occur and how to fix them. https://reproducible-builds.org/ https://diffoscope.org/ https://try.diffoscope.org/

Let’s say, for the positive case, hash comparison is significantly faster.

I feel like that is quite unlikely. Both the hash and bitwise comparisons read both files in both cases. In the not-equal case the hash reads the entirety of both files, so its slower than a start-to-end bitwise comparison, which exits at the first not-equal bit. In the equal case, both read the entirety of both files. Various other bitwise strategies can be faster than start-to-end, rdfind for example checks the start of the file first, then the end, then the rest of the file.

Re: Antigravity 2.0 Tops the OpenSCAD Architectural 3D LLM Benchmark

#159

Post by embedding-shape »

nancyminusone wrote: post #129
jlhawn wrote: post #105

> Antigravity was the only autonomous agent that implemented the Pantheon’s signature interior ceiling pattern: repeated square coffers visible through the oculus. That is seriously really impressive. I looked at the 3D model and didn't even thing to LOOK INSIDE the building before reading this. Here's [1] the 3D model with `show_cutaway` enabled. [1] https://modelrift.com/models/pantheon-benchmark-antigravity-...

I can't decide whether it's good or bad that it has included outside information clearly not present in the prompts to make the model. Clearly its the right thing to do if you want "the Pantheon" but I don't think any draftsman or engineer would find this acceptable work.

Exactly my experience with pretty much any Google LLM, they go above and beyond what you ask, even sometimes when you plead for them to not add comments they end doing so anyways, sometimes with "the user asked for X but it would be better if..." visible in the thinking/reasoning even, really strange behavior and exactly what I do not want from my agents.
Post reply on HN