Table of Contents
Artificial Intelligence Translates the Language of Animals
Kim Kyung-jin, Attorney at Law
The Story of AI Learning to Listen to Whales, Dolphins, Birds, and Bees
Twelve chapters on how AI listens to dolphins, sperm whales, humpback whales, birds, and bees to find rules in their sounds, and what this technology means for its risks and for animal rights. Written in simple sentences a child can read, with verified sources in every section.
Table of Contents
AI Deciphers Ancient Scripts
Kim Kyung-jin, Attorney at Law
Ancient Records Revived by AI
In twelve chapters, this book explains how AI revives records once unreadable, from burned scrolls and wooden slips buried in mud to broken clay tablets. It covers virtual unrolling at Herculaneum, virtual collation of oracle-bone texts, reading Silla wooden tablets, computational analysis of undeciphered scripts, and multispectral archives, with verified references for each chapter.
Table of Contents
Artificial Intelligence for New Materials Design and Rocket Propulsion Engineering
Kim Kyung-jin, Attorney at Law
AI Potentials, Self-Driving Laboratories, and Physics-Informed Machine Learning (PIML)
Ten chapters on how artificial intelligence is changing new materials and rocket propulsion: atomic simulation, generative models, self-driving labs, high-temperature alloys, metal 3D printing, combustion, cooling design, and engine diagnosis and control. Written without equations, with verified sources in every chapter.
AI Library
The Age of Autonomous Scientific Discovery
Kim Kyung-jin, Attorney at Law
AI Scientists and Self-Driving Labs
This book follows how AI scientists and self-driving labs are changing the way science generates and verifies claims. It covers literature-based discovery, natural-language protocols translated into robot commands, multi-agent research systems, closed-loop laboratories, materials search, the verification gap, chains of evidence, research harnesses, journal ethics, and legal responsibility.
AI Library
A New Era of Life Sciences Opened by Artificial Intelligence
Structural Proteomics, Genomic Foundation Models, Autonomous Laboratories, and Global Governance
Kim Kyung-jin, Attorney at Law
This book is a research volume compiled with artificial intelligence. A human selected the materials and structured the work, while AI models drafted the sentences and cross-checked the facts.
AI Library
The Double Structure of Digital Sovereignty
Europe’s Departure from Palantir and the Chains of American Big Tech
Kim Kyung-jin, Attorney at Law
This is a record of 2026, when European intelligence agencies and defense ministries began removing analytics tools from America’s Palantir. It covers the replacement decisions made by France’s General Directorate for Internal Security (DGSI), Germany’s Federal Office for the Protection of the Constitution (BfV), and the Netherlands Ministry of Defense; the incident in which US export controls severed an ally’s ac…
New English Edition
Artificial Intelligence in Horticulture
Kim Kyung-jin, Attorney at Law
Across five chapters and ten sections, this book examines computer vision for crop diagnosis, harvesting robots and autonomous field systems, smart greenhouses and digital twins, precision irrigation and supply-chain quality control, high-throughput phenotyping, and predictive breeding.
New English Edition
Artificial Intelligence in Food Crop Agriculture
Kim Kyung-jin, Attorney at Law
Across six chapters and eighteen sections, the book examines digital agricultural infrastructure, remote sensing, crop diagnosis, yield forecasting, precision irrigation, genomics, molecular breeding, agricultural robotics, climate-smart agriculture, and global food security.
New English Edition
The Future of Forestry and Agroforestry
Kim Kyung-jin, Attorney at Law
Driven by Artificial Intelligence and Digital Innovation
Across five chapters and fifteen sections, the book follows satellites, drones, LiDAR, digital twins, forest-specific language models, wildfire and pest forecasting, forestry robotics, agroforestry, timber traceability, and forest carbon markets.
New English Edition
Smart Livestock Farming: AI Enters the Barn
Kim Kyung-jin, Attorney at Law
Sensors listen, cameras watch, and artificial intelligence helps farmers decide.
