EW AI Coding

Turn an idea, SOP, workflow, or repetitive task into an application that can be built, tested, accepted, delivered, and verified for real value.

EW AI Coding is a free bilingual skill that helps anyone move beyond AI chat. Users can start from an everyday-life idea, learning goal, work problem, SOP, workflow diagram, or repetitive task. EW AI Coding first helps identify what is worth turning into an application, then guides the user one question at a time to create five consistent files ready for Codex, establish recoverable GitHub history, and continue after coding through automated tests, user preview, acceptance fixes, explicitly authorized release, and evidence-based value verification. It can be used for report automation, study assistants, AI travel planners, personal organizers, team workflows, and enterprise applications.

AI 應用專案鍛造工具

EW AI Coding 是免費的中英雙語技能,目的不是讓使用者停留在與 AI 聊天,而是把想法真正推進成可執行的應用。使用者可以從生活構想、學習需求、工作問題、SOP、流程圖或每天重複做的事情開始;EW AI Coding 會先協助判斷哪些工作值得做成應用,再透過一次一題的自然對話整理需求,產生五份可交付 Codex 的一致規格,建立可恢復的 GitHub 開發紀錄,並在 Codex 完成編程後繼續引導自動測試、使用者預覽、驗收修正、明確授權發布與價值驗證。

Install in ChatGPT / 在 ChatGPT 安裝

Office helped people create documents, spreadsheets, and presentations faster. AI Coding addresses the next problem: people should not have to repeat the same rule-based work every day.

AI Coding does not mean everyone must become a software engineer. It means identifying repetitive or time-consuming work, explaining the workflow and decision criteria in natural language, and working with AI to turn that knowledge into a reusable digital tool. People define the goal, rules, exceptions, and acceptance criteria; AI helps with implementation and repeated execution.

Office 時代解決的是「更快完成文件、表格與簡報」;AI Coding 時代要解決的是「不必每天重複做同樣的事」。

AI Coding 不是要求每個人都成為軟體工程師,而是讓人找出工作、學習與生活中重複耗時的部分,用自然語言說清楚目標、流程、資料、規則、例外與驗收方式,再與 AI 共同把經驗轉化成可以重複執行的數位工具。

過去是人去適應軟體;現在是使用者描述真實需求,AI 協助建立適合的工具。Office 讓人具備數位作業能力;AI Coding 讓人具備工具創造與工作自動化能力。

Here, project forging does not mean deploying code to a server. It means turning the real need, users, workflow, inputs, outputs, responsibilities, boundaries, and acceptance criteria into a reliable implementation blueprint before development starts.

這裡的「專案鍛造」不是把程式部署到伺服器,而是在開發前先把真實需求、使用者、流程、輸入輸出、責任、系統邊界與驗收標準整理成可執行的開發藍圖。

• Anyone with a useful idea but no software background.
• Learners building a study assistant, practice tool, or knowledge organizer.
• Individuals building travel, planning, household, or personal productivity tools.
• Employees who want to automate recurring workplace tasks.
• People with an SOP, procedure, workflow diagram, or repeated task who want to know whether it is worth turning into an application.
• Department AI application champions who want to build useful internal tools with Codex.
• Teams turning interviews, meeting notes, and operating experience into an implementation-ready blueprint.
• Builders reviewing whether an existing specification is consistent, safe, and testable.
• 有實用構想、但沒有軟體背景的一般使用者。
• 想建立學習助手、練習工具或知識整理工具的學習者。
• 想建立旅行規劃、生活管理或個人生產力工具的人。
• 想把重複工作自動化的企業員工。
• 已有 SOP、程序書、流程圖或重複工作,希望先判斷是否值得做成應用的人。
• 希望用 Codex 建立部門工具的 AI 應用種子人員。
• 要把訪談、會議紀錄與工作經驗轉成可施工藍圖的團隊。
• 需要檢查既有規格是否一致、安全、可驗收的開發者。

1. Describe the idea or workflow — Start from an idea, SOP, workflow, or repetitive task in ordinary language.
2. Find the worthwhile application — When workflow evidence exists, identify suitable, prerequisite-dependent, human-assist-only, and unsuitable automation opportunities.
3. Work it out together — Clarify inputs, outputs, rules, exceptions, users, responsibility, safety, and delivery target one question at a time.
4. Confirm the first version — Approve one small, complete, useful application blueprint.
5. Create the files — Generate one consistent Codex handoff package.

