In this article, we explore the process of vibe coding, where coding is not just a solitary task but an engaging experience. 在這篇文章中，我們探討了 vibe coding 的過程，這是一種編碼不僅僅是孤獨的任務，而是一種引人入勝的體驗。 The idea is to narrate the coding journey in real-time, allowing readers to witness the development process as it unfolds. 這個想法是實時敘述編碼過程，讓讀者見證開發過程的展開。 This approach provides a unique perspective on coding, making it more relatable and accessible. 這種方法提供了對編碼的獨特視角，使其更具親和力和可及性。
One common challenge developers face is adjusting the font size in their code editor. 開發者面臨的一個常見挑戰是調整代碼編輯器中的字體大小。 In this case, the solution involves using keyboard shortcuts, specifically Control + Shift + Plus, to zoom in on the font size. 在這種情況下，解決方案是使用鍵盤快捷鍵，具體來說是 Control + Shift + Plus，來放大字體大小。 This simple adjustment can significantly enhance readability and improve the coding experience. 這個簡單的調整可以顯著提高可讀性並改善編碼體驗。
The focus of this coding session is on training an AI to play the classic game Tempest. 這次編碼會議的重點是訓練一個 AI 玩經典遊戲 Tempest。 The AI learns through trial and error, gradually improving its gameplay by analyzing its performance across various levels. AI 通過試錯學習，逐步通過分析其在各個關卡的表現來改善其遊戲玩法。 Currently, the AI can navigate through 33 levels, but it struggles with the faster-paced yellow levels, indicating a need for further refinement in its learning algorithm. 目前，AI 可以通過 33 個關卡，但在快速節奏的黃色關卡中表現不佳，這表明其學習算法需要進一步改進。
A crucial aspect of training the AI involves tweaking the reward function, which measures the AI's performance in each frame of the game. 訓練 AI 的一個關鍵方面是調整獎勵函數，該函數衡量 AI 在遊戲每一幀中的表現。 The reward function assigns points based on the AI's actions, such as successfully evading enemies or hitting targets. 獎勵函數根據 AI 的行動分配點數，例如成功躲避敵人或擊中目標。 Adjusting these parameters is essential for encouraging the desired behavior in the AI. 調整這些參數對於鼓勵 AI 的期望行為至關重要。
The AI extracts a wealth of information from the game state, including the current level, player lives, and enemy positions. AI 從遊戲狀態中提取大量信息，包括當前關卡、玩家生命和敵人位置。 This data is crucial for the AI to make informed decisions during gameplay. 這些數據對於 AI 在遊戲過程中做出明智的決策至關重要。 By simplifying the data it processes, the AI can learn more efficiently and focus on mastering the game's mechanics. 通過簡化其處理的數據，AI 可以更有效地學習並專注於掌握遊戲的機制。
As the coding progresses, strategic changes are made to the AI's behavior, such as penalizing it for using powerful abilities prematurely. 隨著編碼的進展，對 AI 的行為進行了戰略性變更，例如對其過早使用強大能力進行懲罰。 These adjustments aim to promote more cautious gameplay, ensuring the AI learns to prioritize survival over aggressive tactics. 這些調整旨在促進更謹慎的遊戲玩法，確保 AI 學會優先考慮生存而非攻擊戰術。 Each modification requires careful consideration and testing to gauge its effectiveness. 每次修改都需要仔細考慮和測試，以評估其有效性。
Testing the AI's performance presents its own set of challenges. 測試 AI 的性能帶來了一系列挑戰。 Running extensive simulations to gather statistically valid results can be time-consuming, often taking hours or even days. 進行廣泛的模擬以獲取統計有效的結果可能會耗時，通常需要幾個小時甚至幾天。 This necessitates a careful balance between making multiple changes and isolating variables to understand their impact on the AI's learning process. 這需要在進行多次更改和隔離變量之間取得仔細的平衡，以了解它們對 AI 學習過程的影響。
The training process is supported by powerful hardware, specifically a Thread Ripper machine with 96 cores and 192 threads. 訓練過程得到了強大硬體的支持，具體來說是一台擁有 96 顆核心和 192 條線程的 Thread Ripper 機器。 This high-performance setup allows for the simultaneous execution of multiple game instances, significantly speeding up the training process. 這種高性能的設置允許同時執行多個遊戲實例，顯著加快訓練過程。 The efficiency of the hardware plays a vital role in the AI's ability to learn and adapt quickly. 硬體的效率在 AI 快速學習和適應的能力中起著至關重要的作用。
As the coding session wraps up, the AI is set to run with the newly implemented reward functions. 隨著編碼會議的結束，AI 將運行新實施的獎勵函數。 The goal is to observe its performance and make further adjustments as necessary. 目標是觀察其性能並根據需要進行進一步調整。 This iterative process of coding, testing, and refining is essential for developing a robust AI capable of mastering the complexities of Tempest. 這種編碼、測試和改進的迭代過程對於開發一個能夠掌握 Tempest 複雜性的強大 AI 至關重要。
Q: What is vibe coding?