Across five chapters and fifteen sections, the book follows precision livestock farming from animal health and reproduction to robotic milking, virtual fencing, digital twins, methane reduction, welfare, and data ownership.
Table of Contents
Han Dong-hoon, Busan Buk-gu Gap: A Record of the 100 Days Before and After the Election (Mar. 26-Jul. 3, 2026)
Kim Kyung-jin
Table of Contents and 13 sections
From March 26 to July 3, 2026, this record follows the spring after expulsion, the Busan Buk-gu Gap by-election, victory as an independent, and the first bill submitted in the National Assembly.
Table of Contents
Artificial Intelligence and Medicine
Kim Kyung-jin, Attorney at Law
AI in clinical care, hospitals, education, and research
AI in medical imaging, risk prediction, treatment planning, hospital operations, education, and research, with patient safety, privacy, and accountability.
[AI Library] Chapter 18. Getting Big Work Done Overnight
Leaving It to AI and Stepping Away - A Complete Beginner's Guide to YOLO Mode
Chapter 18. Getting Big Work Done Overnight
Kim Kyung-jin
Case 17) Restructuring 100 Code Files Overnight
The Scene
Wednesday, 11 p.m. Jang Seo-jun, the development team lead, sat in front of his monitor, turning a problem over in his head. The company's web service codebase had gotten old. The same pattern was repeated across 100 files, and the whole thing needed to be migrated to a new structure. Done by hand, that's a week of work. Open each file one by one, edit the code, run the tests.
Jang Seo-jun decided to try a different approach: hand it off to AI and go to bed.
The Command
First, he created a new branch in Git.
git checkout -b refactor-structure
The idea was to leave the original code untouched and do all the work on a separate branch. If something went wrong, he could just delete the branch and walk away.
He opened Claude Code.
"In all .ts files inside the src folder, replace the old-style import syntax with the new style. Specifically, change 'import { X } from '../utils/helpers'' to 'import { X } from '@/utils/helpers''. After each change, check whether the file compiles without errors. If any file fails to compile, revert it to the original and give me a list of those files."
How It Went
Claude Code scanned the src folder. 112 .ts files. It opened each one, found the pattern, and made the replacement. After each change, it ran the TypeScript compiler to check for errors.
When a file threw an error, Claude Code rolled it back and added it to a list marked "requires manual review."
Processing all 112 files took 45 minutes to an hour. Jang Seo-jun gave Claude Code its instructions and went to sleep.
The Result
Thursday morning, 8 a.m. Jang Seo-jun arrived at the office and checked the terminal. A message was waiting: "108 of 112 files converted successfully. 4 files require manual review."
He ran git diff to see what had changed. The import statements in 108 files had been updated to the new style. The 4 remaining files came with a note explaining that converting them had caused errors elsewhere, so they had been left as-is.
Jang Seo-jun reviewed and fixed the 4 files himself. Thirty minutes, done. What would have taken a week was finished in a single night.
The Cost
1 hour of AI work. Cost: roughly 10 to 17 USD. Had a person done the same work over a week, the labor cost would have run into the millions of won.
The Risk
Having AI modify code at scale overnight carries real risk. That's why the three things Jang Seo-jun did matter.
He created a branch. The original code was never touched. If anything had gone wrong, deleting the branch would have been enough.
He built in a compile check. The instruction wasn't just "make the changes" — it was "make the changes, then verify them."
He told the AI to revert any file that failed. So the AI couldn't force a broken fix through.
If any one of these three was missing, you might have walked into the office that morning to find the codebase in ruins.
How to do the same
When handing off large-scale code changes overnight, always work on a new branch. Include "confirm" and "revert on failure" in the instructions. In the morning, run git diff to review what changed, run your tests, then merge into the original branch.
