https://www.youngju.dev/blog 천천히 올바르게. AI Researcher & DevOps Engineer Youngju's tech blog. GPU/CUDA, LLM, MLOps, Kubernetes AI workloads, distributed training, and data engineering. ko fjvbn2003@gmail.com (Youngju Kim) fjvbn2003@gmail.com (Youngju Kim) Sat, 16 May 2026 00:00:00 GMT https://www.youngju.dev/blog/culture/2026-05-16-modern-statistical-computing-2026-r-4-5-posit-rstudio-stan-pyro-numpyro-brms-jax-tidyverse-polars-deep-dive.en https://www.youngju.dev/blog/culture/2026-05-16-modern-statistical-computing-2026-r-4-5-posit-rstudio-stan-pyro-numpyro-brms-jax-tidyverse-polars-deep-dive.en In 2026 statistical computing has two large currents running in parallel. On one side R 4.5 (April 2025) and Posit (formerly RStudio) commercial stack are cementing the reproducible-research standard. On the other side JAX, NumPyro, and Polars are accelerating Bayesian inference and dataframe analytics on GPUs and TPUs. Between them Stan keeps the absolute lead on NUTS sampling, while brms and rstanarm bring Bayesian methods into the daily toolbox of R users. Marimo overcomes Jupyter limits with reactive notebooks, and Quarto unifies R, Python, Julia, and Observable in one document. This article maps the entire 2026 landscape — R ecosystem, Stan/Pyro Bayesian stack, JAX scientific computing, Polars dataframes, visualization, causal inference, and the Korean and Japanese statistical-computing communities — in one continuous read. Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) statistical-computingr-languagepositrstudiostanpyronumpyrobrmsjaxtidyversedata-tablepolarsmarimoenglish https://www.youngju.dev/blog/culture/2026-05-16-modern-statistical-computing-2026-r-4-5-posit-rstudio-stan-pyro-numpyro-brms-jax-tidyverse-polars-deep-dive.ja https://www.youngju.dev/blog/culture/2026-05-16-modern-statistical-computing-2026-r-4-5-posit-rstudio-stan-pyro-numpyro-brms-jax-tidyverse-polars-deep-dive.ja 2026年の統計コンピューティングは二つの大きな潮流が並行して進む。一方では R 4.5(2025年4月)と Posit(旧 RStudio)の商用スタックが再現可能研究の標準を固め、もう一方では JAX・NumPyro・Polars が GPU と TPU の上でベイズ推論とデータフレームを加速する。その間で Stan は NUTS サンプラーの絶対王者の地位を守り、brms と rstanarm は R 利用者にベイズを日常の道具にする。Marimo はリアクティブノートブックで Jupyter の限界を破り、Quarto は R・Python・Julia・Observable を一つの文書に束ねる。本稿では R 生態系・Stan/Pyro ベイズスタック・JAX 科学計算・Polars データフレーム・可視化・因果推論・韓国と日本の統計コンピューティングコミュニティまで、2026 年の風景を一つの流れで整理する。 Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) statistical-computingr-languagepositrstudiostanpyronumpyrobrmsjaxtidyversedata-tablepolarsmarimo日本語 https://www.youngju.dev/blog/culture/2026-05-16-modern-statistical-computing-2026-r-4-5-posit-rstudio-stan-pyro-numpyro-brms-jax-tidyverse-polars-deep-dive https://www.youngju.dev/blog/culture/2026-05-16-modern-statistical-computing-2026-r-4-5-posit-rstudio-stan-pyro-numpyro-brms-jax-tidyverse-polars-deep-dive 2026년 통계 컴퓨팅은 두 개의 큰 흐름이 동시에 진행 중이다. 한쪽에서는 R 4.5(2025-04)와 Posit(구 RStudio)의 상용 스택이 재현가능 연구의 표준을 굳히고, 다른 한쪽에서는 JAX·NumPyro·Polars가 GPU·TPU 위에서 베이지안과 데이터프레임을 가속한다. 그 사이에서 Stan은 NUTS 샘플러의 절대강자 자리를 지키고, brms와 rstanarm은 R 사용자에게 베이지안을 일상 도구로 만든다. Marimo는 Jupyter의 한계를 반응형 노트북으로 돌파하고, Quarto는 R·Python·Julia·Observable을 한 문서에 묶는다. 