五段自然規劃對話:說出想法或提供流程、找出值得做的應用、一次一題想清楚、確認第一版、產生文件。使用者不必先學會 PRD、架構、API 或程式語言。

After the five-file handoff, invoke EW AI Coding again with the Codex result, project folder, ZIP, repository, preview, test output, or blocker. It identifies the first incomplete stage and continues without restarting the interview:

Idea/workflow → opportunity analysis when relevant → five specifications → GitHub record → Codex development → automated tests → user preview → acceptance fixes → authorized release → value verification

五份文件交接後,使用者可帶著 Codex 結果、專案資料夾、ZIP、GitHub 專案、預覽、測試輸出或阻塞狀況再次叫用 EW AI Coding。系統會從第一個未完成階段繼續,不必重新訪談:

想法/流程 → 必要時進行應用機會分析 → 五份規格 → GitHub 建檔 → Codex 開發 → 自動測試 → 使用者預覽 → 驗收修正 → 授權發布 → 價值驗證

EW AI Coding does not guess how long coding will take. It manages evidence-based states: what is complete, what failed, what is blocked, and what action comes next. Publication remains separate from development and always requires explicit authorization.

EW AI Coding 不猜測 Coding 需要多久,而是管理有證據的狀態:完成了什麼、哪裡失敗、受到什麼阻塞、下一步做什麼。開發與發布分開;公開或正式上線前一定重新取得明確授權。

EW AI Coding supports personal, learning, life, team, and enterprise tools. It increases governance only when the real use requires it:

企業可自行建立一般工具,但涉及敏感資料、跨部門、財務、安全、法律、醫療或設備控制時,必須升級治理與專業審查。AI Coding 能降低開發門檻,不會取消工程責任。

• Codex and deterministic software: interfaces, workflows, databases, permissions, reports, sorting, and state transitions.
• Model specialists: computer vision, time-series detection, retrieval, multimodal models, fine-tuning, evaluation, and inference optimization.
• Domain specialists: CAD geometry, manufacturing, energy, finance, healthcare, or other professional rules.
• Data specialists and infrastructure: data quality, lineage, replay, integration, and reliable real-world data sources.
• Humans: approval of high-risk, irreversible, or professionally accountable decisions.

EW AI Coding 的目的不是把所有需求交給單一軟體商,而是先讓個人或組織掌握問題、規格與驗收,再精準判斷真正需要哪一類專家。

• English and Traditional Chinese are fully supported.
• The skill follows the language of the latest user request.
• The interview, blueprint, and five files remain in one selected language.
• Bilingual deliverables are created only when explicitly requested.
• 完整支援英文與繁體中文,並依使用者最新訊息選擇語言。
• 訪談、藍圖與五份文件保持同一語言;只有明確要求時才產生雙語文件。

• I want to build an app that... / 我想開發一個能夠......的應用。
• 請幫我分析這份 SOP/流程,看看哪些工作適合做成應用。
• I want to improve this workflow or repeated task... / 我想改善這個流程或重複工作......
• Codex has finished coding. What should I do next? / Codex 已完成開發,下一步怎麼做?

• Simple for employees; rigorous for implementation.
• Define the problem before choosing a model or technical stack.
• Analyze a supplied SOP or workflow before assuming the whole process should be automated.
• No coding before the first-version blueprint is approved.
• Ask one main question per turn.
• Do not invent data, laws, business rules, users, departments, or environments.
• Build one complete useful workflow before expanding scope.
• Keep humans in control of high-risk or irreversible actions.
• Map every first-version module to architecture ownership and acceptance evidence.
• Preserve user and organizational ownership of code, data, and knowledge.

