Self-growing neural networks via a custom Rust-to-LLVM compiler
Details
- External ID
- 46411334
- Source
- HN
- Company
- —
- Product
- Self-growing neural networks via a custom Rust-to-LLVM compiler
- Website domain
- github.com
- Launched
- Dec. 28, 2025
- Cohort
- —
- Upvotes
- 8
- Upvotes percentile
- 0.4217557251908397
- Tags
- —
- Fetched at
- Sept. 7, 2026, 9:25 p.m.
- Updated at
- Sept. 7, 2026, 9:25 p.m.
Description
Hi HN,I built NOMA (Neural-Oriented Machine Architecture), a systems language where reverse-mode autodiff is a compiler pass (lowered to LLVM IR).My goal is to treat model parameters as explicit, growable memory buffers. Since NOMA compiles to standalone native binaries (no Python runtime), it allows using realloc on weights mid-training. This makes "self-growing" architectures a system primitive rather than a complex framework hack.I just pushed a reproducible benchmark (Self-Growing XOR) to validate the methodology: it compares NOMA against PyTorch and C++, specifically testing how preserving optimizer state (Adam moments) during growth affects convergence.I am looking for contributors! If you are into Rust, LLVM, or SSA, I’d love help on the harder parts (control-flow AD and memory safety).Repo: https://github.com/pierridotite/NOMA
Enrichment
- Theme
- lightweight and on-device AI runtimes
- Vertical
- Horizontal
- Function
- Dev tools
- Audience
- Developer
- AI stance
- Not AI
- Project type
- Hobby / open-source project
- Normalized one-liner
- self-growing neural networks compiler
- Manually corrected
- False
Could you build this?
No Building a custom compiler in Rust that translates dynamic neural network topologies directly to LLVM IR with custom reverse-mode automatic differentiation passes requires specialized compiler engineering and computer science expertise.
What it would actually take: Requires deep expertise in compiler design (Rust, LLVM C++/Rust bindings, custom IR passes) and numerical methods for automatic differentiation. The developer must implement custom memory allocators for dynamically growing parameter tensors, construct computation graphs lowering to LLVM instructions, and optimize machine code emission without runtime reliance on standard libraries like PyTorch.
Discussion
3 comments analyzed.
Feature requests: Support for complex control flow in reverse-mode AD, SSA Phi-node handling during automatic differentiation
Competitors
Other products that read as similar to this one — 165 launches clear the similarity bar, closest 8 shown.
Attention rank: #88 of 166 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).
Launched 55 days after the earliest competitor.
- Optimizing LiteLLM with Rust · hn · 2025-11-18 · 27 upvotes · similarity 0.45
- lsm-engine · github · 2026-09-12 · 16 upvotes · similarity 0.44
- Forkrun · hn · 2026-03-27 · 151 upvotes · similarity 0.44
- Linear RNN/Reservoir hybrid generative model, one C file (no deps.) · hn · 2026-04-09 · 7 upvotes · similarity 0.44
- Ontological Directed Synthesis Network · ph · 2026-09-22 · 1 upvotes · similarity 0.44
- Lamina · hn · 2025-11-20 · 5 upvotes · similarity 0.43
- I made an open-source Rust program for memory-efficient genomics · hn · 2025-11-13 · 17 upvotes · similarity 0.42
- High-Res Neural Cellular Automata · hn · 2026-06-17 · 208 upvotes · similarity 0.42
Other launches for this product
- No other launches for this product.
Same idea, different domain
Nobody's really built a dev tools tool for Sales yet.