Case 18) Auto-detect an outage at dawn, auto-fix it, and open a PR
The scene
3:17 a.m. The web service goes down. Datadog, the server monitoring tool, catches a spike in the error rate and fires a Slack alert. Normally, the on-call developer wakes up, opens a laptop, checks the logs, patches the code, and ships a fix — an hour gone. While all that is happening, users are sending messages saying the site is down.
This company does it differently. They built a pipeline — an automated workflow — that spins up Claude Code automatically whenever a Slack alert comes in.
The instruction
No one is issuing commands in real time. This is all pre-configured automation. When a Slack outage alert arrives, a webhook — a connection that triggers automatically when a specific event occurs — fires a script, and the script starts Claude Code.
The instruction passed to Claude Code looks like this.
"Analyze the error logs. Find the cause in the error messages and check the related code files. If you can fix it, create a new branch, apply the fix, and open a PR (Pull Request) on GitHub. Prefix the PR title with '[Auto-fix]'. If you can't fix it, post an error analysis report to Slack."
How it played out
3:17 a.m. Claude Code starts automatically. It reads the error logs. The same error keeps repeating: "TypeError: Cannot read property 'email' of null". Something is returning null — an empty value — when it tries to fetch user data.
It tracks down the relevant code file. Line 42 of user-profile.ts reads the user's email, but the code has no handling for the case where user data doesn't exist.
Claude Code creates a new branch and adds a null check — code that verifies whether a value is empty — to line 42. Compilation check. Tests run. All pass.
A PR goes up on GitHub. PR title: "[Auto-fix] Add null check in user-profile.ts"
A message goes to Slack. "Outage detected at 3:17 a.m. Cause: unhandled null in user-profile.ts. Fix PR submitted. Please review and merge."
3:32 a.m. Fifteen minutes from detection to PR.
The result
The on-call developer comes in the next morning, reviews the PR, checks that the code is correct, and merges it. The incident was resolved in 15 minutes, and nobody had to wake up in the middle of the night.
The cost
15 minutes of automated work, roughly $2 to $4. Far cheaper than paying an on-call developer overnight rates.
The risk
An auto-generated fix can be wrong. So this company put two safety mechanisms in place.
It only opens a PR — it doesn't deploy automatically. A person has to review and merge it before anything ships.
PR titles are prefixed with "[Auto-fix]" to distinguish them from human-created PRs. Auto-fix PRs get reviewed more carefully.
What if you had set it to deploy automatically too? The AI could push a bad fix and deploy it on its own, making the incident worse.
How to try this yourself
This setup works when you have a development team of some size and are already using Slack and GitHub. You need to build the pipeline ahead of time: monitoring tool (Datadog, Sentry, etc.) → Slack alert → webhook → Claude runs → PR created. You can have Claude Code build the pipeline itself. Just tell it: "Set up an automated incident-response pipeline for overnight outages."
Case 19) Running 20 sub-agents in parallel to review the codebase
The scene
Friday at 6 PM. A major update is scheduled to ship Monday, and you want to do a full sweep of the code before then. The project has over 2,000 files. One person doing it manually would take two weeks.
The command
Claude Code has a feature called sub-agents. One Claude Code instance can spawn multiple smaller Claude Code instances and run them all at the same time.
"Split the src folder of this project into 20 sections, then run one sub-agent per section. Each sub-agent should inspect the files in its assigned section and look for: unused variables, missing error handling, and security-vulnerable patterns (hardcoded passwords, SQL injection risks). Collect all findings in code_review_results.txt."
How it runs
Claude Code divides the src folder into 20 sections, distributing files as evenly as possible. All 20 sub-agents start at the same time.
Each sub-agent reads through its assigned files and looks for problems. The sub-agents work independently of each other. If one slows down, the others aren't affected.
With 20 running in parallel, a task that would take one agent 5 minutes gets done 20 times faster. The full review takes 15 to 25 minutes.
The result
A file called code_review_results.txt, with findings organized by section. Entries like: "auth.ts line 23: password hardcoded", "db-query.ts line 55: potential SQL injection", "utils.ts: 3 unused variables". Dozens of items like these.
