GoodTechAI

How to Build Resilience: The 3 Elements of Positive Psychology and a 3-Week Action Plan for Professionals

GoodTechAdviser — Sun, 12 Apr 2026 09:59:46 +0900

✍️ GoodTech AI ⏱️ 5 min read

Table of Contents

1. What Is Resilience?
2. Self-Regulation: The Art of Handling Emotions
3. Interpersonal Skills: The Power of Having One Person in Your Corner
4. Positivity: How to Train Optimistic Thinking
5. 3-Week Action Plan
6. Frequently Asked Questions

Are you looking for what resilience really means — and how to actually build it?

Research shows that two out of every three working professionals experience burnout (Maeil Shinmun, 2024). This article breaks down resilience into three core elements drawn from positive psychology, and provides a concrete 3-week action plan you can start today.

What Is Resilience?

Resilience is the capacity to bounce back to a healthy psychological state after experiencing stress or adversity. Psychologists call it the muscle of the mind.

Key fact: 30 to 50% of resilience is genetic — but the remaining 50 to 70% can be developed through deliberate practice. It is a learned competency, not a fixed trait.

Professor Kim Joo-hwan, who developed the Korean Resilience Quotient (KRQ-53) through 20 years of research, identified three core components: self-regulation, interpersonal skills, and positivity.

Self-Regulation: It Is Not About Suppressing Emotions — It Is About Handling Them

The core of self-regulation is recognizing emotions and releasing them in healthy ways.

Awareness Comes First

Think of the three emotions you feel most at work. If you are drawing a blank, that itself is a signal — awareness has already broken down. We can only control what we are aware of.

Watch Out for Displacement

The most common pattern in emotional dysregulation is displacement. You absorb stress from a manager at work and release it on your family at home — the anger lands somewhere it does not belong.

Healthy Ways to Release Emotions

Method	Scientific Effect	Recommended Practice
Moderate exercise (walking, running)	Burnout risk reduced by 62% (Kangbuk Samsung Hospital, study of 7,973 people)	25+ minutes daily
Talking with positive people	Emotional stability restored	2+ times per week
Hobbies that produce a sense of achievement	Improved self-esteem and focus	1–2 times per week
Creative activities (drawing, music, writing)	Emotional catharsis and creative expression	As desired

Interpersonal Skills: The Power of Having One Person in Your Corner

Emmy Werner's landmark 40-year longitudinal study in Kauai, Hawaii found one consistent factor among children who thrived despite poverty and abuse: at least one stable, supportive person in their life.

The key to genuine empathy is not saying "I've been through that too." Try this instead:

"That sounds like it was really hard. What part felt the worst to you?"

Positivity: How to Train Optimistic Thinking

Martin Seligman: "Optimism is not something you are born with — it is a skill you can learn."

Pessimistic Thought Pattern	Characteristic	How to Reframe
Permanence	"I'm always like this"	"This time, it went that way"
Pervasiveness	"Nothing works — this and that"	"Just this situation isn't working"
Personalization	"It's all my fault"	"It's not entirely my fault"

3-Week Action Plan: Start Training the Muscle of Your Mind Today

[Image: section4_image.png]

Week	Core Focus	Daily Practice (5–10 min)
Week 1	Notice your emotions	After work: write down 1 emotion you felt most today and why
Week 2	Gratitude + Strengths	Before bed: write 3 things you are grateful for + 1 thing you did well
Week 3	Connection + Body	Reach out to 1 person who supports you + start a 25-minute daily walk

Burnout is not a sign of personal weakness. It is a mismatch between the demands placed on you and the resources you have to meet them. On your commute home today, invest five minutes. Name one emotion you felt today. That is enough to begin.

Frequently Asked Questions

Q. Is resilience something you are born with?

A. Genetic factors account for only 30 to 50%. The remaining 50 to 70% can be improved through deliberate practice. Anyone can build resilience by consistently training the three elements: self-regulation, interpersonal skills, and positivity.

Q. How long does it take to build resilience?

A. Experts point to about three weeks of consistent practice as the starting point for habit formation. Sustained practice over three months or more can actually alter the brain's response patterns. Small, daily repetition matters more than ambitious starts.

Q. What is the connection between burnout and resilience?

A. Burnout is what happens when resilience runs low. A study of 7,973 workers at Kangbuk Samsung Hospital found that just 25 or more minutes of moderate daily exercise reduced burnout risk by 62%.

GoodTech AI

Technology for the benefit of people — slowly, steadily, one person at a time

Creative Questioning in the AI Era — How Metacognition and the Dunning-Kruger Effect Reveal Your Survival Strategy

GoodTechAdviser — Fri, 10 Apr 2026 09:19:14 +0900

✍️ GoodTech AI ⏱️ 5 min read

AI can generate answers to almost anything these days. So why do some people stagnate even with AI at their fingertips, while others grow several times faster? This post unpacks the root of that gap through three concepts: metacognition, the Dunning-Kruger effect, and divergent thinking — and offers concrete habits any professional can start using today.

Table of Contents

1. The Smartphone Paradox — Tools Don't Make You Wiser
2. The Curiosity Gap Society and the Power of Asking
3. Creativity and Meaning — What AI Cannot Do
4. The Dunning-Kruger Effect and Metacognition — Where Good Questions Begin
5. The 3 Elements of a Great Question
6. Daily Questioning Habits You Can Start Today
7. FAQ

The Smartphone Paradox — Tools Don't Make You Wiser

Imagine a person from the 1950s stepping into a modern subway car. What would their expression be?

In the palm of every commuter's hand sits a device containing the sum of human knowledge accumulated over thousands of years — every language's translator, lectures from the world's top experts, the contents of dozens of libraries. And what are people looking at? Cat videos. Someone's lunch. Arguments with strangers.

A tool works at the level of the person using it. AI is no different.

The Curiosity Gap Society and the Power of Asking

Professor Nophon Sompong at the University of Edinburgh describes our current moment as a "curiosity gap society." Access to information has become perfectly equal — yet knowledge gaps are not narrowing. The reason, he argues, is not the information itself but the difference in knowing what to ask.

World Economic Forum (WEF) 2026 Report

1. GenAI-related course enrollment: +185%
2. Critical thinking course enrollment: +168%

More and more people are recognizing that the heavier you rely on AI, the more deliberately you need to train your thinking. A 2025 CHI study found that knowledge workers who use generative AI more frequently tend to self-report lower critical thinking scores.

When answers come easily, we ask fewer questions. When we ask fewer questions, our thinking muscles weaken. The gap forms between those who break this cycle and those who don't.

Creativity and Meaning — What AI Cannot Do

In 1917, Marcel Duchamp signed a mass-produced urinal and titled it Fountain. The organizing committee rejected it — yet the work went on to reshape the entire course of 20th-century modern art.

The urinal itself hadn't changed at all. The meaning Duchamp assigned to it transformed it into something entirely new.

"Generative AI output is nothing more than a blurry JPEG of accumulated data." — Ted Chiang

AI extracts and recombines patterns from massive datasets. There is connection — but no act of meaning-making.

This is precisely where researcher Koo Bon-kwon, director of People and Digital Lab, defines the human role:

Convergent Thinking: Narrowing toward a single correct answer → increasingly handled by AI
Divergent Thinking: Expanding the field of possibility → the uniquely human domain

AI overwhelms humans at selecting the right answer on an exam. But only humans can ask: "Is this even the right question?" and "Is there a completely different direction we haven't considered?"

The Dunning-Kruger Effect and Metacognition — Where Good Questions Begin

The Dunning-Kruger Effect is a cognitive bias where people with limited knowledge tend to overestimate their ability, while those who know more often experience a temporary dip in confidence as they become aware of how much they don't yet know.

Many people start using AI and immediately think, "This thing can do anything." A few months later, watching someone more skilled, they realize how vague their own prompts had been. That moment of recognition — the valley on the Dunning-Kruger curve — is actually the starting point of real growth.

"I know that I know nothing." — Socrates

"To know what you know and know what you don't know — that is true knowledge." — Confucius

Metacognition is the capacity to ask yourself, "Do I actually understand this properly?" Acknowledging what you don't know is the doorway to better questions. People who cultivate this habit and those who don't will take entirely different trajectories in the age of AI.

The 3 Elements of a Great Question

Researcher Koo Bon-kwon identifies three components that make a question genuinely powerful:

① Foundational Knowledge — Standing on Giants' Shoulders

Newton said: "If I have seen further, it is by standing on the shoulders of giants." Pasteur echoed the same idea: "In science, inspiration visits only the prepared mind." A question without a knowledge base has no direction.

② Questioning Assumptions — The Socratic Method

"Is this premise actually correct?" "Is there something I'm taking for granted that might be wrong?" If you never question the problem itself, you'll only ever search for answers within its frame.

③ Concrete Context

"How should I do this?" and "We're in manufacturing, average employee age is 45, and I'm planning an AI literacy training program from scratch — where do I start?" are completely different questions.

