OpenGravity

OpenGravity (also referenced as 'OpenGraviton' on GitHub) is an open-source project created by a GCSE student in the UK. It enables local inference of large AI models (up to 500B parameters claimed) on consumer Mac hardware using mmap-based layer streaming and Top-K sparsity to reduce memory requirements. User comments reveal several distinct themes: 1. **USE CASES**: Users want it for personal coding, testing quick HTML/CSS/JS ideas in-browser without booting a full dev environment, and as a home server alternative to cloud API costs. 2. **HARDWARE & PERFORMANCE**: Commenters report interest in running models on Mac Mini M4 (16GB base), MacBook Pro M1 Max (64GB), and Mac Studio M3 Ultra (96GB). The author states ~36GB memory works for 72B models. However, skeptical users note that the GitHub benchmarks don't actually show results for larger-than-memory models — only TinyLlama results and size differences. One commenter calls it 'quite sloppy.' 3. **SKEPTICISM ABOUT CLAIMS**: Multiple users question the 140B-on-Mac-Mini claim. One notes that with 1.58-bit ternary quantization, 'you may think you're running a big model but really you're just running a mini version of it.' Another points out that sustained inference latency when streaming layers from disk mid-inference would likely degrade throughput sharply. 4. **IDENTITY & BRANDING**: The project may have been originally called 'Antigravity' or is a clone/reproduction of something called Antigravity. Commenters suggested rebranding to 'Zero Gravity.' The logo resembles an avocado or a '0.' 5. **TOS CONCERNS**: At least one commenter warns that using this approach may violate Gemini/Antigravity Terms of Service and could result in account revocation. 6. **LIMITATIONS**: Users report AI hallucinations removing important code parts, lack of continuous learning for project context, and session/context loss when switching users. 7. **COMMUNITY SENTIMENT**: Overwhelmingly positive and encouraging toward the young developer, with significant HackerNews engagement (multiple +64 upvote comments). However, technical skepticism is present alongside the encouragement. One unrelated Reddit comment about outdoor furniture appears to be noise from the data collection.

No aggregate ratings were found for this product during the last harvest.