NCMNext-Concept Modeling: Fast, Parallel Coarse-to-Fine Text Generationscalingintelligence.stanford.edu·2 pts·matt_d·0
Apples to Apples: Getting Type Mappings Right in LLM-Based Code Translationcodemetal.ai·8 pts·matt_d·0
Heterogeneity in AI: Pathways from Inference to Compound Systems (Dissertation)doi.org·2 pts·matt_d·0
Haskell: Origins, evolution, and future – Peyton Jones – JuliaCon Global 2026 [video]youtube.com·2 pts·matt_d·0
VibeSys Builds a Qwen3.5-397B Engine on MI300A, 2.3× Tuned SGLangsyfi.cs.washington.edu·2 pts·matt_d·0
What mathematicians should know about the Lean Theorem Prover: reliability & AIterrytao.wordpress.com·210 pts·matt_d·58
Ahead-of-Time Compilation of JavaScript to WebAssembly via Static Type Inference [video]vimeo.com·1 pts·matt_d·0
A Self-Pruning Transformer: Extreme KV-Cache Compression w/Universal Attentionarxiv.org·6 pts·matt_d·0
ServeLearnBench: How Well Can Agents Self-Improve from Serving Experience?infini-ai-lab.github.io·1 pts·matt_d·0
Slot Machine Programming and the Hidden Curriculumthelastsoftwareengineer.substack.com·2 pts·matt_d·0