“강의 안 팝니다. 그냥 다 알려드릴게요” AI 공부, 딱 3일만 이 순서대로 해보세요

Main ideas / lessons

• AI “FOMO” is common, especially among people in their 20s–30s. Studying/using AI regularly can reduce anxiety and improve confidence.
• You don’t need paid courses or a deep theory-first path to start using AI effectively.
• Treat AI like a smartphone: learn enough to use it, then look up features only when you need them.
• The key to learning AI is hands-on iteration: generate outputs, review them, refine prompts/workflows, and gradually build reusable “skills.”
• Structure learning into a 3-day plan (Chuseok holiday) with increasing real-world usefulness:
  • Day 1: get familiar with AI functions
  • Day 2: delegate a recurring real work task
  • Day 3: build a small new project and deploy/share it

---

Study methodology / instructions (detailed)

Before the 3-day plan: correct common misconceptions

1. Misconception 1: “To use AI, you must study theory first.”

   • Theory is required mainly for people who want to research/develop AI.
   • Most users should focus on using existing AI tools to improve work, generate materials, and try tasks without mastering all underlying principles.

2. Misconception 2: “To use AI you must subscribe to many sites/tools.”

   • Start with only two:
     • ChatGPT (referred to as “ChiPT” / “Choi Ji-T”)
     • Claude (also mentioned in the video)
   • Use one first to learn basics, then add specialized services later as needed (images/videos/music, etc.).

3. Misconception 3: “AI is too difficult; take paid courses or learn in school.”

   • Use the smartphone analogy: you don’t need every feature—you learn by doing.
   • Pay for/sub to one course/tool (e.g., ChatGPT or Claude) as “lecture-like” cost, then learn through practice instead of watching many lectures.

---

3-day AI holiday plan (3 hours/day, exactly 3 days)

Day 1 — Learn AI by trying many capabilities (5 steps)

1. Install and subscribe to AI

   • Install the app on your computer (not just a website).
   • The speaker uses both ChatGPT and Claude because sometimes one performs better.
   • Suggested beginner pick: ChatGPT (UI preference mentioned).

2. Do deeper research tasks (not just simple Q&A)

   • Instead of asking for one-line facts, ask for source-finding, comparison, and organization.
   • Example task format:
     • “Plan a US trip for a week in September.”
     • Include comparisons like flight time, weather, travel difficulty, estimated cost, and provide sources.

3. Turn research into rich outputs (images/files)

   • Generate images based on research (example: “a Ghibli-style photo”).
   • Create shareable artifacts tied to research:
     • a map course image
     • an Excel/PowerPoint itinerary
     • a responsive website with an embedded map

4. Upload your own files for analysis

   • Put files into ChatGPT (examples mentioned: presentation slides, ledgers, product manuals).
   • Request multi-file summaries and identify sections you must check.
   • Core idea: file input makes AI feel more like a colleague, not just a search tool.

5. Use “Computer Use” to let AI operate your computer

   • Grant permission for the AI tool to control your computer (similar feature exists in Claude too).
   • Example instruction:
     • “Look at the internet window currently open on my computer; scrape all content on this page.”
   • Don’t try to learn everything in advance:
     • If you don’t know how to do something, ask AI how to do it, then try—AI can often set up permissions and execute.
   • Practice principle: when stuck, don’t search first—ask AI to attempt, then collaborate to fix issues.

Day 1 mission completion criteria: after 1–3 hours, you should feel you’ve actually used AI by completing the five steps once.

---

Day 2 — Delegate one real recurring work task (5 steps)

1. Pick one frequent, time-consuming task

   • Examples: collecting/organizing materials, meeting prep, daily writing/content creation, organizing study resources.

2. Explain your exact workflow to AI

   • Provide details:
     • what materials you look for
     • what criteria you use to judge/select
     • the order/sequence of your steps
     • what the final outcome should be
   • Lesson: detailed explanation is enough; “prompt skill” isn’t mysterious.

