Six resources worth following
You do not need forty newsletters. These are the ones Andrew actually uses, and each earns its place for a different reason.
Nate Herk
Hands-on builds of AI agents and automations. Watch him to see what "let the AI do the work" looks like in practice, then steal the patterns for your own capstone.
youtube.com/@nateherkEthan Mollick
Wharton professor writing One Useful Thing. The clearest thinking available on what AI actually changes about work, backed by studies instead of hype.
oneusefulthing.orgStart here: Management as an AI Superpower
Nate Jones
Daily breakdowns of AI news with a career lens. Good for staying current enough to talk credibly about AI in any interview, which is now every interview.
youtube.com/@NateBJonesNotebookLM
Google's grounded research notebook. It only answers from sources you give it, which makes it the perfect rehearsal room. Step 8 lives here.
notebooklm.google.comClaude Use Case Library
Anthropic's library of what people actually build and do with Claude. Don't browse it by hand: give Claude the link and ask it to search the site for your use case, and it returns the pattern that fits your situation.
claude.com/resources/use-casesBattlecards
A single page holding everything you know about one company and one role: the people, the product, the questions, your stories. Step 7 shows you Andrew's real one.
See the Salesforce battlecardTwo ideas that change how you apply
AI prefers AI writing
Most resumes are now read by AI before a human ever sees them, and here is the uncomfortable part: AI screeners score AI-shaped writing higher. Clean structure, standard headings, exact keyword matches from the posting, quantified outcomes. Hand-crafted quirky formatting that used to show personality now just parses badly.
Andrew wrote up what job seekers should know about this, including how to use it without lying about anything.
Write for the machine first, the human second. Both will thank you. Read: AI Resumes, What Job Seekers Should Know
Cover letters are out. Summary intros are in.
Nobody opens the second attachment. Recruiters spend seconds on your application, and a separate letter restating your resume is a relic. What replaced it is the summary intro: two lines at the top of your resume, and the same two lines as the opener of any outreach message, that do the entire cover letter's job. Who you are, the one proof point that matters for this role, why this role.
Two lines is the discipline. If your summary runs longer, you have not decided what your proof point is yet.
Skip the letter. Spend that hour making two perfect lines, then reuse them everywhere. Read: RIP, cover letters (Business Insider)
Eight steps, run chronologically
The order matters. People before postings, research before applications, artifacts before interviews. Each step hands its output to the next one.
Referrals are the way in, especially early
A referral moves your resume from a pile of hundreds to a stack of five. In the first weeks of a search, when you have the most energy and the least momentum, this is where that energy pays best. So before you polish a single pixel, write your tribe list: former colleagues, church, school cohorts, that group chat, LinkedIn first-degree connections. Everyone who already knows your work or your character.
The answer to "do I know anyone there?" is yes more often than people expect. Go look before you assume.
Artifact: your tribe list, names and where they workResearch every company on the tribe list
Now work the list. For each company where someone in your tribe works, have Claude do the deep read: what they build, how they make money, what they shipped lately, what they are struggling with. Research is cheap now. Do far more of it than feels normal, because everyone else is still skimming About pages.
The company research prompt
I'm job hunting. Research [Company] for me: what they do, how they make money, recent news and launches, team culture, who leads the team I'd likely join, and what they're struggling with that someone in a [your role type] seat could help fix. Check their careers page for open [role type] roles. Finish with two answers: is this company worth my time, and what should I say to my contact who works there?
Filter to companies with roles you actually want
Cross-reference the research against live openings. Keep only companies that have a real posted role you would genuinely take, and where the research came back warm. This is your shortlist, and it should be short. Five focused pursuits beat fifty sprayed applications, because every remaining step in this playbook is per-company work.
Artifact: shortlist with the exact JD for each roleTailor your resume to the JD
Your master resume is the template, the job description is the spec. Mirror the JD's own vocabulary, put your strongest matching evidence first, keep it to one page, and make the top two lines the summary intro that replaced your cover letter. There is a Claude Code skill for this now, it reads the JD, maps your evidence against every requirement, runs the verb and one-line audits, and tells you where your evidence is thin.
Everything you need to run it yourself is below: the skill file, Andrew's actual resume rules, and the install steps. Set it up once and every tailoring after that is one sentence: "Tailor my resume to this job description."
Best starting point: Build your own resume-tailoring system, one prompt that interviews you, builds your super resume and house rules, and leaves you with the whole tailoring loop.