Context	Weak Question	Strong Question
Asking AI	"Tell me how to write a report"	"I'm in HR at a manufacturing company writing a quarterly training outcomes report for executives. Key metrics are completion rate and on-the-job application rate. Please suggest an outline: 1-page summary + 3-page body."
In a meeting	"This approach seems fine, doesn't it?"	"What do you see as the biggest risk if this approach fails? And what conditions would we need in place to mitigate that risk?"
While learning	"How does this work?"	"Let me check whether I understood this correctly. My understanding is X — can you tell me if I'm missing anything or got something wrong?"

Daily Questioning Habits You Can Start Today

Practice 1 — Darwin's Objection Log

Charles Darwin never let a contradicting opinion slide by. He actively recorded disagreements and dug into them. When your proposal gets approved in a meeting, don't stop there — ask first: "What's the weakest link in this plan?" That one question changes the quality of the whole conversation.

Practice 2 — Articulate What You Don't Know Before You Search

When curiosity strikes, resist the urge to search immediately. Take 30 seconds to write: "Why don't I know this?" and "Exactly which part am I confused about?" The act of articulating the question deepens thinking. Answers you find this way stick far longer.

"Be patient toward all that is unsolved in your heart. Try to love the questions themselves." — Rainer Maria Rilke

What the most competitive people in the AI era have in common is not faster processing than AI. It's the habit of asking questions that AI cannot imitate.

FAQ

Q. What exactly is metacognition?

Metacognition is "thinking about thinking" — the ability to monitor and regulate your own thought process. In plain terms, it means asking yourself: "Do I really understand this?" and "Where exactly am I getting stuck?" It directly affects learning efficiency and problem-solving ability. In the AI era, it's increasingly recognized as the core skill that determines the quality of your questions.

Q. What is divergent thinking?

Divergent thinking is the opposite of convergent thinking, which narrows toward a single answer. Instead, it expands outward — exploring multiple possibilities and ideas. It works through open questions like "What other approaches exist?" and "Is there a completely different direction?" As AI increasingly handles convergent thinking, divergent thinking becomes the core human advantage.

Q. What is the Dunning-Kruger Effect?

The Dunning-Kruger Effect is a cognitive bias identified by psychologists David Dunning and Justin Kruger in 1999. People with limited knowledge or skill tend to overestimate themselves, while those who gain more knowledge become more aware of their own limitations and may temporarily feel less confident. The same pattern shows up with AI: at first, people think "AI is so easy!" — and then they see someone using it well and realize they'd been doing it all wrong. That moment of recognition is exactly where growth begins.

Just as smartphones didn't make everyone smarter, AI won't automatically make everyone more competitive. The ability to ask good questions starts with one small habit today: taking something you've always assumed and asking, "Is this actually right?"

That small difference creates a different velocity in the age of AI.

GoodTech AI

Technology for the benefit of people — growing one step at a time.

Published on Tistory | Post #056
Category: AI Skills / Future Competencies | March 1, 2026

Using AI as Your Idea Partner— From Brainstorming to Screening, Step by Step

GoodTechAdviser — Wed, 8 Apr 2026 12:56:22 +0900

✍️ GoodTech AI ⏱ 5 min read

Table of Contents

1. How AI Becomes a Brainstorming Partner
2. Dividing Roles Between Humans and AI
3. Setting Up AI with the 5-Element Prompt Framework
4. The Diverge → Validate → Screen Workflow
5. Multi-Session Strategy
6. FAQ

Have you ever typed "give me some ideas" into ChatGPT and felt underwhelmed by the response? If so, it's time to change how you ask. AI responds generically to vague requests — but when you structure your prompt precisely, it surfaces combinations you'd never have thought of yourself.

According to Professor Lee Junghye at Seoul National University's Graduate School of Engineering, AI plays five distinct roles in the idea-creation process. The key is transcending the limitations of traditional brainstorming. AI analyzes vast data to find connections humans miss, and it proposes solutions from multiple angles without unconscious bias. This guide walks you through how to configure AI as a genuine idea partner, step by step.

How AI Becomes a Brainstorming Partner

Professor Lee identifies five roles AI plays in idea generation:

Role	Description
Data-driven insights	Identifies market trends and unmet demand
Idea expansion	Surfaces creative solutions beyond human experience limits
Idea validation	Pre-evaluates feasibility, cost, and potential problems
Visual prototyping	Converts text descriptions into images and prototypes
Unbiased perspective	Diverse viewpoints unconstrained by cultural or experiential background

"AI doesn't replace creativity — it plays a complementary role, helping us explore more possibilities." (Prof. Lee Junghye, SNU Graduate School of Engineering)

Traditional brainstorming depends on participants' experience and group dynamics. People unconsciously filter out ideas that "might sound unreasonable." AI operates without that internal censor.

Dividing Roles Between Humans and AI

Author Park Jong-cheon breaks down workplace capabilities into four types:

1. Logic: Grasping the overall flow of a task → AI handles this
2. Analysis: Deep analysis of a specific domain → AI handles this
3. Creativity: Deciding what to do and how to instruct → Human handles this
4. Integration: Judgment to select the best outcome → Human handles this

"People who use AI well will take work from those who don't. AI doesn't steal jobs — people who use AI do." (Park Jong-cheon)

AI already surpasses humans in logic and analysis — it can read and summarize a 100-page document in seconds. What humans need to focus on is what to assign and what to choose.

Setting Up AI with the 5-Element Prompt Framework

Instead of asking AI vaguely, use a prompt built on five elements:

#	Element	Description	HR Example
1	Role	The persona AI should adopt	"You are an expert HR program designer"
2	Task	What needs to be done	"Propose ideas for a new employee onboarding program"
3	Output format	Form of the deliverable	"A numbered list of 10 ideas"
4	Context	Background information	"Manufacturing company, 50 new hires, 3-month program"
5	Example	Direction you want	"Like 'on-site mentoring' — practical and work-connected"

With these five elements in place, AI produces entirely different results. Park Jong-cheon emphasizes: "The more complex your instructions, the more valuable AI becomes." Detailed instructions feel burdensome to people — but they make AI more precise.

The Diverge → Validate → Screen Workflow

Step 1: Diverge — Ignore Feasibility for Now

"List 20+ possible ideas on this topic. Ignore feasibility for now."

Asking only for "realistic" ideas from the start narrows the scope immediately. This phase maximizes AI's logic and analysis capabilities.

Step 2: Validate — Filter by Adding Context

"From these, evaluate which ones could be implemented tomorrow from the perspective of an HR manager in manufacturing. Criteria: cost, floor-level acceptance, and preparation time."

Step 3: Screen — Develop Through Feedback Loops

"Develop idea #3 further — in a direction that middle managers can accept without resistance."

Professor Lee presents continuous improvement through feedback loops as the core strategy for AI-assisted ideation.

Multi-Session Strategy — Building Specialist Assistants

Park Jong-cheon suggests treating ChatGPT like "having several dedicated assistants" — separate sessions for each purpose, each pre-configured as a specialist.

Example multi-session setup for an HR professional:

1. Session A: Training program designer — idea generation
2. Session B: Report editor — polishing proposal language
3. Session C: Interview designer — developing employee survey questions
4. Session D: Trend researcher — gathering HR industry updates

The key is multi-turn conversation — staying within a single session and building context over time. The longer the context, the better the output.

Park cites a striking statistic: Korea's overall AI utilization rate sits at around 5 out of 100, while the US is at 30. The difference isn't in AI capability — it's in knowing how to give instructions.

FAQ

Q. What if ChatGPT gives me inaccurate information?

A. Provide relevant source material first and ask AI to answer based on that. Saying "Please refer to the following document to answer" dramatically reduces errors.

Q. Do I need a paid subscription?

A. Basic functionality works well enough with the free tier. However, multimodal features like file uploads and image analysis require a paid plan.

Q. How do I decide which screening criteria to use?

A. Park Jong-cheon's "importance vs. urgency" matrix is useful here. AI can surface ideas that aren't urgent but are highly important — so humans can focus on evaluating importance while AI does the discovery.

Closing Thoughts

AI is already an effective tool for opening up the flow of ideas. But where that flow goes is still a human decision. Explore more possibilities with AI — and let your final judgment come from your own experience and context.

Try starting one piece of work today with AI as your partner. Partnership builds from small beginnings.

GoodTech AI

Technology for the benefit of people — one step at a time.

A practical guide for professionals navigating the AI era.

Scheduling a Board Meeting Used to Take Half a Day — Claude AI + Google Forms Did It in 15 Minutes

GoodTechAdviser — Tue, 7 Apr 2026 11:14:05 +0900

GoodTech AI | 5 min read

Table of Contents

1. The Phone Call Problem
2. Asking AI to Design the Survey
3. From Spreadsheet to Full Report in One Prompt
4. Before vs. After: What Actually Changed
5. How to Replicate This Workflow
6. Frequently Asked Questions

It was 2 PM. I hung up the phone after my third call. The first board member said April's fourth week was entirely out. The second could only do Monday and Wednesday. The third needed to "check and get back to me." My notepad was covered in circles and X marks. Two more members to go, and an hour had already passed.

