Zero-power photonic language model–code
Details
- External ID
- 46089764
- Source
- HN
- Company
- —
- Product
- Zero-power photonic language model–code
- Website domain
- zenodo.org
- Launched
- Nov. 29, 2025
- Cohort
- —
- Upvotes
- 18
- Upvotes percentile
- 0.6703056768558951
- Tags
- —
- Fetched at
- Sept. 7, 2026, 9:25 p.m.
- Updated at
- Sept. 7, 2026, 9:25 p.m.
Description
The model uses a 1024-dimensional complex Hilbert space with 32 layers of programmable Mach–Zehnder meshes (Reck architecture) and derives token probabilities directly via the Born rule.Despite using only unitary operations and no attention mechanism, a 1024×32 model achieves coherent TinyStories generation after < 1.8 hours of training on a single consumer GPU.This is Part 1 - the next step is physical implementation with $50 of optics from AliExpress.
Enrichment
- Theme
- scientific computing and deep tech tools
- Vertical
- Horizontal
- Function
- Model & infra
- Audience
- Developer
- AI stance
- AI-native
- Project type
- Hobby / open-source project
- Normalized one-liner
- photonic language model without power requirements
- Manually corrected
- False
Could you build this?
No This project deals with novel research in neuromorphic photonics, physical optics simulations (Mach-Zehnder interferometer meshes, Reck architecture), and physical hardware implementations.
What it would actually take: Building this requires a physics simulation pipeline (using NumPy/PyTorch/JAX) modeling linear optics unitary transformations, complex Hilbert space operations, and Born-rule state collapse, alongside physical laser phase mask hardware prototyping. It requires a PhD-level background in optical computing, quantum/wave mechanics, and non-standard neural network architectures.
Discussion
7 comments analyzed.
Concerns raised: Unclear if detector can meaningfully read output from optical chain, Information loss between electronic and optical domain translation, Lacks real hardware implementation details and reliability validation, Unclear path to scaling beyond GPT-2 level complexity, How to implement non-linear activation functions without amplification
Competitors
Other products that read as similar to this one — 88 launches clear the similarity bar, closest 8 shown.
Attention rank: #29 of 89 (itself plus its competitors, highest first — normalized so YC and Product Hunt are compared fairly).
Launched 28 days after the earliest competitor.
- zkGolf · hn · 2026-07-02 · 69 upvotes · similarity 0.47
- Qumulator · hn · 2026-04-27 · 14 upvotes · similarity 0.45
- Digital to Wave: Wave computing-Memory · ph · 2026-09-14 · 1 upvotes · similarity 0.44
- Coda — Natural language quantum computing · yc · 2026-01-21 · 10 upvotes · similarity 0.42
- genpark-quantum-statevector-simulator-skill · github · 2026-09-09 · 8 upvotes · similarity 0.42
- genpark-quantum-statevector-simulator-skill · github · 2026-09-09 · 8 upvotes · similarity 0.42
- C++ Quantum Simulator written from scratch · hn · 2025-11-08 · 7 upvotes · similarity 0.41
- Open-source distributed quantum compute network · hn · 2026-04-02 · 11 upvotes · similarity 0.41
Other launches for this product
- No other launches for this product.
Same idea, different domain
Nobody's really built a model & infra tool for Fintech yet.