
ChatGPT驚き屋/JunnosukeTachibana
12.1K posts

ChatGPT驚き屋/JunnosukeTachibana
@smartAINews
毎日ブラウザLLMの挙動を観測・検証。 「最近 AI使いこなせていない…?」と思ったら、実は AI側に問題があるかも。 バグやアプデの予兆、影響も発信中。 複数のAIに同じ質問を投げ続け、日々の変化を追跡しています。open AI、Gemini、claudeの22モデル、26年2月5日に56400サンプル突破






・皇宮警察予備隊 存立危機事態対応戦略支援装備(=核兵器)の運用のために設立された組織。 憲法改正により内閣の助言への拒否権を持つ天皇の統帥を受けるため、日本は事実上の核の先制使用を封じているとされる。 #大嘘 #AI画像

Can confirm OpenAI's next model will be called GPT-5.5 Appeared internally recently and have tried the model (should be noted that this probably isn't the release checkpoint of the model). My first impressions are that it is better than GPT-5.4, but not markedly so - it will not be particularly competitive with Mythos/Tiramisu/Capybara from Anthropic Looks like we'll have to wait until GPT-6 for that

デマになると怖いので情報を上げるのを控えていたけど、来週中にもGPT-5.5(Spudかどうかは不明)がリリースされるという噂

Can confirm OpenAI's next model will be called GPT-5.5 Appeared internally recently and have tried the model (should be noted that this probably isn't the release checkpoint of the model). My first impressions are that it is better than GPT-5.4, but not markedly so - it will not be particularly competitive with Mythos/Tiramisu/Capybara from Anthropic Looks like we'll have to wait until GPT-6 for that