If you've ever tried to schedule a meeting with five busy people across ten candidate dates, you know this pain. Each person has a different calendar, different constraints, and different communication preferences. One round of calls is rarely enough.

This time, I tried something different. I asked Claude AI to design a survey for me. The result: what used to take half a day was done in 15 minutes. And the AI suggested a survey option I hadn't even considered — one that turned out to be the key to finding the optimal date.

The Phone Call Problem

Here's what coordinating a board meeting used to look like at my organization. Five external board members, each active in completely different fields. Ten candidate weekdays across two weeks in late April. The goal: find one day when all five can meet.

The traditional process had three critical flaws:

Time sink. Ten minutes per call, five members, and that's just the first round. Factor in missed calls, callbacks, and renegotiation, and you're looking at half a day — sometimes a full day.

Binary answers miss the nuance. Phone calls capture "yes" or "no." But reality is more subtle. Some people could attend if they rearranged their schedule. That conditional availability is invisible in a phone-based process, yet it's often the difference between finding a workable date and not.

Aggregation is a separate job. Even after collecting all responses, you still need to manually cross-reference dates, identify overlaps, rank alternatives, and write up a recommendation report for leadership. According to a 2025 survey by Megazone Cloud and Foundry covering 749 AI/IT professionals in Korea, the #1 reason companies adopt AI is "improving work efficiency and productivity" (70.5%). Schedule coordination is precisely the kind of repetitive, multi-variable task where AI excels.

Asking AI to Design the Survey

My original plan was simple: create a Google Form with ten dates and two options — "Available" or "Unavailable." Then I described the situation to Claude, and the AI pointed out something I'd missed.

"What about adding a 'Tentative — schedule adjustment needed' option?"

That single suggestion changed the quality of the entire scheduling process. With only two choices, a board member who could attend after rearranging something would have to select "Unavailable." The third option — tentative — let people express conditional availability. For the coordinator, this meant clearly distinguishing between "absolutely impossible" and "possible with some effort."

Here's the survey structure Claude designed:

Element	Details
Response format	3-tier selection per date
Option 1	Available
Option 2	Tentative (schedule adjustment needed)
Option 3	Unavailable (completely impossible)
Respondents	5 board members
Candidate dates	10 weekdays across 2 weeks

Claude also drafted the invitation message in a formal business tone appropriate for senior external board members. I sent the Google Forms link via text message, and each member completed it in one to two minutes at their convenience. No more waiting for the right moment to call.

The response from leadership? They praised the survey design. The thoughtful question structure — particularly the "tentative" option — was noticed and appreciated. That was Claude's contribution, not mine.

From Spreadsheet to Full Report in One Prompt

Once all responses were in, Google Forms automatically generated a spreadsheet. I uploaded the Excel file to Claude with a single prompt:

"Analyze the survey results, create optimal schedule alternatives ranked 1 through 4, and if there's no date that works for everyone, give me your recommendation."

Five minutes later, the results appeared. Zero days had unanimous "Available" responses. Under the old system, this would have meant another round of phone calls. But Claude read the data differently.

"April 30 (Thursday) is the only date with zero 'Unavailable' responses. Four members selected 'Available' and one selected 'Tentative.' If the tentative member's schedule can be adjusted, full attendance is possible."

Without the "Tentative" option, this date would have been buried among the others. The survey captured the nuance, and the AI identified the signal within it.

Here's everything Claude produced from that single prompt:

• Priority alternatives table — 4 options ranked by fewest conflicts, most availability
• Word report — Color-coded response heatmap + ranked alternatives + operational recommendations
• Visualization — Chart showing response distribution per date with the optimal date highlighted
• Email drafts — Two versions (concise and detailed) for reporting to leadership

All of this from one uploaded file and one sentence of instruction.

Before vs. After: What Actually Changed

Fifteen minutes. That's how long it takes to brew a cup of coffee. In that time, the survey was designed, sent, collected, analyzed, alternatives were ranked, a report was generated, and an email draft was ready.

Step	Before (Manual)	After (AI + Google Forms)
Collecting availability	Individual phone calls to 5 members; callbacks for missed calls	Google Forms link sent via text; each person responds in 1–2 min
Response granularity	Binary yes/no; conditional availability lost in conversation	3-tier options capture "Available," "Tentative," and "Unavailable"
Aggregation	Manual cross-reference on notepad; pattern-matching by eye	Claude auto-parses Excel → ranks dates by minimum conflicts
Reporting	Separate Word document; manual table creation	Word report with color-coded heatmap + visualization + email draft
Total time	Half a day to full day	~15 minutes

The most striking part wasn't the time saved. It was that the AI created an option the human hadn't considered. I planned for two choices. Claude suggested a third. That third choice — "Tentative" — was the reason we found the optimal date.

Gartner projects that by 2026, 40% of enterprise applications will integrate task-specific AI agents (via SK AX). Microsoft's Copilot Wave 3, announced in March 2026, already automates the full scheduling pipeline: scanning attendee calendars, finding available slots, booking rooms, and sending invitations.

But many organizations aren't ready for those enterprise solutions yet. In Korea, only 5.3% of small and mid-sized businesses have adopted AI (Electronic Times, 2025). That's exactly why the Google Forms + Claude combination matters — it delivers similar results with zero infrastructure, using tools that are already free and familiar.

How to Replicate This Workflow

If you coordinate schedules for multiple people, here's the process in three steps:

Step 1: Describe your situation to Claude. Don't just say "make a survey." Explain who needs to meet, how many candidate dates you have, and what constraints exist. The AI will suggest response options you haven't thought of.

Step 2: Send the Google Form and collect responses. Share the link via text, email, or messenger. Let people respond on their own time.

Step 3: Upload the spreadsheet to Claude. Ask for analysis, alternatives, and a report. One prompt is often enough for all three.

This workflow isn't limited to board meetings. Team dinners, workshop scheduling, training attendance checks — any situation where you need to find a date that works for a group of people.

Frequently Asked Questions

Q. Do I need coding skills to use this workflow?

Not at all. The entire process uses Google Forms (free) and a conversation with Claude. No APIs, no scripts, no technical setup required. If you can send a text message and upload a file, you can do this.

Q. What if more than 10 people need to respond?

Google Forms handles any number of respondents easily. For AI analysis, Claude can process spreadsheets with dozens of rows without issues. For very large groups (50+), consider breaking the analysis into sections or using models with larger context windows.

Q. Can I use ChatGPT instead of Claude for the analysis?

Yes. The workflow works with any capable AI assistant. The key insight — designing better survey options by describing your situation to AI — applies regardless of which tool you use. Claude, ChatGPT, and Gemini can all analyze spreadsheet data and generate reports.

GoodTech AI

Technology for People

Don't Pay for KakaoTalk Backup When Upgrading Your Galaxy — How AI Sorted 31,536 Photos in 9 Minutes

GoodTechAdviser — Mon, 6 Apr 2026 16:00:19 +0900

GoodTech AI. 5min read

1. The Saturday Evening SOS Call
2. Smart Switch Backs Up KakaoTalk for Free
3. The Backup Worked — But 44 Folders of Chaos Remained
4. How Claude Code Sorted 343GB in 9 Minutes
5. Your Galaxy Upgrade Checklist
6. Frequently Asked Questions

The Saturday Evening SOS Call

My sister called on a Saturday evening.

"I switched to a new Galaxy, but I can't transfer my photos. It's been three days."

She had just bought a new Galaxy and needed to move nine years of photos and videos from her old phone. App data and KakaoTalk messages had already been transferred, but the photos — over 31,000 of them — were too large for any app-based transfer.

Here's what she tried:

Attempt 1: A cheap OTG USB from Coupang (KRW 40,000 / ~$29). It claimed 2TB capacity. After waiting through an "18 hours remaining" message, it crashed repeatedly over three days. The seller deleted the listing and disappeared. Money and time — gone.

Attempt 2: Direct USB connection. Errors on every attempt. No luck.

Attempt 3: Samsung Smart Switch. This one finally worked — but what came next was a different kind of nightmare.

She called me. I opened Smart Switch on my computer. Then I opened Claude Code.

Smart Switch Backs Up KakaoTalk for Free

Before I helped my sister, let me share my own mistake.

When I upgraded my Galaxy recently, my biggest worry was KakaoTalk — South Korea's dominant messaging app, used by virtually everyone. Conversation histories, photos, group chats. Losing them felt unthinkable.

So I paid for KakaoTalk's premium backup service. Just to be safe.

Then I discovered that Samsung Smart Switch transfers KakaoTalk data for free — including open chats. I had paid for something that was already available at no cost.