3. Refine iteratively with feedback

   • Expect mistakes or unwanted items.
   • Don’t give up—keep providing feedback and requests until the output format is usable.
   • Example refinement logic (industry news):
     • Exclude product promotions/ads.
     • Include only new models/features or items that affect your work methods.
     • Summarize key points with links.
     • Mark features you should try yourself separately.
     • Keep the same organization/format going forward.

4. Convert the workflow into a reusable “skill”

   • Create a saved process so you don’t re-explain from scratch each time.
   • Include:
     • what data to check
     • what content to include
     • how to structure the results
   • Ask AI:
     • “Turn this workflow into a skill so we can use it the same way next time.”

5. Test the skill with new data

   • Add new inputs or a different time period to verify:
     • it finds what you want
     • it removes unnecessary content
     • it outputs results correctly
   • Fix issues by instructing AI to correct them.
   • Optional upgrade: automate scheduled execution (e.g., every Monday 8 AM) so the task runs automatically.

Day 2 goal: your AI workflow becomes something you can delegate continuously with minimal manual effort.

---

Day 3 — Build something you couldn’t do before, then deploy/share (5 steps)

1. Choose a postponed “make it” idea

   • Example: a simple website to record daily study time, show a weekly bar chart, and show time remaining until a target.

2. Specify the desired look and requirements

   • Describe:
     • input fields (date + study time)
     • visualization (bar graph for a week)
     • mobile/desktop usability
   • Share reference links/images if available.

3. Delegate creation end-to-end to AI (including coding & file setup)

   • Ask AI to build the project:
     • create folders/files
     • write code
     • handle everything until the UI opens
   • If errors occur:
     • use AI’s error reporting and ask it to find the cause and fix it.
   • Coding lesson: even developers can rely on AI for code—focus on what result you want and give good instructions/feedback.

4. Use and iterate like a real user

   • Manually input data to verify:
     • graph correctness
     • whether design/layout feels right
     • what needs improvement on mobile (e.g., button size, aesthetics)
   • Provide thorough feedback because AI can’t perfectly infer preferences immediately.

5. Deploy and share

   • If possible, ask AI to distribute the site so it works via a link.
   • Open it on your phone and send the link to a friend.

Day 3 success condition: you can share a working result publicly (via link) and have practiced the full cycle: build → test → improve → deploy.

---

Extending beyond the 3 days: connect AI to everyday tools

• Use MCP (Model Context Protocol) to connect AI to tools such as:
  • Notion
  • Google Drive
  • Gmail
  • and other services via plugins
• What it enables:
  • AI can directly read/edit/work with data inside existing services.

• Example requests:

  • “Find last month’s meeting materials in my Google Drive.”
  • “Review content ideas in Notion and choose a topic.”
  • “Check in GitHub/Gitterb where the error occurred.”

• Combine MCP with Day 2/Day 3 “skills”:

  • Example ideas:
    • weekly industry-news skill → send via email
    • website design improvements via Figma
  • Mentioned workflow example: connecting a Figma plugin to AI to generate card news.

---

Final recommendations after 3–4 days

• Keep AI “on by default” in your daily workflow—like natural delegation, not memorizing prompt tricks.
• Subscribe to AI-related content (magazines/news/posts) to stay updated on new capabilities.
• The speaker’s main point: reducing FOMO comes from repeatedly using AI and noticing what’s newly possible.

---

Speakers / sources featured

Speaker (primary)

• Unnamed narrator/creator
  • Refers to themselves as a developer
  • Mentions receiving planned travel help from “Teacher Chaechi”

Tools / platforms mentioned

• ChatGPT (“Choi Ji-T”, “Chappy/Chappy’s UI”, “PT” as shorthand in subtitles)
• Claude
• Coursera (as an example platform for theory courses)
• MCP (Model Context Protocol)
• Notion
• Google Drive
• Gmail
• Figma
• Excel
• PowerPoint
• GitHub / “Gitterb” (referenced in subtitles; likely GitHub)