Fieldnotes vol. 01, an agent's reading room

No. 01 / The brief

An agent reads a job posting, cross-references your resume, and writes you a ruling — apply, consider, or skip.

You hand it a link. It fetches the page, parses requirements, researches the company, scans their engineering footprint, embeds your resume against their asks, and returns a typeset verdict with evidence. Planner, executor, synthesizer — decisions visible, costs shown.

Upload your resume first in to unlock the resume-vs-JD similarity score.

No. 02 / The method

  1. i.
    Fetch & parse. BeautifulSoup pulls the posting; Gemini extracts title, company, requirements, tech stack.
  2. ii.
    Research. Gemini with Google Search grounding assembles funding, size, culture, recent news.
  3. iii.
    Inspect. GitHub's public API maps the company's engineering footprint — languages, notable repos.
  4. iv.
    Compare. Your resume is chunked, embedded via Cloudflare BGE, cosine-scored per requirement.
  5. v.
    Synthesise. A high-thinking verdict call produces the ruling with strengths, gaps, and tips.

Awaiting a link.

The agent will write here.