Developer builds 8M-parameter LLM lab, asks about sharing

A developer built and trained an ~8M-parameter LLM from scratch in PyTorch as a lab for implementing research-paper ideas. Training ran 60,000 steps; next steps are KV caching, PagedAttention and Orca. He asked whether the community welcomes posts about these experiment blogs.

• Developer releases manual transformer autograd in PyTorch2026-10-28
• Practical notes for GPU kernel dev setup2026-10-27
• Deepity releases C++/CUDA predictive coding library, edges JAX baseline on MNIST2026-10-27
• Raschka releases from-scratch reasoning model video course2026-10-26
• Developer proposes speculative decoding using 2-bit draft plus RAM residual2026-10-22