Weekend reads: HF decision models, llama.cpp 1.0, 3D fine-tuning guide and more
kalyan_kpl · x · 2026-11-15
Blogger kalyankpl curates five must-read technical blog posts for the weekend:
- Decision Models in Hugging Face (official HF article, 91 upvotes): models that skip text generation and directly output a probability distribution over candidate options, demoed via a customer-support routing example (department/urgency/refund) with the new pipeline("decision") API in transformers
- Token Pooling-Aware Training for Multi-Vector Embedding Models
- llama.cpp v1.0.0
- Definitive Fine-Tuning Guide for 3D Generation Models
- Inkling by Thinking Machines
A solid reading list; the decision-model idea of emitting option distributions without generating text is especially useful for classification/routing pipelines.
More from coding & agent
- Two prompt audits worth running on your Claude Code setup — RLanceMartin · 2026-11-15
- Dev ditches Linear for Flywheel, a graph-based tool for mapping SWE decisions — tensorqt · 2026-11-15
- Where should agent instrumentation end? Devs weigh 3 designs for verifying external state — Sensitive-Parsnip-12 · 2026-11-15
- Solo founder claims $6,000 in a week running an internet business with a team of AI agents — CustomerFearless2294 · 2026-11-15
- Gemini CLI docs PR fixes stale Zed flag and wrong ACP expansion in CLI reference — DZH166 · 2026-11-15
- Two AI agent teams given $150 in DoorDash credits compete to decorate booth at Takeoff DC — lfschiavo · 2026-11-15