blogHow to write a technical blog post about your own experiments, the kind a strong deep-learning researcher would publish: the story arc from the surprise that started the work to open questions and related work, a voice after Sasha Rush, Phillip Isola and Lilian Weng, model posts to learn from (Carte
paper-voiceThe prose voice of pre-2023 papers by Kaiming He, Joseph E. Gonzalez and Alex Dimakis, measured on 15 arXiv sources, plus Dimakis's own writing notes ("Ideas for better writing", 2018), for drafting or rewriting research-paper prose that reads as LLM-written. Load it with writing:style for any paper
plotOne matplotlib style for charts in posts and papers, after Epoch AI's charts: Instrument Sans at Epoch's measured sizes, margins and gaps; a light grid both ways and only the baseline axis; the quantity above each column and panel names inside; Google's GM2 tones, toned by mark size; layout set in i
preflightMechanical submission preflight for a LaTeX paper directory: compile gate, page count against the venue cap, build-log scan for undefined citations and overfull boxes, leftover placeholders and \\textcolor{red} marks, correspondence-problem density, and .bib hygiene. Deterministic, no model judgment
styleDefault writing standards. Use whenever producing English prose the user will send or publish: emails, message drafts, blog posts, tweets, documentation, grant proposals, and conference/journal paper drafts (NeurIPS, ICML, ICLR, ACL, CVPR, COLM, EMNLP, arxiv). Applies 18 canonical rules everywhere;