[
  {
    "authorId": "OpenAI",
    "text": "Introducing GPT-5.5 — our most capable model yet. Improved reasoning, faster response times, and native tool use. Available today in the API.",
    "createdAt": "2026-04-16T22:00:00.000Z",
    "url": "https://x.com/OpenAI/status/100000000000000001"
  },
  {
    "authorId": "AnthropicAI",
    "text": "Claude Opus 4.7 is here. Extended thinking with thought signatures enables multi-step reasoning chains that persist across conversations.",
    "createdAt": "2026-04-16T21:30:00.000Z",
    "url": "https://x.com/AnthropicAI/status/100000000000000002"
  },
  {
    "authorId": "GoogleDeepMind",
    "text": "Gemini 3.1 Pro achieves state-of-the-art on mathematical reasoning benchmarks with our new thinking_level parameter.",
    "createdAt": "2026-04-16T20:15:00.000Z",
    "url": "https://x.com/GoogleDeepMind/status/100000000000000003"
  },
  {
    "authorId": "kaborosu",
    "text": "GPT-5.5のtool useがやばい。MCPサーバー3つ同時に繋いで、指示一発で調査→コード生成→テスト実行まで全自動で回った。エージェント時代が本当に来た。",
    "createdAt": "2026-04-16T23:00:00.000Z",
    "url": "https://x.com/kaborosu/status/100000000000000004"
  },
  {
    "authorId": "ai_database",
    "text": "速報: MicrosoftがGitHub Copilot Workspaceを全ユーザーに無料開放。リポジトリ単位でエージェントが動き、Issue→PR→マージまで自動化。",
    "createdAt": "2026-04-16T19:00:00.000Z",
    "url": "https://x.com/ai_database/status/100000000000000005"
  },
  {
    "authorId": "ylecun",
    "text": "New paper from FAIR: 'World Models for Robotic Manipulation' — our approach achieves 94% success rate on unseen objects using learned physics priors.",
    "createdAt": "2026-04-16T18:30:00.000Z",
    "url": "https://x.com/ylecun/status/100000000000000006"
  },
  {
    "authorId": "sama",
    "text": "We're releasing the OpenAI Agents SDK v2 today. Build production-grade agents with built-in memory, tool orchestration, and safety guardrails.",
    "createdAt": "2026-04-16T17:00:00.000Z",
    "url": "https://x.com/sama/status/100000000000000007"
  },
  {
    "authorId": "emaborai",
    "text": "Cursorの新機能Background Agentを使ってみた。寝てる間にPR作ってレビューコメントまで付けてくれた。朝起きたらマージするだけ。生産性が次元変わる。",
    "createdAt": "2026-04-16T22:30:00.000Z",
    "url": "https://x.com/emaborai/status/100000000000000008"
  },
  {
    "authorId": "MetaAI",
    "text": "Llama 4 Behemoth is now available. 2T parameter MoE with 256 experts. Open weights for research and commercial use.",
    "createdAt": "2026-04-16T16:00:00.000Z",
    "url": "https://x.com/MetaAI/status/100000000000000009"
  },
  {
    "authorId": "hardmaru",
    "text": "The gap between open and closed models continues to shrink. Llama 4 Behemoth matches GPT-5 on most benchmarks. Exciting times for the open source community.",
    "createdAt": "2026-04-16T16:30:00.000Z",
    "url": "https://x.com/hardmaru/status/100000000000000010"
  },
  {
    "authorId": "ai_and_design",
    "text": "Figma AIの新機能、コンポーネントの意味を理解してデザインシステムに沿った提案をしてくれる。デザイナーの仕事が設計判断に集中できるようになる。",
    "createdAt": "2026-04-16T15:00:00.000Z",
    "url": "https://x.com/ai_and_design/status/100000000000000011"
  },
  {
    "authorId": "StabilityAI",
    "text": "Stable Diffusion 4 is here. Real-time generation at 60fps, native video support, and unprecedented prompt adherence. Try it now on stability.ai.",
    "createdAt": "2026-04-16T14:00:00.000Z",
    "url": "https://x.com/StabilityAI/status/100000000000000012"
  },
  {
    "authorId": "tech_and_future",
    "text": "AI Agent同士が自律的に交渉してタスクを分担する研究が出た。マルチエージェント協調がいよいよ実用段階に入ってきている。",
    "createdAt": "2026-04-16T13:00:00.000Z",
    "url": "https://x.com/tech_and_future/status/100000000000000013"
  },
  {
    "authorId": "GoogleAI",
    "text": "Announcing Gemini Nano 2 — on-device AI that runs entirely on your phone. No internet needed. Privacy-first design with 3B parameters.",
    "createdAt": "2026-04-16T12:00:00.000Z",
    "url": "https://x.com/GoogleAI/status/100000000000000014"
  },
  {
    "authorId": "aisafety_news",
    "text": "EU AI Act enforcement begins today. Companies deploying high-risk AI systems must now demonstrate compliance with transparency and safety requirements.",
    "createdAt": "2026-04-16T11:00:00.000Z",
    "url": "https://x.com/aisafety_news/status/100000000000000015"
  },
