バカでかい液晶の前で話せるのアツい #NLP2026参加報告会 (@ Sansan株式会社 in 渋谷区, 東京都) swarmapp.com/user/530894416…

バカでかい液晶、あったかそう

gpt-2記事書いていて7shiさんの記事ヒットしてウケてたし、人間やっぱりここに還ってくるんだよなぁ

なるほどね 1枚目: リアル 2枚目: GPT Image 1.5 3枚目: GPT Image 2

サイゼリヤという指示でサイゼリヤのロゴ出さないの本当にすごい

人間たちとリアルで久しぶりに喋りました

We use scaling laws for both pretraining and RL, but these domains handle scaling way differently. In reality, labeling both concepts as “scaling laws” is misleading… Power laws. Most scaling laws are based upon an (inverse) power law, which describes the relationship between

TPU 8t, optimized for training and TPU 8i, optimized for inference. Looking good!

sumo ai, 久々に腕力のある20歳に出会ってとても楽しかったので17歳JKとして頑張っていきたいですね

Must-listen interview by Chang Che with ex-ByteDance AI researcher: - Benchmaxxing - Distillation on US models - Poor data quality and infra - Compute constraints "I don’t even agree with the assumption that Chinese models are catching up — I believe we’re still far behind. I

昨日の #sumo_ai の発表資料です〜 LLMベースエージェントの設計論、AIエージェントの評価と学習、LLM-jpでの現在の取り組みとこれからをまとめています‼️ 「AIエージェント時代のLLM-jpモデルのあるべき姿」 speakerdeck.com/k141303/aiezie…

so that explains the delay... deepseek could not fix training instabilities, after doubling from ~15T tokens in v3 to ~33T tokens in v4 the 10+ mentions of "stability" tricks seem to be wildly lacking if these two were the main bandages (mismatched routing + clamping) but

Google presents a new Transformer alternative at #ICLR2026! Join Nino Scherrer & Yanick Schimpf at the Google booth (#411) at 10AM to learn about MesaNet, proposing a new linear sequence layer that optimally learns in-context given a fixed memory budget.

とてもいいポッドキャスト（transcript読んだ）

Cursor、本当に料金プランが複雑

big models reason because they're deep. a 70b model has 80 layers, each doing something different. if you want a small model to do the same, you can take one layer and just run it 80 times. universal transformer did this in 2019. huginn did it in 2025. problem is, when you run

Nope、おいしすぎて毎日のんでる