The cost
25 minutes of runtime, roughly $28–$49. Because 20 sub-agents hit the API simultaneously, the cost runs higher than other cases. Still trivial compared to what 20 people doing the same code review would cost.
The risk
If 20 sub-agents all try to edit files at the same time, you get conflicts. That's why this example instructs them to "inspect only — don't make any changes." Read-only work carries no conflict risk.
If you do want them to make fixes as well, each sub-agent should work on a separate branch.
Watch the costs. As the number of sub-agents grows, the bill grows with it. Running 50 instead of 20 means 2.5 times the cost. It's worth setting an API usage cap in advance.
How to try this yourself
The sub-agent feature is available starting with Claude Code 2026. It works well for large projects (500 or more files). For small projects, a single Claude instance is enough without sub-agents. Start with 3 to 5 sub-agents, and if the results are good, increase the number.
A failure story: the agent that repeated the same error 300 times overnight
One developer handed off a code refactoring job to AI at night and went to sleep. By morning, the terminal had more than 300 lines of error messages. The same error message, repeated 300 times.
Here's what happened. The AI modified the code, but the tests failed. The AI said "I'll fix it" and tried again. Failed again. Fixed again. Failed again. This went on 300 times.
A person would have stopped around the third attempt and thought, "My whole approach must be wrong." The AI didn't stop. Same fix, same failure, same fix again. On and on.
The cost was the real problem. Those 300 attempts ran up about $85 in API charges. In a single night. The code was right back where it started, and the money was gone.
There are two ways to prevent this.
Set a retry limit. Add something like "if the same error comes up 3 times, stop and report back" to your prompt. That one line would have saved $85.
Set an API spending cap. You can set a daily usage limit in the Anthropic dashboard. If you cap it at around $14, the API stops once you hit that amount. Even if the AI runs all night, it stops at $14.
[End of Part 5] Use cases at a glance
Case 1. 50 email reply drafts. Time: 3 minutes, Cost: $0.35.
Case 2. Meeting recording turned into meeting minutes. Time: 10 minutes, Cost: $1.40.
Case 3. Summary report from 10 research papers. Time: 30 minutes, Cost: $4.20.
Case 4. Markdown converted to print-ready PDF. Time: 8 minutes, Cost: $0.85.
Case 5. 12-month sales figures tallied. Time: 2 minutes, Cost: $0.28.
Case 6. Duplicate vendors identified. Time: 1 minute, Cost: $0.35.
Case 7. 200 receipts transferred to Excel. Time: 40 minutes, Cost: $3.50.
Case 8. YouTube comment sentiment analysis. Time: 8 minutes, Cost: $1.40.
Case 9. Personal blog built from scratch. Time: 30 minutes, Cost: $2.80.
Case 10. Neighborhood group schedule. Time: 10 minutes, Cost: $0.35.
Case 11. Flower shop booking system. Time: 1 hour, Cost: $4.55.
Case 12. Automated browser testing. Time: 10 minutes, Cost: $1.05.
Case 13. Gmail sorting and reply drafts. Time: 5 minutes, Cost: $1.40.
Case 14. Slack weekly summary. Time: 3 minutes, Cost: $0.56.
Case 15. Auto-updating a Notion page. Setup time: 10 minutes, then fully automatic.
Case 16. Tracking competitor websites. Setup time: 15 minutes, then fully automatic.
Case 17. Restructuring 100 code files. Time: 1 hour, cost: 20,000 won.
Case 18. Automated response to a late-night outage. Time: 15 minutes (automatic), cost: 4,000 won.
Case 19. Code review across 20 sub-agents. Time: 25 minutes, cost: 55,000 won.
One thing every case had in common: we committed to Git before starting, then checked the results with our own eyes.
Part 6. How to Survive When Things Go Wrong
Kim Kyung-jin
Lawyer · Former Member of the National Assembly · AI Policy Researcher
© 2026 Kim Kyung-jin. All rights reserved.