Wednesday, August 12th at 8 PM · Online Zoom · Free
Host: Yang Eun-seong, Development Lead
Hello, I am Captain Developer Yang Eun-seong. I have some news for those who have taken the C++ Development Environment and Modern CMake course.
I run a company by myself. Since I have to repeatedly create the same types of work, I couldn't avoid using AI. However, while it seemed to do well when making the first one, the results kept changing with the second and third ones. So, I had to explain everything all over again every single time—who I am, what tone I should use, and what was disappointing about the last time.
I thought adding more information would work. I added background, attached materials, and even passed over all the previous conversations. But it was still the same. The problem was that I did it all with words every time. Words don't stick around. They disappear as soon as you open a new window.
The reason we save build settings as files is so that people don't have to remember and adjust them every time. The criteria we leave for AI follow the same principle. What to do first, where to stop, what to consider a success, and what the final output should look like. If you define those in a file, you get the same result every time.
In the end, what's truly important is this: It's not information that makes AI produce results. It's judgment and decision-making.
On the evening of August 12th, I will open that structure exactly as it is on the screen. It won't be a session for explaining with slides; instead, I'll open the actual files I use and run them right there using the transcript from last week's live session.
What we cover
Why does a different result come out every time you give the same repeated instructions? — Why it's not a matter of the amount of information
How the tasks once handled by programming languages have shifted to me
Standards · Procedures · Templates · Assets — What should be written down in advance so you don't have to explain it again?
Extracting different results from the same recording by simply changing the purpose — Improvements → Summary → Q&A (Demonstration)
How to stop humans from doing all the inspection — Programs handle the format, AI handles the content
Real-time Q&A — "Can this be applied to my work as well?"
For people like this
Those who end up re-checking the AI-generated results from scratch anyway
Those who repeatedly perform tasks with fixed formats, such as meeting minutes or reports
Those who find themselves repeating background explanations every time they start a new conversation
Those who use Claude Code or Cursor but get inconsistent results
Participants Only
1:1 Mentoring for 30 minutes — One person
Free access to the lecture covered in this live session — one person
Summary notes and the outputs generated by the demo that day — For everyone who fills out the response form
Apply
https://forms.gle/12sQjhxyzxpkuxcn9
Once you submit, the Zoom link will appear immediately and will also be sent to the email address you provided. Questions will only be accepted via Zoom.
What is being handled is documentation, not code. It is actually better suited for tasks with set processes and specific output formats, such as meeting minutes or reports. To follow along, you will need a Claude Code or Codex account, and those costs are separate.