In April 2024, Samsung and Kakao partnered to integrate KakaoTalk data migration directly into Smart Switch (ZDNet Korea, 2024). The only requirements:

1. Android OS 12 or higher(Galaxy S21 and newer models qualify)
2. KakaoTalk must NOT be installed on the new phonebefore running Smart Switch

That second point is critical. If you set up your new Galaxy and install KakaoTalk before completing the Smart Switch transfer, your chat history won't restore. The order matters — and Samsung community forums are filled with users who learned this the hard way.

Feature	KakaoTalk Paid Backup	Smart Switch (Free)
Cost	Monthly or per-use fee	Free
Chat history	Yes	Yes
Open chats	No	Yes (since Apr 2024)
Photos & videos	Yes	Yes
Requirement	In-app settings	KakaoTalk not installed on new phone

Don't make the same mistake I did.

The Backup Worked — But 44 Folders of Chaos Remained

Back to my sister's situation.

Smart Switch PC backed up her phone successfully. 343GB landed on an external hard drive. So far, so good.

Then we opened the backup folder.

SM-S916N/SM-S916N_01041335606/SM-S916N_20260308213456/
├── Photo/DCIM/Camera/
├── PHOTO_ORIGIN/
├── Video/DCIM/Camera/
├── VIDEO_ORIGIN/
├── CONTACT/
├── MESSAGE/
└── ... (44 subfolders total)

SM-S916N_01041335606. What does that even mean? There were 44 subfolders with cryptic names. Finding where the actual photos lived meant clicking through each one manually.

My sister said, "I'm too scared to touch anything. What if I delete something by accident?"

Here's what was inside:

Folder	Files	Size	Type
Photo/DCIM/Camera	28,974 photos	246GB	.jpg
PHOTO_ORIGIN	56 photos	430MB	.jpg
Video/DCIM/Camera	2,493 videos	96GB	.mp4
VIDEO_ORIGIN	13 videos	970MB	.mp4
Total	31,536 files	343GB	—

Nine years of memories. From 2017 to 2026. A nephew's first birthday, family vacations, holiday gatherings. Losing even one folder was not an option.

How Claude Code Sorted 343GB in 9 Minutes

I opened Claude Code and described the situation:

"This folder contains a Galaxy Smart Switch backup. Search the web to understand the backup structure, then organize photos into a 'Photos' folder and videos into a 'Videos' folder with clear Korean names."

Claude Code first searched the web to understand Samsung Smart Switch's backup structure. It confirmed that Photo/DCIM/Camera contains camera photos and PHOTO_ORIGIN stores pre-edit originals. No human needed to open 44 folders one by one.

Then it created clearly named Korean folders and moved everything:

/Volumes/PRO-G40/S23 Backup/
├── 사진 모음 (Photos)/
│   ├── 카메라 (Camera)/      → 28,974 photos (246GB)
│   └── 원본사진 (Originals)/ → 56 photos (430MB)
├── 동영상 (Videos)/
│   ├── 카메라 (Camera)/      → 2,493 videos (96GB)
│   └── 원본동영상 (Originals)/ → 13 videos (970MB)
└── backup/                    → Other data (contacts, messages, etc.)

Since everything stayed on the same drive, it used the mv command — changing file paths without physically moving data. Instant. Then it verified the file count: zero files missing.

The comparison speaks for itself:

Method	Time	Cost	Result
Coupang OTG USB	3 days	$29	Failed (seller vanished)
Direct USB	12 hours	$0	Repeated errors
Smart Switch + Claude Code	Backup 60 min + Sort 9 min	$0	Complete success

Three days became nine minutes.

Your Galaxy Upgrade Checklist

The average smartphone replacement cycle in South Korea is 31 months (Counterpoint Research, 2025) — significantly shorter than the global average of 48 months. If you bought a Galaxy S23 at launch, you're due for an upgrade right about now.

And in South Korea, 82% of smartphone users choose Galaxy. Among users in their 50s, it's 89%. Among those 60 and older, 92% (Gallup Korea, 2025).

Here's what to remember:

1. Install Smart Switch PC first— Free download from Samsung's official website. It backs up photos, videos, contacts, calendars, notes, and device settings in one go.

2. Do NOT install KakaoTalk on your new phone first— Complete the Smart Switch restore before opening KakaoTalk. If you reverse this order, your chat history is gone.

3. If the backup folders look overwhelming, ask AI— One line to Claude Code: "Organize this folder for me." That's all it takes.

Frequently Asked Questions

Q. Does Smart Switch really transfer KakaoTalk data for free?

Yes. Since April 2024, Samsung and Kakao's partnership enables full KakaoTalk data transfer — including open chats — through Smart Switch at no cost. The key condition: KakaoTalk must not be installed on the new phone before running Smart Switch.

Q. Does this work for Galaxy S23, S24, and S25 models?

Yes. Any Galaxy running Android 12 or higher supports this feature. That includes Galaxy S21 and all newer models (S22, S23, S24, S25 series).

Q. Do I need coding skills to use Claude Code for file organization?

No. Claude Code understands natural language. You describe what you want — "sort these photos by date" or "organize this backup folder" — and it handles the rest. It searched the web on its own to understand Samsung's backup structure.

Final Thoughts

When I showed my sister the organized folders, she went quiet for a long moment.

"...That took nine minutes?"

Nine years of memories — a nephew's first birthday in 2017, a family trip in 2019, a holiday photo from last month — all sitting neatly in folders she could actually read.

Technology isn't hard. Not knowing where to ask is what makes it feel hard. Smart Switch was free. Claude Code was ready to help. Both had been there all along — she just didn't know.

A single photo holds the air of that day. Keeping that air safe just got a little easier.

#SamsungGalaxy #SmartSwitch #KakaoTalkBackup #PhotoBackup #ClaudeCode #AIFileOrganization #SmartphoneUpgrade #DataTransfer #GalaxyS25 #GalaxyS24

GoodTech AI

Technology for the benefit of people

References:
ZDNet Korea, "Samsung-Kakao partnership enables KakaoTalk data transfer via Smart Switch," April 2024
Samsung Service, "Galaxy Smart Switch Backup/Restore Guide" (samsungsvc.co.kr)
Counterpoint Research / Chosun Biz, "South Korea smartphone replacement cycle: ~31 months," January 2026
Gallup Korea, "Galaxy market share 82% in Korea; 50s: 89%, 60+: 92%," 2025
Samsung Newsroom, "Samsung x Kakao collaboration story for user convenience"

Human Creativity in the AI Era: Why the Top 10% Will Always Outpace Machines (2026)

GoodTechAdviser — Sun, 5 Apr 2026 15:08:35 +0900

GoodTech AI · 5 min read

A few years ago, a major American fast-food chain installed an AI ordering system at its drive-throughs. Customers asking for a coffee would find ice cream added to their order. Repeating the same words twice produced entirely different results. The system was eventually pulled.

That story felt reassuring at first. But the reassurance did not last long. Because it turned out that while AI struggled to take a drive-through order, it was rapidly getting very good at report writing, drafting proposals, and analyzing data — work that many of us have spent years building expertise in.

So the question shifted: not "Will AI replace me?" but "What can AI not replace?"

In 2025, a research team at the University of Montreal set out to answer exactly that. They ran a large-scale creative thinking assessment with 100,000 human participants alongside GPT-4, Claude, and Gemini (Scientific Reports, 2025). The results were striking in both directions.

Some AI models already exceeded the average human creativity score. That part got the headlines. But the full picture matters more: the top 50% of human participants outperformed every AI tested. And the top 10% maintained a substantial lead that no model came close to closing.

The conclusion is precise: AI has surpassed average human creativity. It has not touched the upper range of human creative ability.

This raises an important follow-up: what separates the top 10% from the average — and how do you get there?

1. Why AI Gets Stuck at Average
2. Creativity Is Connection: What Einstein Proves
3. Lessons from Monet and John Cage
4. The Real Risk: Letting AI Do All Your Thinking
5. 3 Ways to Protect and Build Your Creativity
6. FAQ

Why AI Gets Stuck at Average

To understand why AI creative output tends toward the middle, you need to understand what AI is fundamentally doing. Every response an AI generates is a statistically informed prediction: given everything it has learned, what is the most plausible continuation of this input?

That is an extraordinary capability for many tasks. But it is also a structural ceiling on originality.

Christian Terwiesch, a professor at Wharton Business School, described the dynamic clearly in a 2025 study:

"If ChatGPT is used as the sole creative advisor, the ideas will soon run out because they are too similar to each other. The AI model tries to average out the most likely completion based on the given input."

His research found that groups using ChatGPT for creative ideation produced individually higher-quality ideas — but across the group, the ideas converged. Different people ended up generating similar outputs. Diversity of thought collapsed. (Wharton Business School, 2025)

This pattern has shown up in creative industries globally. When AI-generated content flooded online platforms, audiences began noticing a sameness — technically competent work that felt emotionally thin. In South Korea, AI-generated webcomics prompted a reader backlash when the content felt indistinguishable from other AI work. Major platforms responded by restricting competition entries to human-created work. Audiences sensed the absence of something they could not always name: genuine perspective.

