Yachen Li

AI Projects

Summary

AI tools I have built

Bili Clipper, a Chrome extension that saves Bilibili video subtitles to Obsidian or Notion in one click.

AI recruiting agent

I have not built this tool. I am currently working on the Opportunity Hunter Project, an early-stage exploration of how to help individuals discover opportunities.

My Three Stages with AI

01

Built a vibe-coding project

02

Realized I was producing AI slop, and started systematically learning how to work with AI

03

Now teaching myself the theory of how AI works underneath

01Bili Clipper

Problem

Obsidian's official Web Clipper can save YouTube videos as notes, but it does not support Bilibili, a major Chinese video platform. Many Bilibili videos have no transcript, so the only option was taking notes by hand.

Process · from problem to product

01
Get inspired by an existing product
Seeing what was already possible made me realize my problem could be solved with software.
02
Search and verify, instead of just trusting AI
The AI first said no such tool existed. My own search turned up similar projects, and I started comparing their approaches.
03
Build it with a coding agent
The agent wrote all of the code. I told it what I wanted, tried what it built, and pushed back when something seemed off, like its first setup plan:
AI's first plan
Extra Obsidian plugin + local API setup
→
Revised plan
Works with Obsidian, no extra plugin or local API
04
Ship, collect feedback, fix
I published it on the Chrome Web Store and open-sourced the code. Users then reported that multi-part videos always pulled subtitles from part one, so I fixed the bug and released an update.

Reflection

GitHub Chrome Web Store Read the full retrospective →

02Opportunity Hunter

Problem

Many AI tools already help people find opportunities such as jobs, yet most of my job-seeking classmates do not use them. Some have never heard of them, some cannot set them up, and some find them too complicated. There is a wide gap between what these tools can do and what ordinary users are actually able to use.

Goal

Help ordinary people use AI to discover and win opportunities they would otherwise never know about or reach.

Approach

Find the users first: find an audience or community → listen to what they need → collect feedback → build the simplest thing and let them try it → iterate on feedback → MVP

Start with people
Find an Audience / Community
Observe the problems job seekers and students actually run into
↓
Content & Education
Share articles, guides, real examples, and methods in public
↓
Lightweight Tools
Prompt → Template → Skill / Plugin
Test whether ordinary users can reuse them easily
↓
Community Feedback
Watch for needs, friction, real results, and whether people come back
↓
Potential AI Agent / Product
Turn only proven, valuable workflows into a product
A future possibility, not a current commitment
↻ Any stage can loop back based on feedback. The steps are not a fixed sequence.

Progress

Next Step

I plan to keep writing about different directions. Job hunting is a big topic, so I will use reader feedback to find one specific pain point: something people care about enough and can pick up easily. That is what I want to build a tool for.

Opportunity Hunter
A general framework for discovering opportunities
Career
Jobs and career opportunities
Current focus
Jobs
Public and hidden openings
Companies
Companies and teams worth following
People
Potential contacts, mentors, recruiters
Events
Industry events, talks, recruiting events
Communities
Communities and professional groups
Education
Study and academic opportunities
Later
Programs
Schools, majors, and degree programs
Research
Research, advisors, and lab openings
Funding
Scholarships and grants
Academic Events
Lectures, conferences, and exchanges
Projects & Programs
Competitions, fellowships, short programs
Career is where I am starting. Education follows the same logic and is a possible later extension, not yet a content plan.