이 글은 R 생태계·Stan/Pyro 베이지안 스택·JAX 과학 컴퓨팅·Polars 데이터프레임·시각화·인과추론·한국과 일본의 통계 커뮤니티까지 통계 컴퓨팅 풍경을 한 흐름으로 정리한다. Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) statistical-computingr-languagepositrstudiostanpyronumpyrobrmsjaxtidyversedata-tablepolarsmarimo https://www.youngju.dev/blog/culture/2026-05-16-notebooks-2026-jupyter-marimo-quarto-observable-deepnote-hex-positron-deep-dive.en https://www.youngju.dev/blog/culture/2026-05-16-notebooks-2026-jupyter-marimo-quarto-observable-deepnote-hex-positron-deep-dive.en The notebook ecosystem split again in 2026. Jupyter Lab 5 and Notebook 7 became the de facto standard, but Marimo (June 2024) attacked the "Hidden state" problem head-on with a reactive model, and Posit reincarnated RStudio as Positron. Quarto unified R, Python and Julia into one publishing pipeline, while Mike Bostock turned Observable into a static-site generator with Observable Framework. Deepnote and Hex split the collaborative SaaS axis, and Curvenote, Polars Notebook and Pluto.jl fill the niches. From Kakao and Toss in Korea to Mercari and ZOZO in Japan — we redraw the map across four categories: classic, reactive, publishing, SaaS. Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) notebookjupytermarimoquartoobservabledeepnotehexpositronrstudiorpythonjuliadata-sciencereactive2026deep-diveenglish https://www.youngju.dev/blog/culture/2026-05-16-notebooks-2026-jupyter-marimo-quarto-observable-deepnote-hex-positron-deep-dive.ja https://www.youngju.dev/blog/culture/2026-05-16-notebooks-2026-jupyter-marimo-quarto-observable-deepnote-hex-positron-deep-dive.ja 2026年のノートブック生態系はまた分岐した。Jupyter Lab 5 と Notebook 7 が事実上の標準になった一方で、Marimo(2024年6月)はリアクティブモデルで「隠れた状態(Hidden state)」問題に正面から挑み、Posit は RStudio を Positron として生まれ変わらせた。Quarto は R・Python・Julia をひとつの出版パイプラインに統合し、Mike Bostock の Observable Framework はノートブックを静的サイトジェネレーターに変えた。Deepnote と Hex は協業型 SaaS の二大軸として分かれ、Curvenote・Polars Notebook・Pluto.jl が隙間を埋める。韓国の カカオ・トス、日本の Mercari・ZOZO の実例まで含め、クラシック・リアクティブ・出版・SaaS の 4 分類でノートブック地図を描き直す。 Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) notebookjupytermarimoquartoobservabledeepnotehexpositronrstudiorpythonjuliadata-sciencereactive2026deep-dive日本語 https://www.youngju.dev/blog/culture/2026-05-16-notebooks-2026-jupyter-marimo-quarto-observable-deepnote-hex-positron-deep-dive https://www.youngju.dev/blog/culture/2026-05-16-notebooks-2026-jupyter-marimo-quarto-observable-deepnote-hex-positron-deep-dive 2026년 노트북 생태계는 한 번 더 갈라졌다. Jupyter Lab 5와 Notebook 7이 사실상 표준이 되었지만, Marimo(2024년 6월)의 반응형 모델이 "Hidden state" 문제를 정면으로 깼고, Posit은 RStudio를 Positron으로 다시 태어나게 했다. Quarto는 R·Python·Julia를 하나의 출판 파이프라인으로 묶었고, Mike Bostock의 Observable Framework는 노트북을 정적 사이트 생성기로 바꿨다. Deepnote와 Hex는 협업 SaaS 두 축으로 갈라졌으며, Curvenote·Polars Notebook·Pluto.jl이 틈새를 채운다. 카카오·토스의 분석 환경, Mercari·ZOZO의 데이터팀 노트북 활용까지 — 클래식·반응형·출판·SaaS 4분류로 노트북 지도를 다시 그린다. Sat, 16 May 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) notebookjupytermarimoquartoobservabledeepnotehexpositronrstudiorpythonjuliadata-sciencereactive2026deep-dive