jesswang.freedom

給朋友們一些回饋~ 透過我的連結註冊將拿到 20% 的現貨以及合約手續費佣！🔥🔥🔥 這很多了，想讓朋友們交易贏在起跑點，少一些手續費磨損♥️ Bitget 是一間野心很大，也很用心打造產品的交易所 馬上3/2 要出 HaIo IEO，並搭配超多優惠活動 partner.bitget.com/bg/puy586… #Bitget #Web3 #Crypto

A cool Taiwanese singer We love him 🙋🏼‍♂️ #別戀 youtube.com/shorts/BxtiXDqx7… 來自 @YouTube

我個人很佩服一種人 - 真正從底層爬上來的人 很多人瞧不起暴富的土豪 覺得他們很沒氣質只會穿金戴銀 但是真正認識了這些暴富土豪你才會知道 他們的來時路遠比你要你想像的艱難 有一些好的共同特質是我歸納出來並且想要效法的 當然壞的部分就請自行過濾😅 1. 不內耗 他們基本上不內耗只外耗 不會自己糾結 也不會過多的顧慮他人感受 一個糾結整天想東想西的人我覺得是很難可以真正成事的 他們一旦遇到了有可能會內耗的事情 就會直接拋出來換成外耗別人 然後就繼續專注在自己的事務 相信大家最近也看到了孫宇晨的新聞😂就是這種體現 2. 行動力 在AI 的時代，執行面的成本已經越來越被降低 而真正缺的就是可以真正落地執行的能力和決心 他們這種人有可能不是頂尖聰明 但是決心魄力野心非常人可及 一般人還在思考準備計劃 他們什麼都還沒有就先衝了 在這時代行動力遠遠比才智重要 通才也比專才有意義 差距會越來越大 AI 時代屬於有行動力以及通才的人 3. 洞悉漏洞 他們常常看事情視角與我們不同 同一件事情他們可以想出非常多不同的面相 博弈論是個好的方法訓練這種思維 舉例來說 拿考上台大這件事情 他們做的絕對不會是乖乖唸書😅 而是從結果倒推想盡各種手法進入台大 一旦可以進入了就會轉到別的事情繼續行動 不知道大家能否懂我在什麼 好的面向值得學習 可以爬上來的其實都絕非泛泛之輩 那些大陸土豪也是 他們炫耀也是正常的 共勉之

I recently came across a term that I think perfectly describes where AI is heading: Graph Engineering. If you break down how we work with AI, there are roughly four layers. 1. Prompt: Ask the right question This is the most basic layer. If you define the problem incorrectly, even the smartest AI will just give you a very sophisticated answer to the wrong question. But we’re slowly realizing that prompts are only the beginning. ⸻ 2. Context: Give it the right information A lot of the time, the model isn’t the problem. The problem is: You gave it the wrong data. 😂 Garbage in, garbage out. Even the best model in the world can only turn bad context into beautifully written garbage. So one of the most important abilities of an AI system is knowing: What information does this task actually need right now? Not simply dumping everything you have into the context window. ⸻ 3. Loop: Let AI handle complex tasks Real-world work is rarely completed in a single step. It usually looks more like: Analyze → Execute → Check → Find problems → Fix → Execute again. In other words: a Loop. This is where AI agents start becoming interesting. AI is no longer just “answering your question.” It can keep working until it reaches a desired outcome. ⸻ But once you go one step further, you run into another problem: Too much freedom isn’t always a good thing. Which brings us to the fourth layer. 4. Graph: Define the workflow If you’re doing creative writing, giving AI freedom makes sense. But if you’re doing things like: financial reconciliation, customer refunds, contract review, risk control, reporting, or code review… these tasks already have SOPs. What you really want isn’t: “AI, figure out what you should do next.” It’s more like: Validate the data → collect missing information → execute → verify the result → send exceptions for human review. Where can tasks branch? What can run in parallel? Where must something be checked? When should the system retry? You define these things in advance. That’s how I think about Graph Engineering. It’s not about restricting AI. It’s about putting AI’s intelligence inside a system that can produce stable, repeatable results. And that’s why I increasingly believe the most valuable people in the AI era won’t necessarily be the ones who are best at writing prompts. They’ll be the ones who can answer this question: “What is the correct process for getting this work done?” Prompts can be optimized by AI. Context can be retrieved automatically. Loops can be handled by agents. But if the workflow itself is a mess, even the smartest model will just help you create chaos more efficiently. 😂 The truly scarce skill in the AI era may not be simply knowing how to “use AI.” It may be the ability to take a vague human task and turn it into a clear, verifiable, repeatable workflow. That, to me, is what makes Graph Engineering so interesting. #AI #GraphEngineering #AIAgent #AgenticAI #PromptEngineering #ContextEngineering #LLM #Workflow #SoftwareEngineering #ArtificialIntelligence

好演講值得一聽

【我們眺望的星空】網頁活動進行中！參與活動，生成屬於自己的星星，幫助莫寧點亮星海，可獲取星聲等獎勵！starsalign-api.kurogames-glo… #鳴潮 #莫寧

跟 Gemini 一起做了一個夢幻字典 簡單記錄一下 成果非常滿意，完全符合我個人需求 1. 有吉祥物隨時給予情緒價值😂（有喜怒哀樂隨時眨眼陪伴學習） 2. 同時多語言支援（給想要學第二外語的絕佳工具，學習第二外語下同時複習第一外語）🌍 3. 全文本即時音頻（這邊不是用AI，就是用基礎的Web Speech API確保一定有聲音，聲音有配置到很自然溫暖的語速）💬 4. AI 即時聊天繼續探討（就跟基本的chatgpt一樣，但是有用prompt做設定繼續討論你要查的內容）🧠 5. 字或句生圖加深記憶（Gemini api) 6. 記事本記錄你有興趣儲存之後再看的單字或句子，並且可以用你儲存的單字生成小故事加深學習記憶📒 7. 設定可以隨時更改母語以及想要學的語言⚙️ 8. 非同步併發所有請求，階段性渲染畫面，效能高，一般的AI app效能是個很大的痛點😓 9. 比我之前用過的字典好用太多，一般的字典會列出太多不是主要meaning 的意思，太多相對無用資訊佔據腦容量