• Student installation / 學員安裝
• Educator publishing guide / 教師發佈指南
• Support / 支援
• Privacy Policy / 隱私權政策
• Terms of Service / 服務條款

具象職人:讓經驗具象,讓知識傳承,讓 AI 成為員工。

我們協助企業將員工經驗、專業判斷與工作流程,轉化為可保存、可複製、可執行的數位知識與應用;進一步串聯企業流程、真實資料與工業現場,打造能理解任務、執行工作並持續累積能力的 AI 數位員工,讓個人經驗成為組織資產,讓企業智慧得以規模化傳承。

Embodied Worker: Make experience tangible. Preserve knowledge. Make AI an employee.

We help enterprises transform employee experience, professional judgment, and workflows into digital knowledge and applications that can be preserved, replicated, and executed. By connecting business processes, real-world data, and industrial sites, we build AI digital employees that can understand tasks, perform work, and continuously accumulate capability—turning individual experience into organizational assets and enabling enterprise intelligence to be transferred at scale.

The free EW AI Coding journey remains self-service through specifications and guided continuation. Organizations that need consulting guidance—or projects involving machines, sensors, Edge AI, industrial data acquisition, production equipment, energy systems, or on-site integration—may receive an optional next-step notice without blocking the free workflow.

For enterprise application development guidance or physical-world and industrial implementation, contact Embodied Worker Co., Ltd. / 具象職人股份有限公司 at pohsun@embodiedworker.com.

免費的 EW AI Coding 流程會完整交付五份文件並提供後續引導,不要求登記、付費或申請顧問服務。企業若需要顧問進一步指導應用開發,或專案涉及設備、感測器、Edge AI、工業資料採集、生產設備、能源系統與現場整合,才會顯示可選擇的延伸提示,不阻斷免費流程。

Before generating the five files, EW AI Coding confirms where the finished application will run and what the user must receive: a hosted Web URL, Windows installer or portable app, macOS app, Ubuntu/Linux package or service, Android phone/tablet app, iPhone/iPad app, multi-platform build, container, or source code. Packaging, signing, installation, update, and uninstall requirements are carried into architecture and acceptance.

在產生五份文件前,EW AI Coding 會確認應用最終在哪裡使用,以及要交付網址、Windows 安裝包或可攜版、Mac App、Ubuntu/Linux 套件或服務、Android 手機/平板 App、iPhone/iPad App、多平台版本、容器或原始碼。打包、簽章、安裝、更新與卸載要求會同步寫入架構與驗收文件。

v0.8.1 makes all three ChatGPT conversation starters bilingual in English and Traditional Chinese while keeping each complete card within the 128-character platform limit. The routes follow project maturity—an application idea, a workflow or repeated task, and post-coding continuation—so the first two cards elicit the user's intended application or problem instead of merely classifying identity. Selecting 1, 2, or 3 activates the corresponding guided Skill flow.

v0.8.1 將三張 ChatGPT 提示卡全部改為英文與繁體中文並列,且每張完整雙語文案均不超過平台 128 字元限制;三個入口依專案成熟度分為「應用想法」、「流程或重複工作」與「Coding 完工後續作」。前兩張卡直接引導使用者說出想開發的應用或想改善的問題,而不是只分類使用者身分;選擇 1、2、3 後由 Skill 進入相應引導流程。

v0.8.0 expands EW AI Coding from idea-to-build guidance into application opportunity discovery. Users can now start from an idea, SOP, workflow diagram, procedure, interview notes, or repetitive task. When workflow evidence exists, EW AI Coding first identifies suitable automation opportunities, tasks that need prerequisites, AI-assist-only work, and work that should remain manual. It then guides the selected opportunity one question at a time into the same five Codex-ready specifications and continues through GitHub, Codex development, automated testing, preview, acceptance, explicit release authorization, and value verification. Personal and everyday-life applications remain first-class; enterprise governance appears only when the actual scope requires it. The public brand remains EW AI Coding.