  {
    "authorId": "kaggle",
    "text": "New competition: $1M prize for building an AI system that can reliably detect AI-generated scientific papers. Submissions open until June 2026.",
    "createdAt": "2026-04-16T10:00:00.000Z",
    "url": "https://x.com/kaggle/status/100000000000000016"
  },
  {
    "authorId": "ai_business_jp",
    "text": "日本のAIスタートアップの資金調達額が2025年比で3倍に。特にエンタープライズ向けRAGとエージェント基盤に投資が集中。",
    "createdAt": "2026-04-16T09:00:00.000Z",
    "url": "https://x.com/ai_business_jp/status/100000000000000017"
  },
  {
    "authorId": "nvidia",
    "text": "Blackwell Ultra GPUs now shipping. 2x the inference performance of H200 at the same power. Built for the age of AI agents.",
    "createdAt": "2026-04-16T08:00:00.000Z",
    "url": "https://x.com/nvidia/status/100000000000000018"
  },
  {
    "authorId": "dev_tools_ai",
    "text": "GitHub Actions now supports native AI agent steps. Define an agent in YAML, give it tools, and let it solve issues autonomously in your CI pipeline.",
    "createdAt": "2026-04-16T07:00:00.000Z",
    "url": "https://x.com/dev_tools_ai/status/100000000000000019"
  },
  {
    "authorId": "amasaki",
    "text": "Claude Codeのスキル機能を使い始めて2週間。同じ作業を3回やったらスキルにする、というルールを徹底したら作業速度が5倍になった。",
    "createdAt": "2026-04-16T22:45:00.000Z",
    "url": "https://x.com/amasaki/status/100000000000000020"
  },
  {
    "authorId": "veraborai",
    "text": "Vercel v0がついにフルスタック対応。フロントだけじゃなくAPIルートもDBスキーマも一発で生成してくれる。プロトタイピングの速度が異次元。",
    "createdAt": "2026-04-16T21:00:00.000Z",
    "url": "https://x.com/veraborai/status/100000000000000021"
  },
  {
    "authorId": "research_ml",
    "text": "Interesting trend: more papers using 'test-time compute' scaling instead of pre-training scaling. The paradigm shift from bigger models to smarter inference is real.",
    "createdAt": "2026-04-16T20:00:00.000Z",
    "url": "https://x.com/research_ml/status/100000000000000022"
  },
  {
    "authorId": "aicoding_daily",
    "text": "Cursor、Windsurf、Claude Code、GitHub Copilot Workspace。AIコーディングツールの選択肢が増えすぎて逆に迷う時代。結局はコンテキスト管理が上手いツールが勝つ。",
    "createdAt": "2026-04-16T19:30:00.000Z",
    "url": "https://x.com/aicoding_daily/status/100000000000000023"
  },
  {
    "authorId": "huggingface",
    "text": "Introducing SmolAgent 2.0 — build agents in 10 lines of Python. Now with native MCP support and automatic tool discovery.",
    "createdAt": "2026-04-16T18:00:00.000Z",
    "url": "https://x.com/huggingface/status/100000000000000024"
  },
  {
    "authorId": "ai_regulation",
    "text": "米国AIガバナンス法案が上院を通過。企業はAI利用の透明性レポートを四半期ごとに公開することが義務化される見込み。",
    "createdAt": "2026-04-16T17:30:00.000Z",
    "url": "https://x.com/ai_regulation/status/100000000000000025"
  },
  {
    "authorId": "robotics_ai",
    "text": "Boston Dynamics Atlas now powered by Gemini. The robot can understand natural language instructions and plan multi-step physical tasks autonomously.",
    "createdAt": "2026-04-16T15:30:00.000Z",
    "url": "https://x.com/robotics_ai/status/100000000000000026"
  },
  {
    "authorId": "indie_hacker_jp",
    "text": "AIツールだけで月収100万円のSaaSを一人で運営する人が増えてきた。技術力じゃなくて課題発見力が差になる時代。まさにAI-Drivenの世界。",
    "createdAt": "2026-04-16T14:30:00.000Z",
    "url": "https://x.com/indie_hacker_jp/status/100000000000000027"
  },
  {
    "authorId": "perplexity_ai",
    "text": "Perplexity Pro now includes real-time web search with citations, code execution, and file analysis. All in one conversational interface.",
    "createdAt": "2026-04-16T13:30:00.000Z",
    "url": "https://x.com/perplexity_ai/status/100000000000000028"
  },
  {
    "authorId": "edge_ai_news",
    "text": "Apple Intelligence 2.0 launches with on-device reasoning capabilities. Siri can now complete multi-step tasks without sending data to the cloud.",
    "createdAt": "2026-04-16T12:30:00.000Z",
    "url": "https://x.com/edge_ai_news/status/100000000000000029"
  },
  {
    "authorId": "ml_ops_daily",
    "text": "The real bottleneck in AI deployment isn't the model — it's evaluation. Teams spending 60% of their time building eval frameworks instead of shipping features.",
    "createdAt": "2026-04-16T11:30:00.000Z",
    "url": "https://x.com/ml_ops_daily/status/100000000000000030"
  }
]