The Spray-and-Pray Ceiling
Most job hunts are a numbers game played badly. You skim a portal, paste the same CV at fifty roles, lose track of which ones, double-apply at two of them, and hear nothing back. The spreadsheet you keep forgetting to update is not a system. It is a guilt file.
Two bottlenecks cause this. Judgment, which roles are worth your week. Memory, never forgetting and never applying twice. Double-applying signals the back room you are doing something odd, and you get binned. The fix is not a better spreadsheet. It is a filter that judges, and a memory that never forgets.
Two Files: Who You Are, How You Decide
Stop thinking of AI as a thing that writes your cover letter. Start thinking of it as a recruiter you hired. The first week they are useless, because they do not know you. Feed them two files and within a week they read a hundred postings for you, rank the five worth your time, flag the ghost jobs, and never forget.
The Copy-Paste Pack
The two starter files, the shortcut that fills them for you, and the decision dial the evaluator returns.
The cv.md Starter Template
[Your Name]
[City, Country] | email | phone | linkedin | github | work permit status
## Summary
[Two to four lines. Years of experience, the intersection you work at, the roles you want, the roles you are moving away from. One sentence of philosophy, optional.]
## Experience
### [Company] - [Role] (Month Year to Month Year, Location)
- [Bullet with a real metric.]
- [Bullet with a real metric.]
[repeat per role]
## Skills
[Grouped: languages, frameworks, platforms, methods.]The CLAUDE.md Decision Rules
[One line: this is my job-hunt command center, a filter that tells me which postings are worth my time and a memory that never forgets.]
## Scoring rubric
Every role gets a score from 1 to 5 across five dimensions:
1. Match with my CV
2. Fit with the roles I actually want
3. The compensation signal
4. Culture and red flags
5. Posting legitimacy (a real opening, or a ghost job)
## Dealbreakers
- No B2B-only contracts, I am a contractor.
- [add yours: remote only, won't relocate, salary floor, etc.]
## Guardrail
The AI never auto-submits anything. It evaluates and recommends. I decide and act.The "Interview Me" Prompt
Rather than fill cv.md field by field, hand the model the empty file and let it interview you. Paste this into Claude Code or Claude.ai in a folder that holds your cv.md:
APPLY, SKIP, or TAILOR
A dealbreaker in the posting forces SKIP regardless of score. Otherwise the call follows the score and the gaps. Each result appends one row to tracker.md: date, company, role, score, call, link.
Keep It Sharp
Every role you reject for a reason you forgot to write down is a missing line in CLAUDE.md.
Every Rejection Is a Missing Line
You will skim a posting, feel "no," and move on. That feeling is data. Ask yourself what triggered it, then add the line to CLAUDE.md so the next hundred postings get checked automatically.
Run this loop every time and within a month the evaluator is filtering in your judgment, against your dealbreakers, without you re-reading a single posting. That is the shift from applicant to operator.
Do One Thing After This
Open Claude Code in an empty folder. Create cv.md with the starter above. Create CLAUDE.md with your rubric, your dealbreakers, and the guardrail. Then paste a real job posting and ask it to evaluate. Watch it score, flag the dealbreakers, and tell you APPLY, SKIP, or TAILOR. That is how tonight becomes real. Not by reading another posting. By filtering.
The AI recommends. You decide. It never auto-submits.