pftq
1.7K posts

pftq
@pftq
Self-taught. Coding, film, music, games, trading. Ambidextrous & plays tennis with two rackets. Fears boredom. Rarely sleeps. @AutodidacticI @TechTraderAI

「おはよう」って言われるのを待っているねこ

Doing my usual poker simulation test with the new @Grok 4.2 multi-agent, and it cleanly breaks past the 70% win rate that all older models have been stuck under. The multi-agent makes a big difference. @xai @elonmusk And GPT-5.4 as usual doing worse with each update.

congrats on the launch



海外の人に聞きたい! リプライ(返信)にGIFや絵文字を使ったらアカウントはロックされないんですか?


In your opinion what is the biggest threat to humanity


@mrmikeMTL Language, in all of its forms. The Gospel, poetry, music, novels... all impossible without the invention of language. "In the beginning was the Word, and the Word was with God, and the Word was God." John 1 : 1

信じられないかもしれないけど 日本では 自分の金で買った家に 死ぬまで税金を払い続けます 買う時も、住んでる時も、売る時も 全部取られます しかも死んだあと 子供にまで相続税がかかります 自分の家なのに。






【HappyHorse-1.0 要約】 happyhorse-ai.com artificialanalysis.ai/video/arena ■ 核心 中国発の新星オープンソースモデル。15Bパラメータながら、テキスト-to-ビデオと画像-to-ビデオの両方で現在Artificial Analysis Video Arena総合1位を獲得。 特に「複数カット(multi-shot)」の安定性と、細かいプロンプトへの忠実度が現行どの閉源モデルよりも優れていると評価されています。 ■ 主要特徴 ・ビデオ+音声を1モデルで同時生成(音声付き動画がそのまま作れる) ・8ステップ推論で超高速(256pなら2秒、1080pでも38秒程度) ・リポジトリ不要・多言語対応(日本語・英語・中国語・韓国語など) ・近日中にモデル重量・コード完全公開予定 ■ 背景 これまでは閉源モデル(Veo、Kling、Runwayなど)が圧倒的だった動画生成領域で、初めて「オープンソース+音声付き+爆速+高品質」が揃った衝撃モデル。 Justine Moore(a16z)が本気で推している点からも、本物度が非常に高いです。