問：什麼是 vibe coding？
A: Vibe coding is an engaging approach to coding where the development process is narrated in real-time, allowing readers to witness the coding journey as it unfolds.
答：Vibe coding 是一種引人入勝的編碼方法，開發過程以實時方式敘述，讓讀者見證編碼過程的展開。
Q: How can I adjust the font size in my code editor?
問：我如何調整代碼編輯器中的字體大小？
A: You can adjust the font size in your code editor by using the keyboard shortcut Control + Shift + Plus to zoom in.
答：您可以通過使用鍵盤快捷鍵 Control + Shift + Plus 來放大代碼編輯器中的字體大小。
Q: What is the focus of the AI training session discussed?
問：討論的 AI 訓練會議的重點是什麼？
A: The focus is on training an AI to play the classic game Tempest, where it learns through trial and error and analyzes its performance across various levels.
答：重點是訓練一個 AI 玩經典遊戲 Tempest，通過試錯學習並分析其在各個關卡的表現。
Q: What is the purpose of the AI's reward function?
問：AI 的獎勵函數的目的是什么？
A: The reward function measures the AI's performance in each frame of the game, assigning points based on actions like evading enemies or hitting targets to encourage desired behavior.
答：獎勵函數衡量 AI 在遊戲每一幀中的表現，根據躲避敵人或擊中目標等行動分配點數，以鼓勵期望的行為。
Q: What kind of data does the AI extract from the game state?
問：AI 從遊戲狀態中提取什麼樣的數據？
A: The AI extracts information such as the current level, player lives, and enemy positions, which are crucial for making informed decisions during gameplay.
答：AI 提取的信息包括當前關卡、玩家生命和敵人位置，這對於在遊戲過程中做出明智的決策至關重要。
Q: What strategic changes are made to the AI's behavior?
問：對 AI 的行為進行了什麼戰略變更？
A: Strategic changes include penalizing the AI for using powerful abilities prematurely to promote cautious gameplay and prioritize survival over aggression.
答：戰略變更包括對 AI 過早使用強大能力進行懲罰，以促進謹慎的遊戲玩法並優先考慮生存而非攻擊。
Q: What challenges are faced in testing the AI's performance?
問：在測試 AI 性能時面臨什麼挑戰？
A: Testing the AI's performance can be time-consuming, requiring extensive simulations to gather statistically valid results, which necessitates balancing multiple changes and isolating variables.
答：測試 AI 的性能可能會耗時，需要進行廣泛的模擬以獲取統計有效的結果，這需要在多次更改和隔離變量之間取得平衡。
Q: What hardware is used for training the AI?
問：用於訓練 AI 的硬體是什麼？
A: The training process is supported by a Thread Ripper machine with 96 cores and 192 threads, allowing for the simultaneous execution of multiple game instances.
答：訓練過程得到了擁有 96 顆核心和 192 條線程的 Thread Ripper 機器的支持，允許同時執行多個遊戲實例。
Q: What are the next steps after the coding session?
問：編碼會議後的下一步是什麼？
A: The next steps involve running the AI with the newly implemented reward functions and observing its performance to make further adjustments as necessary.
答：下一步是運行 AI，使用新實施的獎勵函數，並觀察其性能以根據需要進行進一步調整。
