{
    "news": [
        {
            "date": "2025-06-07",
            "video_link": "https://www.youtube.com/watch?v=4CV-fB5EVJs",
            "twitter_text": "New Benchmark Results: Claude-sonnet-4 & Claude-opus-4 and Join Us in the Discord!",
            "twitter_link": "https://x.com/haoailab/status/1931438794756313530"
        },
        {
            "date": "2025-06-03",
            "video_link": "https://www.youtube.com/watch?v=m6i9L6-pgu4",
            "twitter_text": "New Benchmark Results: How do top open-source models like Deepseek r1 & Qwen 3 perform on games?",
            "twitter_link": "https://x.com/haoailab/status/1929997363407708646"
        },
        {
            "date": "2025-04-28",
            "video_link": "https://www.youtube.com/watch?v=OEQRhBKYxIE",
            "twitter_text": "Grok-3-mini-beta Joins the Battle: Outperforms Gemini 2.5 Flash, Challenges O3-mini Across Games — Full Grok-3-beta Power Yet to Come. 🚀",
            "twitter_link": "https://x.com/haoailab/status/1917309598861779021"
        },
        {
            "date": "2025-04-24",
            "video_link": "https://www.youtube.com/watch?v=NB1-5aKV9v4",
            "twitter_text": "Zero-Shot AI Gaming Showdown: O3 Multi-Modal Might Sweeps Sokoban & 2048, Lands Top-2 in Phoenix Wright & Candy Crush",
            "twitter_link": "https://x.com/haoailab/status/1915464349558460422"
        },
        {
            "date": "2025-04-15",
            "video_link": "https://www.youtube.com/watch?v=q8PMW870yp8",
            "twitter_text": "Ace Attorney AI Revolution: O1 & Gemini 2.5 Pro lead in courtroom reasoning, while GPT-4.1 matches older models. Cost analysis reveals Gemini 2.5 Pro's 6-15x efficiency over O1.",
            "twitter_link": "https://x.com/haoailab/status/1912231343372812508"
        },
        {
            "date": "2025-04-08",
            "video_link": "https://www.youtube.com/watch?v=yoEo2Bk7PGA",
            "twitter_text": "LLaMA 4 Maverick hacks traditional benchmarks but struggles with real gameplay—our transparent leaderboard exposes the new AI challenge.",
            "twitter_link": "https://x.com/haoailab/status/1909712259326394519"
        },
        {
            "date": "2025-04-01",
            "video_link": "https://www.youtube.com/watch?v=uFVpNor7l_E",
            "twitter_text": "Google's Gemini 2.5 Pro redefines AI gameplay: its multi-modal edge outperforms o1 & Claude 3.7 in Sokoban.",
            "twitter_link": "https://x.com/haoailab/status/1907140718650704204"
        },
        {
            "date": "2025-03-18",
            "video_link": "https://www.youtube.com/watch?v=b-Uyz3W4yIg",
            "twitter_text": "Candy Crush Saga's Hidden Complexity: Top AI Models Take the Challenge",
            "twitter_link": "https://x.com/haoailab/status/1902095369808601551"
        },
        {
            "date": "2025-03-14",
            "video_link": "https://www.youtube.com/watch?v=3aYDCSa3AWI",
            "twitter_text": "2048 Mastery: Only Two AI Models Crack the Code to Surpass Random Play",
            "twitter_link": "https://x.com/haoailab/status/1900645722095317255"
        },
        {
            "date": "2025-03-06",
            "video_link": "https://www.youtube.com/watch?v=59enV32MBUE",
            "twitter_text": "Sokoban Showdown: o3-mini Dominates by Reaching Level 4",
            "twitter_link": "https://x.com/haoailab/status/1897792946646421514"
        },
        {
            "date": "2025-02-28",
            "video_link": "https://www.youtube.com/watch?v=nixMIJZYAgg",
            "twitter_text": "Super Mario AI Revolution: Claude-3.7 Sets Unprecedented Gameplay Benchmarks",
            "twitter_link": "https://x.com/haoailab/status/1895557913621795076"
        }
    ]
}