v0.8.0 將 EW AI Coding 從「把想法整理成可開發規格」進一步升級為「先發現值得做的應用,再完成開發閉環」。使用者可以從想法、SOP、流程圖、程序書、訪談紀錄或重複工作開始;當有流程資料時,系統會先區分「適合現在做」、「補足條件後適合」、「只適合 AI 輔助」與「不適合自動化」的工作,再由使用者選擇值得推進的項目,透過一次一題的引導產生五份 Codex 規格,並接續 GitHub、Codex 開發、自動測試、預覽、驗收、明確授權發布與價值驗證。一般民眾與企業共用同一核心流程,只有實際需要時才提高企業治理強度;對外品牌持續使用 EW AI Coding。

v0.7.0 extends EW AI Coding beyond the five-file handoff. Users can return after Codex work and continue from repository recovery through automated testing, preview, acceptance fixes, explicit release authorization, and value verification. It adds evidence-based lifecycle states, PROJECT_STATUS.md, optional VALUE_REPORT.md, GitHub privacy defaults, and a strict separation between development completion and production release.

v0.7.0 將 EW AI Coding 從五份文件交接延伸為完整閉環。使用者可在 Codex 編程後再次叫用,從專案恢復、自動測試、預覽、驗收修正,繼續到明確授權發布與價值驗證;新增以證據判斷的專案狀態、PROJECT_STATUS.md、可選的 VALUE_REPORT.md、GitHub 隱私預設,以及開發完成與正式發布分離機制。

v0.6.0 adds mandatory delivery-target discovery. The interview now distinguishes Web, Windows, macOS, Ubuntu/Linux, Android, iPhone/iPad, and multi-platform delivery; records the expected package; and adds build, signing, installation, update, rollback, and uninstall requirements to the five-file handoff.

v0.6.0 新增必要的交付平台訪談,區分 Web、Windows、macOS、Ubuntu/Linux、Android、iPhone/iPad 與多平台成品;明確記錄交付包,並把建置、簽章、安裝、更新、回滾與卸載要求寫入五份文件。

v0.5.0 expands EW AI Coding from an enterprise-first tool into a general AI application forge for work, learning, and everyday life. It adds a clear AI Coding capability statement, new personal and learning use cases, an AI travel planner example, and proportional governance that appears only when the project requires it.

v0.5.0 將 EW AI Coding 從企業優先定位擴展為適用於工作、學習與生活的通用 AI 應用鍛造工具;新增 AI Coding 能力主張、個人與學習場景、AI 旅行助手範例,並改為只有在專案需要時才提高治理強度。

v0.4.0 adds an optional, relevance-based path from a completed software blueprint to enterprise application consulting or physical-world and industrial implementation support. The notice appears only after the complete free five-file handoff and does not collect user information inside the skill.

v0.4.0 新增情境式引流能力:完整免費交付五份文件後,企業若需要顧問進一步指導應用開發,或專案涉及設備、感測器、Edge AI、工業資料採集與現場整合,才會顯示聯繫具象職人股份有限公司的可選提示;技能內不收集使用者資料。

v0.3.4 corrects the public Directory and ChatGPT composer icons so the new gold elephant brand asset is shown consistently. Functional behavior is unchanged from v0.3.3.

v0.3.4 修正外掛目錄與 ChatGPT 輸入框圖示,統一顯示新的金色大象商標;功能內容與 v0.3.3 相同。

v0.3.3 establishes EW AI Coding as an Enterprise AI Application Pre-Deployment Forge, makes English the primary product language with Traditional Chinese support, adds enterprise application risk routing, and clarifies the boundary between AI Coding, model specialists, domain experts, data infrastructure, and human approval.

v0.3.3 將 EW AI Coding 正式定位為「企業 AI 應用開發前置部署鍛造工具」,以英文為主、繁體中文為輔,新增企業應用風險分流,並明確區分 AI Coding、模型專家、領域專家、資料基礎設施與人工核准的責任。

v0.3.0 simplified the visible interview into four natural conversational stages and generated the five files as one consistent package.

v0.2.0 added full Traditional Chinese and English interaction and language-aware deliverables.