Dimension	AI Creativity	Human Creativity
Starting point	Patterns in past data	Personal desire and conviction
Output	The most statistically probable answer	New territory beyond existing patterns
Driver	Optimization within known space	Emotion, lived experience, culture
Ceiling	Cannot originate outside trained distribution	Can break the pattern itself

AI builds a better average. Creativity starts where the average ends.

Creativity Is Connection: What Einstein Proves

Steve Jobs defined creativity as "connecting things." Philosopher David Bohm described its result as "creating a harmonious, whole, and beautiful order."

Einstein is the cleanest proof of both definitions.

He took two concepts that no physicist had joined before — time and space — and connected them into a single unified framework: spacetime. The Special Theory of Relativity (1905) did not emerge from finding a pattern in existing data. It came from a question Einstein had held for years: what would it feel like to ride alongside a beam of light?

That question was not in the data. It came from Einstein's particular obsession — what you might call a deep personal aspiration. The eventual answer reshaped physics, enabled atomic clocks, and made GPS possible.

An AI trained on thousands of physics papers would be excellent at summarizing the state of physics as of its training date. It would not generate the concept of spacetime for the first time, because that concept did not exist in the data yet. It had to be created — and creation, at its root, comes from desire, not distribution.

This maps neatly onto Maslow's hierarchy of needs. Each major technological era has pushed humanity up the ladder. The Industrial Revolution met basic physiological and safety needs at scale. The internet era addressed belonging and esteem. The argument now is that AI could be the tool that enables self-actualization — not by thinking for us, but by clearing away enough routine work that we have real space to think for ourselves.

The paradox of AI is that it could either accelerate or suffocate this development, depending entirely on how we use the time it frees up.

Lessons from Monet and John Cage: What Creative Thinkers Have in Common

What do the people who land in the top 10% actually do differently?

Claude Monet was working in 19th-century France, when the art establishment expected painters to depict churches and mythological scenes. The prevailing style had been trained, through centuries of patronage and tradition, to produce a particular kind of output. Monet sat alone in front of a pond of water lilies. He trusted what he personally found beautiful: natural light on water. Impressionism was not a trend he followed. It was something he initiated by refusing the data set entirely.

An AI trained on that same art-historical corpus would have produced a technically accomplished church painting, well-averaged across thousands of examples. Monet stepped outside the corpus.

John Cage did something even more radical. In 1952, he premiered a composition called 4'33" — four minutes and thirty-three seconds during which the performer sits at an instrument and plays nothing. What the audience hears is the ambient sound of the room: chairs shifting, people breathing, wind, traffic. Cage was declaring that all of it is music. This idea was not derivable from any existing musical data. It required someone willing to question the definition of the category itself.

James Kaufman, a professor of educational psychology at the University of Connecticut, frames the distinction this way (UConn Today, January 2026):

"AI is excellent at generating ideas, but evaluating which ideas are worth pursuing, and making the judgment call about direction — that remains a human responsibility. Creativity requires more than idea generation."

The common thread across Monet, Cage, and Einstein is not genius in the popular sense. It is the willingness to hold a personal conviction strongly enough to go somewhere the existing data does not point. And conviction, by definition, cannot be averaged.

The Real Risk: What Happens When You Let AI Do All Your Thinking

A 2024–2025 survey of working professionals asked respondents to identify the most important skill for the AI era. Creativity and innovation ranked first at 37.1% — ahead of communication and collaboration (34.9%) and AI proficiency itself (34.3%).

Even the people who use AI most effectively recognize that creativity is what matters most.

But a separate government survey of AI users in 2025 found that 60.4% identified declining creativity as the biggest downside of generative AI — more than misinformation, privacy concerns, or job displacement.

This is the central tension. AI frees up time. But if the freed-up time goes toward consuming more content, doing more shallow tasks, or simply outsourcing more thinking, the net effect on creativity is negative.

Frank Herbert's science fiction novel Dune opens with a civilization that banned thinking machines — not because the machines failed, but because humans who delegated all their thinking eventually lost the capacity to think. Herbert called this the Butlerian Jihad. The warning is not that AI will revolt. It is that the atrophy happens quietly, through convenience.

The question is not whether to use AI. The question is what you do with the time AI gives back.

3 Ways to Protect and Build Your Creativity Right Now

1. Use AI for repetitive work, not directional thinking— Let AI handle research, summarization, formatting, and first drafts. Reserve the question ofwhatto create andwhyfor yourself. The moment you ask AI "what should I do?" you have outsourced the part of the work that builds creative muscle.

2. Invest reclaimed time in human thinking— If AI saves you 30 minutes on a task, do not immediately fill that 30 minutes with more tasks. Spend some of it thinking without a screen. Talk to a colleague about something you do not yet understand. The connections that drive original ideas form in unstructured time.

3. Set direction yourself, then hand off execution— Instead of prompting AI with "write me something about X," try "here is the angle I want to take, here is why I think it matters, now help me develop it." This keeps creative ownership with you and uses AI as a craft amplifier rather than a direction-setter.

The goal is not to avoid AI. It is to remain the person who decides what is worth making — and why.

Start small. Ten minutes at lunch, away from any screen, asking yourself: what do I actually want to create? What connection am I trying to make?

Monet, Cage, and Einstein did not begin by breaking all the rules. They began by paying close attention to what they personally found compelling — and refusing to let the existing data talk them out of it.

FAQ

Q. Has AI actually been proven to outperform humans creatively?

Partially. The 2025 University of Montreal study (Scientific Reports) tested 100,000 humans against GPT-4, Claude, and Gemini on standardized creative thinking tasks. Some AI models exceeded the average human score. However, the top 50% of human participants outperformed all AI, and the top 10% maintained a significant gap. AI has surpassed average human creativity — not human creativity at its best.

Q. Does using AI make you less creative over time?

There is reason for concern. A 2025 government survey of AI users found 60.4% identified creativity decline as the top downside of generative AI. The Wharton study found that groups relying heavily on ChatGPT for ideation converged on similar ideas — a phenomenon called creative homogenization. The risk is not AI itself but how the saved time is spent. Using AI for execution while preserving your own directional thinking appears to protect creative capacity.

Q. What is the most effective way to build creativity in the AI era?

Use AI to eliminate routine cognitive load, and redirect that freed-up energy toward the things AI cannot do: making unexpected connections, holding a strong personal conviction, asking questions that do not yet have answers. Creativity grows through unstructured thinking time and genuine engagement with problems — neither of which AI can substitute.

Sources: University of Montreal / Scientific Reports (2025) — 100,000-person creativity study; Wharton Business School, Christian Terwiesch (2025) — AI creative ideation research; UConn Today, James Kaufman (January 2026) — AI and creative judgment; South Korean Ministry of Science and ICT, Intelligent Information Society User Panel Survey (2025

GoodTech AI

Technology for good, one person at a time

Anthropic Asked 81,000 People What They Want From AI — The Answer Wasn't 'Faster Work'

GoodTechAdviser — Sat, 4 Apr 2026 08:30:28 +0900

GoodTech AI | Reading time: 5 min

Table of Contents 1. Why This Survey Is Different
2. What 80,508 People Actually Want
3. The 81% Verdict — With a Catch
4. Light and Shade: Five Tensions That Define AI Today
5. A Global Emotional Map of AI
6. What This Means for Your Workplace
7. Frequently Asked Questions

159 countries. 70 languages. 80,508 people.

Those numbers sound like a UN General Assembly roll call. But this isn't geopolitics — it's something arguably more personal. In December 2025, Anthropic deployed a variant of its AI Claude as an interviewer, asking users one simple question: What do you hope for — and fear — from AI?

The top answer surprised me. As someone who has spent years in HR and corporate training, I expected "make my work faster" to dominate. It didn't. The number-one vision — chosen by 18.8% of respondents — was "professional excellence," which sounds like a productivity play on the surface. But dig deeper, and these people weren't asking to produce more. They wanted to automate the repetitive grind so they could reclaim time — for family dinners, neglected hobbies, or simply a moment to breathe.

In other words: people don't want AI to help them work faster. They want AI to help them live better.

Why This Survey Is Different

Most AI surveys give you checkboxes: Positive? Negative? Neutral? Anthropic took a radically different approach. They deployed an AI interviewer that conducted open-ended conversations in 70 languages, letting respondents describe their hopes and fears in their own words. The company calls it "the largest and most multilingual qualitative study in history."

This matters because qualitative research at this scale was, as one analyst noted, "flatly impossible even two years ago." Traditional focus groups max out at a few hundred participants. Survey forms lose nuance. Anthropic used AI to achieve both scale and depth — an ironic proof of concept for the very technology being evaluated.

A critical caveat: Every respondent was already a Claude user. This is a self-selecting, AI-friendly population. The 67% net-positive sentiment likely runs higher than you'd find in the general public. Keep that in mind — but also recognize that these are people speaking from actual experience, not hypotheticals.

