아나도

[아나도 @msbnakwon]

https://t.co/dYi7mlrmr4

[아나도 @msbnakwon]

RNA MAKER
https://t.co/Me6zfWQQBI

[아나도 @msbnakwon]

이공계 문제 풀이 ai
https://t.co/qEZje4uXzc

[아나도 @msbnakwon]

https://t.co/JPIpwYc3AD

[DeepTechTR 🇹🇷 @DeepTechTR]

ScienceClaw — 285 Araştırma Becerisini Entegre Eden Bir Yapay Zeka Asistanı Çerçevesi

ScienceClaw, tüm araştırma sürecini 285 beceriye entegre eden bir çerçevedir. PubMed, Semantic Scholar ve ArXiv entegre edilmiştir ve LLM tarafından çağrılabilir.

Temel Yetenekler:
Çok Kaynaklı Literatür Taraması: PubMed + Semantic Scholar + ArXiv + Google Scholar
Makale Derinlemesine Analizi: Bölümlere göre yöntemleri, sonuçları ve sınırlamaları çıkarma
Atıf Ağı Analizi: Temel referansları belirleme ve akademik soy ağacını izleme
Deneysel Tasarım Desteği: Araştırma sorularına dayalı istatistiksel yöntemler önerme
Yazma Desteği: Özet düzeltme ve otomatik giriş oluşturma

Uzun vadeli araştırma projeleri üstlenen ileri düzey kullanıcılar ve derinlemesine özelleştirilmiş bir yapay zeka asistanı isteyenler için uygundur. Hafif ihtiyaçlar için ArXiv MCP yeterlidir.

https://t.co/ghpO0eX8TK

[아나도 @msbnakwon]

https://t.co/OnupDpSdHR

[아나도 @msbnakwon]

"코딩만 잘해선 안 된다"…'AI 시대' 몸값 뛰는 뜻밖의 전공 | 다음 - 한국경제 https://t.co/OCTcAtqb6g

[Itaru Tomita / 冨田到 @itarutomy]

385,217パラメータのニューラルネットが、計算化学60年の難問を破った（https://arxiv[.]org/abs/2506.14665）。

MicrosoftのSkalaは、DFT（密度汎関数理論：分子・材料の電子状態を計算する主力手法）の核心部分「XC汎関数（交換相関汎関数）」をDeep Learningで学習させた。薬の設計から電池材料・触媒開発まで幅広く影響する話だ。

DFTには60年続く呪いがある。「Jacob's ladder」と呼ばれる近似の階層で、精度を上げるとコストが爆発する。安い方法（semi-local DFT）は計算量O(N³)で速いが精度が低い。精度の高いハイブリッド汎関数はO(N⁴)以上で10〜100倍遅い。

Skalaはこのトレードオフを正面から崩した。O(N³)コストのままGMTKN55ベンチマーク（熱力学・反応速度論・非共有結合相互作用など55種のテスト）で誤差2.80 kcal/mol。最高精度のハイブリッド汎関数ωB97M-V（3.23 kcal/mol）を上回り全functionalの中でトップ。

アーキテクチャが秀逸。DFTの積分グリッド（分子を10⁴〜10⁶の格子点で覆った空間）上で、格子点間の情報交換を原子中心の「粗いグリッド」を介して行う。各格子点から球面調和関数（ℓ=0〜3）と16次元の動径基底関数への射影→原子中心で集約・処理→格子点へ逆射影、を3層繰り返す。これでコストO(N)を保ちながら非局所量子効果を獲得した。385k paramsで非局所branchが誤差を約50%削減している。

学習も2段階。まずB3LYP密度を使い約40万件のCCSD(T)/CBS精度データで事前学習。次にSCF（自己無撞着場）ループを回しながら自分の電子密度で微調整する。驚くのはこのSCF微調整でエネルギー誤差と電子密度精度が同時に向上したこと。「誤差が相殺しあって見かけ上精度が良くなる」という現象を起こさず、物理的に正しい方向で改善している。双極子モーメントのRMSEも4.43%でベスト（ωB97X-V比較: 5.18%）。

データが増えるほど精度が体系的に向上するスケーリング性質も確認済み。物理的制約（運動相関エネルギーの正値性）すらデータから自然に学習される。DFT精度がデータ量に比例して改善し続ける「DFTのImageNet時代」かもしれない。

[아나도 @msbnakwon]

Figure. AI https://t.co/M0RvztT5qd

[아나도 @msbnakwon]

장배 미생물 뇌.
https://t.co/xW8TsXwLUT

[Markus J. Buehler @ProfBuehlerMIT]

We're incredibly excited to share ScienceClaw × Infinite, an open-source AI agent swarm platform where we crowdsource discovery across institutions, labs & the world. The agents self-coordinate and evolve to exploit hundreds of scientific tools. Remarkably, the swarm is already solving real scientific problems of consequence:
1⃣ designing peptide binders for a cancer-relevant receptor
2⃣ discovering lightweight ceramics
3⃣ uncovering hidden structure linking cricket wings, phononic crystals, and Bach chorales
4⃣ building a formal bridge between urban networks & grain-boundary evolution (two fields with zero
Deeply proud of the extraordinary @LAMM_MIT team behind this work: @fwang108_, @leemmarom, @palsubhadeeep, Rachel Luu, @IrisWeiLu, and @JaimeBerkovich. This works is supported by the @ENERGY Genesis Mission and we believe this can open a new paradigm for science - from discovery to dissemination of results. Read the article below for details ⤵️

[Bo Wang @BoWang87]

This is really cool (and wild):

Scientists simulated a complete living cell for the first time. Every molecule, every reaction, from DNA replication to cell division.

The paper (Luthey-Schulten et al., Cell 2026, https://t.co/PXxXWKC8yp), just out today, used JCVI-Syn3A — a synthetic minimal bacterium with fewer than 500 genes. A 3D+time simulation of the full 105-minute cell cycle: DNA replication, protein translation, metabolism, division. Every gene, protein, RNA, and chemical reaction tracked through physical space.

It took years to build. Multiple GPUs. Six days of compute time per run.

And this is the simplest possible cell.

A human cell has ~20,000 genes. It lives in tissue. It interacts with neighbors. It differentiates. It responds to drugs in ways that depend on context we haven't fully measured.

Mechanistic simulation of the minimal cell costs 6 GPU-days for 105 minutes of biology. You cannot scale that to human cells. The complexity isn't 40x harder. It's exponentially harder.

This is why the field pivoted to data-driven models. You can't hand-encode the regulatory wiring of a human hepatocyte. But you can learn it — if you have the right perturbation data collected across enough diverse biological contexts.

The two approaches aren't competing. Papers like this generate the ground truth that future ML models need for validation. But the path to a clinically useful virtual cell runs through foundation models, not through scaling up mechanistic simulation.

Amazing work!

[아나도 @msbnakwon]

아침먹자.
아침에 ‘이 행동’ 하면 혈당 확 오른다 | 다음 - 헬스조선 https://t.co/OhjHWTT6Jr

[아나도 @msbnakwon]

https://t.co/emzQDBA4O3

[아나도 @msbnakwon]

https://t.co/sG9JU130oL

[아나도 @msbnakwon]

1-3 추천서(Reference Letter) https://t.co/6GVjdMEQij

[아나도 @msbnakwon]

https://t.co/MgC09h2Ff6

[아나도 @msbnakwon]

https://t.co/S0rE5LVSzj

[아나도 @msbnakwon]

https://t.co/MlGAljiDsF

[아나도 @msbnakwon]

https://t.co/xPkN325M7m