rasbt proposes an LLM building-block API to cut switching costs in open-source LLM research
TheTuringPost · x · 2026-11-17
Sebastian Raschka argues the open-source LLM ecosystem needs an "LLM building-block API." Today, components are scattered across codebases, making it painful to swap GQA for MLA, dense FFNs for MoE, attention for Mamba2/Gated DeltaNet, or standard residuals for attention residuals in experiments and ablations—even with LLM-assisted coding.
He envisions it built on top of PyTorch or CUDA, but carefully engineered and maintained rather than weekend vibe-coded, and efficient enough for researchers to trust. Others echoed the call for "LEGO for LLMs" as every new open model rewrites the same attention block with a twist.
Related event: Raschka calls for building-block API for open LLMs(2 posts)→
More from coding & agent
- Decision Models upgrades playground with image input, showing probabilities and per-request cost — airesearch12 · 2026-11-17
- Microsoft owns every repo on GitHub—so why does Copilot still suck at coding? — prasenx · 2026-11-17
- Baseten open-sources Rust/CUDA DeepSeek engine; agent on Vera Rubin finds optimizations speeding Blackwell by 10% — baseten · 2026-11-17
- shadcn's underrated tip: ask your coding agent "are you sure?" before it writes any code — shadcn · 2026-11-17
- Stanford's CS329A Self-Improving AI Agents course: 9 free lectures, 10+ hours on YouTube — techNmak · 2026-11-17
- 'Stuck at 99%': a single-file HTML toy progress bar with fully synthesized Web Audio — prasenx · 2026-11-17