What 80,508 People Actually Want From AI

Anthropic classified responses into nine visions:

Rank	Vision	Share	What It Really Means
1	Professional Excellence	18.8%	Automate routine → focus on strategic work
2	Personal Transformation	13.7%	Mental growth, self-awareness, wellbeing
3	Life Management	13.5%	Reduce cognitive overload
4	Time Freedom	11.1%	Reclaim time for relationships
5	Financial Independence	9.7%	Economic security through AI
6	Societal Transformation	9.4%	Solve poverty, disease, climate
7	Entrepreneurship	8.7%	Build and scale businesses
8	Learning & Growth	8.4%	Acquire knowledge, develop skills
9	Creative Expression	5.6%	Bring artistic visions to life

Here's what jumped out at me: if you add up ranks 2 through 5, 48% of respondents wanted something that has nothing to do with work output. They want to grow as people, organize their chaotic lives, get their evenings back, and achieve financial stability. This isn't a tool request — it's a quality-of-life aspiration.

When I run AI training sessions for employees, I always ask: "What would you use AI for?" The most common answer is "write reports faster." But when I follow up with why, the real answer comes out: "So I can leave on time." The data from 80,000 people confirms what I hear in break rooms every week.

✅ The 81% Verdict — With a Catch

When asked whether AI has actually delivered on their vision, 81% said yes — at least partially. Here's how the impact breaks down:

Impact Area	Share	Significance
Productivity	32.0%	Faster work, automation of repetitive tasks
Unmet Expectations	18.9%	AI fell short of hopes
Cognitive Partnership	17.2%	Brainstorming partner, thinking companion
Learning	9.9%	New skills and knowledge
Technical Accessibility	8.7%	Building things without coding
Research Synthesis	7.2%	Processing large volumes of information
Emotional Support	6.1%	Judgment-free guidance

The uncomfortable truth hides in the economic breakdown. Freelancers and entrepreneurs reported economic gains at 47%, while employees at organizations reported just 14%. Same AI, dramatically different outcomes. Why? Freelancers can apply AI to their business the day they learn it. Employees wait for organizational approval, security reviews, and policy guidelines. It's not a tool problem — it's a structure problem.

⚖️ Light and Shade: Five Tensions That Define AI Today

The survey's most striking finding is what Anthropic calls "Light and Shade" — the discovery that AI's benefits and harms aren't opposing camps. They're two sides of the same capability.

The top three concerns:

Rank	Concern	Share	The Fear
1	Unreliability	26.7%	Confidently wrong answers, verification fatigue
2	Jobs & Economy	22.3%	Being replaced, wage stagnation
3	Loss of Agency	21.9%	Losing the ability to think independently

On average, each respondent expressed 2.3 concerns, and 89% had at least one. Most people who love AI are also anxious about it.

Anthropic's research team put it perfectly: "The things people love most about AI are often the very things they fear." This paradox manifests as five tensions:

1. Learning vs. Cognitive Atrophy — 33% said AI expanded their learning opportunities. But 17% felt their own thinking was deteriorating. Educators reported witnessing cognitive decline at 2.5–3x the average rate — teachers see it happening in real time.

2. Time Savings vs. Illusory Productivity — 50% saved time. But 19% felt busier. The saved hours get filled with more tasks.

3. Better Decisions vs. Unreliability — 22% made better decisions with AI. But 37% were exhausted by hallucinations. One lawyer said: "I'm slowly forgetting how to think for myself."

4. Economic Empowerment vs. Displacement — 28% gained economically. But 18% feared replacement.

5. Emotional Support vs. Dependency — 16% valued AI's judgment-free emotional guidance. But 12% feared becoming dependent on it.

A Global Emotional Map of AI

Globally, 67% expressed net-positive sentiment toward AI. But the regional variation tells a deeper story:

Most optimistic: Sub-Saharan Africa, South Asia, Latin America
Most cautious: Western Europe, Oceania, North America

The pattern is counterintuitive until you think about it: wealthier nations already feel AI in their workplaces — and with that comes real anxiety about job displacement. In developing economies, AI hasn't fully penetrated the job market yet, so it's seen as a ladder — a tool for entrepreneurs to access global markets without traditional infrastructure.

East Asia presents a unique case. While Western respondents worried about governance — surveillance, regulation gaps, institutional control — Korean and Japanese respondents worried about something more personal: the loss of meaning. Not losing a job, but losing the purpose within a job. It's a philosophical concern that speaks to East Asian cultural emphasis on craftsmanship and personal mastery.

One Korean respondent captured it: "I memorized AI's answers and got good grades, but I hadn't actually learned anything. That's when I felt the most guilt."

Korea by the numbers: According to Korea Herald and OECD data, 51.8% of Korean workers use AI at work — nearly double the US rate of 26.5%. Korea adopts fast, but organizational AI policies lag far behind. The gap between rapid adoption and slow institutionalization is Korea's defining AI challenge right now.

What This Means for Your Workplace

If you're a manager, HR professional, or team lead, here's what 80,000 voices are telling you:

1. Reframe AI training. Don't sell "productivity tools." Sell "time you get back." The data shows people want life quality, not work quantity.

2. Close the structure gap. The 47% vs. 14% freelancer-employee divide isn't about capability — it's about organizational friction. Reduce approval layers. Create safe-to-experiment zones.

3. Watch for cognitive atrophy. If educators see it at 2.5x the average rate, it's happening in your teams too. Build "AI-free thinking time" into workflows.

4. Address the meaning question. Especially in East Asian workplaces, AI anxiety isn't about losing a paycheck — it's about losing purpose. Help employees find the parts of their work that AI can't replace.

5. Acknowledge the paradox. Don't pretend AI is all upside. The 89% who have concerns aren't Luddites — they're realists. Create space for honest conversations about both light and shade.

The 81,000 voices in this survey aren't asking us to choose between embracing AI and resisting it. They're asking for something harder: the wisdom to take the benefits while staying alert to the costs we can't yet see.

Frequently Asked Questions

Q. What was the biggest surprise in Anthropic's 81,000-person AI survey?

The top finding was that people don't primarily want AI for faster work — 48% of the top visions relate to personal growth, life management, time freedom, and financial independence rather than workplace productivity. The study also revealed that AI's benefits and fears aren't opposing forces — they emerge from the same capabilities, creating five recurring tensions.

Q. How does AI adoption in South Korea compare to other countries?

South Korea has one of the highest AI adoption rates globally: 51.8% of Korean workers use AI at work, nearly double the US rate of 26.5% (Korea Herald/OECD). However, Korean respondents expressed unique concerns about the loss of personal meaning and purpose in work — a more philosophical anxiety compared to Western concerns about governance and regulation.

Q. Is the Anthropic survey representative of the general public?

Not entirely. All 80,508 respondents were existing Claude users, making them an AI-friendly, self-selecting group. The 67% net-positive sentiment likely runs higher than you'd find in the general public. However, this also means the reported concerns (89% had at least one) carry extra weight — even AI enthusiasts are worried.

Claude Code for Non-Developers: How to Automate Work from the Terminal Without Coding (Part 2)

GoodTechAdviser — Fri, 3 Apr 2026 08:41:42 +0900

GoodTech AI · 5 min read

1. What Is Claude Code and How Is It Different from Cowork?
2. How to Install Claude Code in 5 Steps
3. 5 Prompt Tips for Non-Developers
4. Real-World Automation Cases
5. Pricing: Is $20/Month Enough?
6. Limitations You Should Know
7. FAQ

What Is Claude Code and How Is It Different from Cowork?

I remember the first time I opened a terminal window. A black screen with a blinking cursor. Nothing else. It felt like touching the wrong thing would break my computer. So I closed it. I was an HR professional -- terminals were not for people like me.

That changed the moment I typed claude and pressed Enter.

Bloomberg reported that AI coding agents are triggering a "productivity panic" across the tech industry -- and adoption is spreading beyond developers to non-technical professionals (Bloomberg, February 2026).

If Cowork is "using an app to give AI instructions," Claude Code is "talking to AI directly through the terminal to create, modify, and automate files." You still use plain language. The difference is scope.

Feature	Cowork	Claude Code
Interface	Web/app UI	Terminal (CLI)
File manipulation	Limited	Create, edit, delete directly
Integrations	Basic	10+ MCP (Slack, Notion, etc.)
Automation	Simple tasks	Complex multi-step workflows
Best for	AI first-timers	Ready to open a terminal

The word "terminal" stops many people. But once you open it, it is just a conversation window. Type "organize these files for me" in plain English and it works.

How to Install Claude Code in 5 Steps

Prerequisite: Claude Pro ($20/month) or higher. Free plan does not support Claude Code.

Step	Action	Command
1	Subscribe to Pro/Max	claude.com/pricing
2	Open Terminal	Cmd+Space, search "Terminal"
3	Install Homebrew	Paste from brew.sh
4	Install Claude Code	brew install claude-code
5	Log in	claude login

Windows: Install Git for Windows first, then follow the same steps. Full feature parity.

Tip: If the terminal feels intimidating, use VSCode's integrated terminal. After installation, type claude -- that one word is your first milestone.

