deep learning architectures and research tools
Recent window: last 5.2 months (2026-04-25 → 2026-09-29), compared with the prior 5.2 months.
These products offer novel machine learning model implementations, training optimizations, and low-level systems engineering utilities. They are primarily built for AI researchers, computer scientists, and systems developers experimenting with foundational algorithms. Unlike high-level AI SaaS applications, this cluster focuses on architectural research, model internals, and technical computing experiments.
Metrics
- Stage
- heating
- Recent count
- 51
- Prior count
- 9
- Total count
- 66
- Momentum
- 300.00
- Attention
- 0.49
- Crowding
- 0.18
- Concentration
- 0.82
- Opportunity
- 0.62
Opportunity components
- Attention
- 0.49
- Low crowding
- 0.82
- Momentum (normalized)
- 1.00
- Low concentration
- 0.18
Monthly trajectory
Source split
- github
- 39 (0.76)
- hn
- 11 (0.22)
- ph
- 1 (0.02)
- yc
- 0 (0.00)
Dominant source: github · Divergence: 0.76
Similar themes
- scientific computing and algorithmic tools (0.58)
- local inference engines and compact models (0.53)
- interactive 3D and creative web experiments (0.52)
- AI infrastructure and inference optimization (0.52)
- voice AI and speech tools (0.49)
- experimental AI and simulation tools (0.47)
- experimental utilities and interactive visualizers (0.46)
- computational geometry and spatial algorithms (0.46)
Members
| Name | Source | Upvotes ▲ | Launched |
|---|---|---|---|
| spivak-lean | GITHUB | 42 | 2026-09-26 |
| GLiFormer | GITHUB | 46 | 2026-09-11 |
| SHDL | HN | 48 | 2026-01-28 |
| STEPQuant | GITHUB | 49 | 2026-09-29 |
| RLT | GITHUB | 51 | 2026-09-13 |
| TurboGPT: train 22KiB transformer in 13s | HN | 56 | 2026-09-29 |
| magnet-finder | GITHUB | 63 | 2026-09-10 |
| Aha-Engine | GITHUB | 63 | 2026-09-10 |
| transformer-architecture | GITHUB | 67 | 2026-09-11 |
| A-Survey-on-Looped-Transformers | GITHUB | 101 | 2026-09-09 |
| lcs-recomp | GITHUB | 149 | 2026-09-24 |
| MacMind | HN | 159 | 2026-04-16 |
| Duplicate 3 layers in a 24B LLM, logical deduction .22→.76. No training | HN | 265 | 2026-03-18 |
| Covalent-MAS | GITHUB | 337 | 2026-09-19 |
| electrical-engineering | GITHUB | 416 | 2026-09-20 |
| recurrent-looped-tranformer | GITHUB | 853 | 2026-09-12 |