[Developer Notes] The Development Log of CANSLIM.BLOG (CANSLIM Research)

I want to share the latest development progress of CANSLIM.BLOG (CANSLIM Research). It started by simply using AI to automatically post my trading homework to Discord. Then, it became a website, and recently, we successfully developed AI models that simulate “classic traders.” From idea to reality, it only took two weeks of full-time work. During this time, the most common question I get is: “What exactly is CANSLIM.BLOG?”

Building a 90% Automated AI Research Platform

Simply put, it is a financial website run by Artificial Intelligence. Currently, about 90% of the process is automated. The remaining parts need manual adjustment mostly because external data connections frequently change.

You might not know this, but the data the website collects is not just regular public information. Most of it is expensive paid data (costing tens of thousands of dollars a year). The AI processes this massive amount of data and presents it to you in the easiest way possible. My ultimate goal is to build a high-quality media platform run entirely by AI, without needing to hire reporters or editors. I believe this will save huge costs and is the unavoidable future of the media industry.

The Logic Behind Our 5 AI Traders

Right now, we have successfully built AI models of 5 of the most famous and successful traders from the past half-century. Because of high server costs, they cannot run 24/7 yet. Therefore, we schedule their trades during two golden time slots: 1 hour after the market opens and 30 minutes before the market closes. This helps avoid the market noise caused by retail investors and accurately follows the trading rhythm of big institutional investors.

Every day, these 5 AI traders strictly follow these steps:

• Data Scraping & Analysis: They scan the whole market to find target stocks that fit their specific trading style.
• Strategy Planning: They sort the stocks into three categories: “Ready to Trade,” “Keep Watching,” and “Avoid.”
• Public Decisions: Their entire thinking process is made public. This allows readers to see how top-tier AI models make decisions, serving as a great educational tool.

Overcoming Human Weakness: Strict Discipline and Risk Control

Making long-term profits in the stock market is basically a “math game.” You must strictly cut your losses and let your profits run to grow your capital. However, the win rate for these breakout trading strategies is often only 10% to 30%. This means out of 100 trades, you have to endure 70 to 90 losses. This pattern goes completely against “human nature,” and most regular investors simply cannot handle it. But giving this job to an AI—which has no emotions and only calculates win rates and follows discipline—makes it much more efficient.

To prevent the AI from overtrading, I specifically added a “Market Breadth” algorithm. It’s like fishing: if there is a massive storm outside, going out to sea will only get you hurt. When market conditions turn bad, the AI knows to switch to “Risk-Off” mode and hold cash. The market has been bad over the past month, so some AI models might look like they are losing money on paper. However, they are actually doing better than the overall market (when an index drops 3%, highly volatile stocks can crash 20% to 40%). Even on days when the AI decides not to trade, it still shares its Watchlist. Explaining the market through calculated data is a great learning experience in itself.

Inviting You to Beta Test and Future Plans

The algorithms are still being improved, and we will review the results in 3-month cycles. If my budget allows in the future, I plan to upgrade some models to run full-time during trading hours, which will take things to the next level. Because this project started as a personal study tool, CANSLIM.BLOG is currently completely free. If you register, I will unlock all premium features for you in the backend. (Note: Due to WordPress system rules, you must click the confirmation link in your email after registering here (Free Plan) Before I can upgrade your account manually.) I hope you will try it out and give me feedback. If you believe in this idea and want to sponsor CANSLIM Research, it doesn’t have to be money. Providing computing power, or simply leaving a short recommendation on LinkedIn with your profile picture, would mean the world to me.

Finally, thank you all for your continued support!

Andrew Wong

📰 【開發者手記】CANSLIM.BLOG 交易大模型開發日誌

還是想跟大家匯報一下 CANSLIM.BLOG (CANSLIM Reasearch) 的最新開發進展。從最初單純利用 AI 將個人交易功課自動發佈到 Discord，到建立網站分享，乃至近期成功研發出模擬「經典交易者」的 AI 大模型。從構思到落實，全職投入開發其實只花了短短兩週。這段期間最常被問到的問題是：「CANSLIM.BLOG 究竟是什麼？」

打造 90% 自動化的 AI 數據研究平台

簡單來說，這是一個由人工智慧自動營運的財經網站。目前大約 90% 的流程已實現自動化，剩下需人工微調的部分，主要是因應外部數據接口的變動。大家可能不知道，網站背後爬取的並非一般公開資訊，絕大部分是價值不菲的付費數據（每年訂閱費高達數萬元）。AI 處理這些龐大數據後，會以最淺顯易懂的方式呈現給大家。我的最終願景，是建構一個無須僱用記者與編輯、由 AI 全自動產出高質量分析的媒體平台。我相信，這不僅能大幅節省成本，更是未來媒體業的必然趨勢。

五大 AI 交易員的運作邏輯

目前，我們已成功將 5 位過去半世紀最頂尖、策略具可複製性的著名交易員，建構成 AI 大模型。受限於伺服器成本，它們暫時無法 24 小時全天候運作，因此我們將交易指令集中在兩個黃金時段：開盤後 1 小時與收盤前 30 分鐘。這能有效避開散戶造成的市場雜音，並精準跟隨機構投資者在尾盤出手的節奏。

每天，這 5 位 AI 交易員會嚴格執行以下流程：

• 數據爬取與分析： 掃描全市場，尋找符合自身風格的目標股票。
• 沙盤推演： 分類出「可執行」、「需持續觀察」與「未達標須避開」的標的。
• 公開決策： 將思考過程完全公開，讓讀者觀摩頂級大模型的決策邏輯，發揮投資教育的價值。

克服人性弱點：嚴守紀律與風險控管

要在股市長期獲利，本質上是一場「數學遊戲」：嚴格限制虧損，讓利潤奔跑，進而使資金規模不斷擴大。然而，這類突破性交易策略的勝率往往只有 10%－30%。這意味著在 100 次交易中，你必須忍受 70 到 90 次的虧損。這種模式極度「違反人性」，多數散戶根本無法承受。但交由沒有人類情感、只精算勝率與紀律的 AI 來操作，反而更加高效。此外，為了避免 AI 過度交易（Overtrade），我特別在模型中加入了「市寬（Market Breadth）算法」。就像大市掛起十號風球時，強行出海釣魚必然焦頭爛額；當市寬變差時，AI 會自動判定為 Risk-off 並增持現金。過去一個月大市轉差，部分 AI 模型帳面上或許處於虧損，但它們的表現其實已優於大盤（當指數下跌 3% 時，高 Beta 股票往往暴跌 20%－40%）。即使在不交易的日子，AI 依然會分享它的觀察名單（Watchlist），這種透過數據視覺化解釋市場的方式，本身就是極佳的學習。

誠邀公測與未來展望

目前算法仍持續優化中，我們將以 3 個月為一期來檢視成果。若未來預算允許，我計畫將部分模型升級為全交易時段的 Agent，邁向更高的境界。這個專案源於學術研究與個人功課整理，因此 CANSLIM.BLOG 目前完全免費。只要您註冊，我就會在後台為您解鎖所有進階功能（註：因 WordPress 系統限制，註冊後請務必至信箱點擊確認信，我才能手動為您升級）。期待大家親自體驗並給予反饋。如果您認同這個理念，並願意贊助 CANSLIM Research，無論是提供算力支援，或是在 LinkedIn 留下一句推薦證言與頭像，對我來說都是莫大的鼓勵。

最後，感謝大家持續的支持！

Andrew Wong