5 Prompt Tips for Non-Developers (From the Creator)

From Boris Cherny, creator of Claude Code (Source: gpters.org, 2026):

1. Start with Plan mode

Type /plan and Claude Code shows its plan before executing. Approve or modify. The easiest way to prevent mistakes.

2. Use /clear between major tasks

Long histories slow responses. /clear resets the context. Make it a habit at each task boundary.

3. Undo with Esc Esc

Press Escape twice to roll back via Checkpoints & Rewind. Note: changes after the rollback point are permanently lost.

4. Be specific with natural language

Weak: "Organize files"
Strong: "Sort Excel files in this folder by date, create monthly folders, and move files accordingly"

5. Create a CLAUDE.md file

Save recurring rules ("Always respond in English," "Keep reports under one A4 page"). Claude Code reads it automatically every session -- like onboarding notes for a new team member.

Real-World Automation Cases for Office Professionals

If you work in HR, you know this scenario: end-of-month training completion reports. Opening dozens of Excel files, checking names, departments, and completion status by eye.

I told Claude Code: "There are Excel files in this folder. Read the name, department, and completion status columns from each file, merge them into one, and create a separate sheet listing everyone who hasn't completed training." Minutes later, done.

Task	Before (Manual)	After (Claude Code)
Excel data aggregation	Copy-paste across files	Auto-read folder, merged file
Report draft	2-3 hours template filling	Auto-generate from data
File organization	Move files one by one	Auto-sort by date/name
Email drafts	Write each from scratch	Template-based auto-generation
Job postings	Write each JD from scratch	Auto-variation from existing JDs

Reported results: document creation 60% faster, communication errors 25% lower, and 90% savings vs. outsourcing in some cases (retn.kr, 2026).

Pricing: Is Pro at $20/Month Enough?

Plan	Price	Claude Code	Best For
Free	$0	Not available	--
Pro	$20/mo	Available	Recommended entry
Max 5x	$100/mo	Available	4+ hrs daily
Max 20x	$200/mo	Available	Power users
API	Variable	Available	Caution: cost spikes

For non-developers, Pro at $20/month is the right starting point. API pay-per-use can spike unpredictably -- one complex hour reportedly cost $100. Start flat-rate.

Limitations You Should Know Before Starting

Think of Claude Code as a "junior team member." Great work with clear instructions, but critical outputs need human review.

1. Always review outputs -- reports, contracts, and important documents need human verification.

2. Rewind is irreversible -- Esc Esc rollback permanently deletes later changes. Copy files first.

3. Permission prompts -- Claude Code asks before modifying files. Inconvenient but protective.

4. Account suspension risk -- stay within official terms of service.

Frequently Asked Questions

Q. Do I need to learn coding to use Claude Code?

A. No. Claude Code accepts natural language instructions. You need to open a terminal and understand basic folder structures, but no coding knowledge is required.

Q. How is Claude Code different from ChatGPT or the regular Claude app?

A. ChatGPT and the Claude app operate in a chat window. Claude Code runs in your terminal and can directly read, modify, create, and delete files on your computer.

Q. Is my company data safe when using Claude Code?

A. Claude Code processes through Anthropic's servers. Check your company's security policy first and start with test data.

Open the Terminal. Just Once.

The terminal screen is no longer black and intimidating. Type claude and a conversation window appears. Type "organize these files for me" in plain English and it works.

You did not learn to code. You learned to talk to AI from a different window.

If Part 1 introduced Cowork, Claude Code is the next door. You do not need to stand in front of it for long. Open a terminal today. Just once.

GoodTech AI

Technology for human good

Silicon Valley Parents Are Dropping Coding for Philosophy— What AI-Era Talent Really Looks Like

GoodTechAdviser — Tue, 31 Mar 2026 09:01:05 +0900

HR & Organizational Culture

Silicon Valley Parents Are Dropping Coding for Philosophy
— What AI-Era Talent Really Looks Like

GoodTech AI · March 15, 2026 · 7 min read

Table of Contents

1. The Liberal Arts Revival: Why Now?
2. From Adjectives to Verbs: Rethinking Performance Evaluation
3. Middle Managers in the AI Agent Era: Most at Risk?
4. Three Things Every Worker Should Prepare Now
5. Frequently Asked Questions

In early 2026, McKinsey quietly changed its hiring criteria.

Where computer science and data science graduates once had a clear advantage, the consulting giant started actively recruiting philosophy, history, and literature majors. (Fortune, 2026) Around the same time, something equally telling happened in Silicon Valley: tech-industry parents began pulling their children out of coding camps — and enrolling them in philosophy courses.

This isn't a passing trend. It's a structural shift, and Korean HR professor Hwang Seonghyeon (formerly Google HR and Kakao's Chief People Officer, now at Gachon University) offers a sharp explanation:

"AI now handles the 'How.' What's left for humans is the 'What' and the 'Why.'"

As someone who has spent nearly two decades in HR at a Korean manufacturing company, I've watched this transformation unfold in real time — and the implications for everyday workers are bigger than most people realize.

The Liberal Arts Revival: Why Now?

Just two years ago, the career advice in Silicon Valley was simple: learn to code. Parents debated which bootcamp would give their children the best edge. But a fundamental question has changed. It's no longer "Can my child write code?" — it's "Can my child decide what should be built and why?"

Professor Hwang points to a clear logic behind this shift. When AI can write code, analyze data, and even generate designs, the human role moves upstream — to problem definition. And problem definition requires the kind of thinking that liberal arts education cultivates: critical analysis, ethical reasoning, and the ability to ask the right questions.

The numbers back this up. PwC's 2025 Global AI Jobs Barometer shows that formal degree requirements are declining across AI-related roles. McKinsey's CEO has predicted that "within 18 months, every employee will collaborate with at least one AI agent." (Fortune, 2026)

When AI handles execution, humans must handle direction. And direction requires context-reading ability — a fundamentally humanistic skill.

Here's what this looks like in practice. When I recently built a workflow automation tool using Claude Code, the AI handled the coding flawlessly. But the hardest part — and the part that determined whether the tool actually solved the problem — was defining exactly what the problem was. The quality of my question shaped the quality of the AI's output entirely.

From Adjectives to Verbs: Rethinking Performance Evaluation

Every evaluation season in Korean companies follows a familiar pattern. Managers write phrases like "diligent," "good team player," and "needs improvement." Employees read these assessments — and have no idea what to actually change.

Professor Hwang calls this the adjective trap. His proposed solution is disarmingly simple: switch to verbs.

He uses a baseball coaching analogy. Telling a batter "your swing looks good" (adjective) changes nothing. Telling them "lower your swing angle by 15 degrees" (verb + number) creates real improvement.

This shift becomes even more critical in the AI era. To give effective instructions to an AI agent, you need specificity. "Write a good report" produces mediocre results. "Extract all items with year-over-year revenue decline exceeding 10% from Q1 data and analyze three root causes" produces actionable output.

The language of human feedback and AI instruction is converging — toward verbs and numbers.

Type	Adjective-Based (Old)	Verb-Based (New)
Positive feedback	"Responsible and diligent"	"Shared draft 3 days before deadline and incorporated team input"
Improvement feedback	"Poor communication"	"Summarize progress in under 2 minutes at weekly meetings"
AI instruction	"Make a good report"	"Extract items with >10% YoY decline from March data"

A 2025 survey of 206 Korean HRD professionals confirms this trend. While AI proficiency ranked first (69.9%) among skills to strengthen in 2026, communication (58.6%) and problem-solving (37.6%) followed closely. (HRD Article, 2025) Knowing how to use AI is table stakes. Knowing how to speak to AI — and to people — in precise, actionable language is the differentiator.

Middle Managers in the AI Agent Era: Most at Risk?

At the HR EXChange 2026 conference, Professor Hwang presented a projection that should make every middle manager pay attention:

Human-to-AI work ratio: Currently 9:1 → Within 3 years, 5:5 → Eventually, 1:9.

Today, humans handle 90% of work while AI assists with 10%. But as AI agents mature, this ratio inverts. Machines will process 90%; humans will handle 10%. The critical question becomes: what does that 10% consist of?

According to Professor Hwang, it's setting direction and assigning meaning.

He frames AI development in five stages:

Stage	Name	Description
1	Chatbots	Answering questions (where most organizations are today)
2	Reasoners	Solving complex problems step by step
3	Agents	Making autonomous decisions and taking action (entering now)
4	Innovators	Creating genuinely new solutions
5	Organizations	AI operating as organizational units

At Stage 3, organizational structures change fundamentally. When an AI agent reporting directly to the CEO can process the workload of hundreds of employees, the information-relay function of middle management structurally disappears.

Gloat's 2026 report confirms this reality: 40% of roles at Global 2000 companies already require collaboration with AI agents, and AI agent-related job postings surged 986% in a single year.

