Episode 28 | Programmer Unemployment Anxiety? Built an AI Job-Application System with Claude Code and Open-Sourced It

Image Analysis

Image 1 — Facts (only records what is directly visible)

- Readable text (top-to-bottom, large-to-small):

- Top large white text: 网站推荐 (Website recommendation)

- Directly below, yellow text (near top center): 用AI筛公司 (Use AI to screen companies)

- Middle small yellow text: vs

- Directly below, yellow text: 公司用AI筛你 (Companies use AI to screen you)

- Mid-image white text with black outline overlaying rabbit image: 我也要选择公司呀！(I also want to choose the company!) — the last character resembles “呀”; see "unreadable content"

- Image marks:

- Top-right shows a black-and-white GitHub logo (octocat inside a rounded square)

- Interface / screen elements (visible but no readable numbers):

- Right monitor shows an interface containing multiple colored blocks / block matrix (multi-color small squares); to the left or right of that area are several thumbnails or columns (positions and arrangement not fully distinguishable)

- Right monitor also shows a small rabbit avatar thumbnail (near the colored blocks area)

- Left monitor shows a dark-background window, appearing to be a code editor or application interface (window panes and lines of text visible, but text unreadable)

- Visible objects and scene:

- Foreground: a close-up of a realistic/cartoon rabbit face (round face, upright ears, fur texture), expression with squinted eyes, eyebrows lowered, mouth tight — conveying displeasure or dissatisfaction

- Background: a desk setup with keyboard, mouse, desk cables, two monitors and blurred office environment

- No numerical chart data readable: the image contains no readable values, scales, or chart labels — only colored blocks and thumbnails

Unreadable or blurred content (explicitly noted):

- Small text on the screens (both left and right monitors) is unreadable

- Labels or explanatory text around the colored blocks on the right screen are unreadable

- Any small text or icons beneath the rabbit image or on the desk are not identifiable due to resolution/blur

Inferences (marked as inferences, not asserted facts):

- The header text appears to contrast two scenarios — “website uses AI to recommend companies” vs “companies use AI to screen you” — this is an interpretation of semantics, not a step shown in the image

- The right-screen UI is inferred to be some design tool, dashboard, or product showcase (based on colored block and thumbnail layout)

- The rabbit’s expression is inferred to communicate displeasure/complaint (subjective judgment)

Ignored content note:

SkillPI Creation Analysis

Purpose of the piece:

- The source (a Douyin/TikTok video title and screenshot) presents a visual and topical demonstration around “AI in job search / screening.” The title claims the author built an AI job-application system using Claude Code; this indicates an intent to show or share a project that applies Claude Code to recruitment/ job-matching workflows. The exact implementation and repository details must be verified. Source URL: https://v.douyin.com/6CZfkeFoBfY/ .

Image evidence (direct, supportable observations):

- The header text juxtaposition (“Use AI to screen companies” vs “Companies use AI to screen you”) supports a theme of two-way filtering/recommendation.

- A GitHub logo in the top-right suggests an association with a code repository or open-source release.

- Dual monitors, one resembling a code editor and the other a colored-block visualization, indicate the project likely combines code development and a front-end/dashboard for visualization.

- The anthropomorphic rabbit face conveys emotional context (anxiety or displeasure), aligning with the title’s framing of programmer anxiety.

Related skills and stepwise approach (general reproducible workflow based on title/image; specific APIs/commands must be confirmed against actual project docs or Claude Code docs):

1. Define scope and data: decide features (resume parsing, job description indexing, matching/scoring, personalized recommendations) and prepare datasets (resumes, job postings) with privacy safeguards.

2. Design matching logic: choose ranking strategy (keyword matching, semantic similarity, rule-based filters, weighting).

3. Prototype in Claude Code: implement core orchestration or prompt-based logic in Claude Code (text preprocessing, feature extraction, calling models for semantic matching or rationale generation). Verify Claude Code’s supported features and API usage in its documentation.

4. Embeddings and retrieval: convert text to embeddings and store in a vector store for semantic search if needed.

5. Backend/frontend integration: implement matching endpoints and a UI/dashboard for candidates and company views (the colored blocks may represent tags, clusters, or visual summaries).

6. Evaluation: run offline/online evaluations (relevance metrics, human review) and iterate on prompts, thresholds, and weighting.

7. Open-source and deploy: prepare documentation, examples, and deployment instructions; publish repository (e.g., on GitHub) and include deployment scripts or demo.

Prerequisites to reproduce:

- Access to the project source repository and documentation (if the goal is faithful reproduction).

- Confirmation of Claude Code capabilities and how it is used in the project (APIs, runtime environment, model variants).

- Datasets for matching (resumes and job descriptions) or synthetic/anonymized alternatives, plus compliance with privacy/legal requirements.

- Infrastructure for embeddings/vector search and sufficient compute to run embedding/model calls.

Items to verify before claiming equivalence or reuse:

- Whether the project’s source code is publicly available and under what license (the GitHub logo is suggestive but not definitive).

- Exact role of Claude Code in the implementation (what parts it handles, whether it’s orchestration, prompt execution, code generation, etc.).

- Data provenance and privacy handling for any resume or candidate data used.

- Dependencies on third-party APIs or paid services and the related costs.

- Any performance metrics or user-evaluation results — the title/screenshot do not provide verifiable performance data.

Notes and limitations:

- This interpretation is based solely on the provided screenshot and title; it does not access the full video or repository. Implementation specifics and claims of reproducibility require examining the actual code or project documentation.