But here's the nuance that matters: management isn't disappearing — it's transforming. McKinsey Global Institute found that people management skills are now the fastest-growing skill demand, newly required across 138 job categories. (McKinsey MGI, 2025)

The manager who relays information will be replaced. The manager who designs human-AI collaboration will thrive.

Three Things Every Worker Should Prepare Now

Distilling Professor Hwang's insights into practical advice for working professionals, three priorities emerge:

1. Practice defining problems before solving them.

Before telling AI "do this," spend five minutes asking yourself: "What is the actual problem I'm trying to solve?" When I use Claude Code, I've learned that a precisely defined problem produces dramatically better results than a vague request. The quality of your question determines the quality of AI's answer.

2. Replace adjectives with verbs in your daily language.

When giving feedback to a colleague, try replacing "good job" with "the trend line you added on page 3 made the quarterly comparison much clearer." When instructing AI, replace "make it better" with specific action verbs and measurable criteria. This habit compounds — in both human and AI interactions.

3. Become a context provider.

Professor Hwang defines the future human role as "context provider." AI processes data, but humans assign meaning to that data. "Why does this number matter?" "Who needs to see this, and in what order?" — the ability to read context and provide it to both AI and colleagues is the emerging core competency.

A survey by the Korea Chamber of Commerce found that while 69.2% of Korean companies now consider AI skills in hiring, communication and collaboration (55.4%) and domain expertise (54.9%) rank immediately after. (KCCI, 2025) AI literacy is becoming baseline; the ability to understand people and read context is becoming the differentiator.

Frequently Asked Questions

Q. Do I need to learn coding to stay relevant in the AI era?

Not necessarily. As Professor Hwang argues, AI increasingly handles the "How" — including coding. What matters more is the ability to define problems clearly (the "What") and understand why they matter (the "Why"). Focus on developing critical thinking, communication precision, and context-reading skills alongside basic AI literacy.

Q. Are middle managers really going to be replaced by AI agents?

The information-relay function of middle management is at risk, but management itself is evolving rather than disappearing. McKinsey's research shows people management skills are actually the fastest-growing demand area. Managers who can design human-AI collaboration workflows and provide strategic context will remain essential.

Q. What's one practical thing I can do starting tomorrow?

Before giving any instruction to AI (or a colleague), pause and ask: "What exactly is the problem I'm trying to solve?" Then phrase your request using specific verbs and numbers rather than vague adjectives. This single habit — precision in problem definition — is the foundation of every skill discussed in this article.

Final Thoughts

One sentence from Professor Hwang has stayed with me:

"Technical implementation is moving to AI's domain. What remains for humans is the capacity for thought — deciding what to create."

You don't need to master coding. You don't need to memorize Excel formulas. AI will handle those things. But asking "Why does this report matter?" and "Who does this project serve?" — that's still a human job.

Tomorrow, before you ask AI to do something, try one thing: pause and ask yourself, "What exactly is the problem I'm solving?" That single question can transform the output you get.

Tools are getting smarter every day. May the questions we ask grow a little deeper, too.

GoodTech AI

Technology for the benefit of people

Claude AI for Professionals: A Step-by-Step Guide from Signup to Cowork (Part 1)

GoodTechAdviser — Mon, 30 Mar 2026 09:30:19 +0900

GoodTech AI · 6 min read

1. What Is Claude AI and How Is It Different from ChatGPT?
2. Claude Free vs. Paid Plans
3. How to Write Effective Prompts for Claude
4. Claude Projects: Your Persistent AI Workspace
5. Claude Cowork: The 2026 Agent Feature
6. Your Learning Roadmap
7. FAQ

Claude AI step-by-step guide for professionals

What Is Claude AI and How Is It Different from ChatGPT?

When a colleague casually mentioned "I drafted the report with Claude" during a team meeting, I felt a jolt of surprise. After 18 years in HR, I had been using ChatGPT on and off, but Claude was just a name I had heard. "It's probably the same thing," I thought. "And where would I even start?"

As it turned out -- it was much easier than I expected.

Claude is an AI assistant built by Anthropic, a US-based AI safety company. While ChatGPT excels at general-purpose conversational search, Claude stands out in long document analysis and writing quality. It can process over 150,000 words in a single session, meaning you can feed it an entire policy manual and get a coherent summary back.

Most common feedback from professionals: "The writing feels like a human wrote it" and "It hallucinates less than other tools" (Source: Gamsgo). As of August 2025, Claude reached 160 million monthly visitors -- a 10x increase in one year. 60% of Fortune 500 companies have integrated it into workflows (Source: Backlinko).

Claude AI learning stages isometric illustration

1. Go to claude.ai
2. Click "Sign up" and register with Google or email
3. Start chatting immediately after signup

No app installation required. Mobile apps (iOS/Android) exist, but the desktop browser is better for getting started.

Claude Free vs. Paid Plans: Which One Do You Need?

Plan	Price	Key Features	Best For
Free	$0	50-80 msgs/day, Sonnet 4	First-time users
Pro	$20/mo	5x capacity, Opus 4.6, Projects, Cowork, Code	Daily professional use
Max 5x	$100/mo	5x Pro capacity	Heavy users
Max 20x	$200/mo	20x Pro capacity	Power users

If you use AI 30-60 minutes daily, the free plan works. For 2+ hours daily or Projects/Cowork/Code access, Pro at $20/month is the practical choice.

How to Write Effective Prompts for Claude

Professional using Claude AI for workplace tasks

The single biggest change when I started using Claude was how I asked questions. "Summarize this document" produces average results. Restructure like this:

Basic structure: [Context] + [Task] + [Constraints] + [Format]

"You are a senior HR specialist with 10 years of experience. Read the attached HR policy document and summarize the top 3 changes. Write it as an email for director-level readers."

Claude-specific tip: Use XML tags

Claude responds with particular precision when you use XML structure:

<context>Manufacturing company HR team, 300 employees</context>
<task>Draft 2025 performance evaluation criteria</task>
<format>Organize as a table by category, A4 one-page length</format>

It feels awkward at first, but once learned, it applies to virtually every work task. In Korea, I use this for performance evaluations, job descriptions, and interview question sets -- tasks that used to take twice as long by hand.

Claude Projects: Your Persistent AI Workspace

Projects are independent workspaces organized by task. Upload documents once, and you never re-explain context.

Setup: Left sidebar → "+ New Project" → Name it → Upload documents → Start chatting inside the project.

I created a "2025 Performance Review" project with MBO criteria and competency frameworks. Now "Summarize what changed from last year" just works. In Korean manufacturing HR, where specialized terminology matters, pre-setting context produces noticeably more accurate results.

Projects require Pro or higher subscription.

Claude AI 4-step learning roadmap infographic

Claude Cowork: The 2026 Agent Feature Explained

Launched January 2026, Cowork is Claude's agent capability -- it directly reads, writes, and executes files on your computer.

Feature	Description	Use Case
File handling	Browse, read, write files	"Summarize all Excel files in this folder"
Multi-step automation	Sequential task execution	"Clean data, write report, save"
Browser control	Web search, scraping	"Find latest competitor press releases"
Office creation	Generate Excel/PPT	"Create a PPT from this data"

Windows version available with Mac feature parity. Install Claude Desktop with Pro or higher to start.

Note: Cowork is in Research Preview. Features may change before official release.

Your Learning Roadmap: Follow This Order

Claude AI 4-step learning roadmap infographic

Step	Timeline	Goal	Key Action
1	1-2 days	Sign up & first chat	Summarize one work document
2	3-7 days	Prompt basics	Practice [Context]+[Task]+[Constraints]+[Format]
3	1-2 weeks	Projects & Cowork	Create task-specific projects
4	Part 2	Claude Code	Terminal basics, Node.js install

Free courses: Anthropic's official AI Fluency and Claude Code certifications at anthropic.com/learn. Coursera's "Intro to Claude AI" is also excellent for beginners.

Frequently Asked Questions

Q. Is Claude better than ChatGPT for workplace writing?

A. Claude excels at long-form document analysis and natural-sounding writing. ChatGPT is stronger for general search and has a larger plugin ecosystem. Many Korean professionals use both -- Claude for drafting, ChatGPT for brainstorming.

Q. Do I need to know English to use Claude effectively?

A. No. Claude handles Korean, Japanese, Spanish, and many other languages well. Most Korean professionals interact entirely in their native language.

Q. What should I prepare for Claude Code (Part 2)?

A. Install Node.js 18+, Git, subscribe to Claude Pro, and review basic terminal commands. The black terminal screen does not mean you need to become a developer.

One Step at a Time

You do not need to master everything at once. Go to claude.ai today and throw one work task at it. "Summarize this document" is enough.

In Part 2, we cover Claude Code -- automating work from the terminal without writing code. Get comfortable with Claude.ai first, and Part 2 will feel natural.

One step at a time.

#ClaudeAI#ClaudeGuide#AIForProfessionals#ClaudeProjects#ClaudeCowork#Anthropic#AIProductivity#WorkplaceAI#PromptEngineering

GoodTech AI

Technology for human good