Install the skill (one-time setup)
1. Make the folders: mkdir -p ~/.claude/skills/resume-jd-tailor 2. Save the skill text (next section) as: ~/.claude/skills/resume-jd-tailor/SKILL.md 3. Write your own house rules file at: ~/.claude/RESUME_RULES.md Start from Andrew's real rules (below) and swap his specifics, project names, links, job titles, for your own equivalents. 4. Open Claude Code, hand it your master resume and a JD, and say: "Tailor my resume to this job description."
The skill file, SKILL.md
--- name: resume-jd-tailor description: Take an existing resume as the template and tailor it to match a specific job description. Use when the user says "tailor my resume", "format my resume to this JD", "match my resume to this job", or provides a resume plus a job posting. Preserves the template's structure and styling, rewrites content to mirror the JD's language and priorities. --- # Resume JD Tailor Tailor a resume to a job description using the existing resume as the template. The template's layout, sections, and styling are the contract, only the content moves. ## Inputs 1. Template resume, the user's master resume. 2. Job description, pasted text, a file, or a URL to fetch. 3. House rules: ALWAYS read ~/.claude/RESUME_RULES.md first and follow every rule in it. Those rules override this skill. ## Process ### 1. Parse the JD - Extract the role title, team, and the top 6-8 requirements in the JD's own words. - Note the exact vocabulary the JD uses. Mirror their terms, recruiters and AI screeners match on them. ### 2. Map evidence - For each requirement, find the strongest matching bullet, project, or story in the template resume. - Flag requirements with thin evidence, tell the user rather than inventing experience. Never fabricate or relocate accomplishments to a different employer. ### 3. Rewrite the summary as the cover letter Cover letters are out, summary intros are in. The summary at the top of the resume does the cover letter's job: who you are, the one proof point that matters for THIS role, why this role. Max 2 lines. ### 4. Tailor the body - Reorder bullets so the strongest JD matches appear first. - Rewrite bullets to lead with the JD's language, staying truthful. - Keep it AI-screener friendly: standard section headings, plain formatting, exact keyword matches, quantified outcomes. ### 5. Audit before delivering - Verb audit: no repeated starting verbs across the whole resume. - Line audit: every bullet fits one line, resume fits one page. - Link audit: all links real and working, never placeholder #. - Re-check every rule in ~/.claude/RESUME_RULES.md. ## Output - Same format as the template (Markdown in, Markdown out). - Name it Resume_<Company>_<Role> next to the template. - Close with a short gap report: which JD requirements are strong, which are thin, so the user knows what to prepare a forward-looking interview answer for.
Andrew's actual RESUME_RULES.md (swap his specifics for yours)
# Resume Generation Rules
When generating or tailoring a resume, ALWAYS follow these rules:
1. Unique starting verbs - No verb repetition across the entire
resume. Run a verb audit every time.
2. No "philosophy" in Educator bullet - Never use the word
"philosophy" in the Educator line under Thought Leadership.
3. Commas instead of em dashes - Use commas, not em dashes, in
bullet points.
4. All bullets must fit on 1 line - No wrapping. If a bullet is
too long, trim it.
5. Resume must fit on 1 page - Use tight CSS spacing and concise
language.
6. DocOps full title - "DocOps Governance Workbench | Automated
Copy Compliance Tool | Python, LangChain"
7. Four core projects - SAGE, DocOps Governance Workbench,
QuickRamp, YouTube Shorts Pipeline. Always include all four.
8. No "AI Governance Framework" - Removed from all future
generations (duplicate of DocOps).
9. Darden PM Specialization - Include "Product Management
Specialization (Darden)" where relevant.
10. Cisco listed as Splunk - The Cisco/Splunk manager role should
say "Splunk", not "Cisco". End date is "Jan 2026".
11. Working links - LinkedIn, portfolio, and article links must be
real URLs, never placeholder # hrefs.
12. SAGE subtitle - Always include "Hackathon Winner | Figma Make,
Pinecone, Veo 3" in the SAGE project subtitle.
13. Summary max 2 lines - The resume summary section must not
exceed 2 lines.
## Cover Letter Rules
1. No Workday mentions - Never reference Workday in cover letters.
Focus on Splunk experience only.
2. Commas instead of em dashes - Use commas, not em dashes,
throughout cover letter text.
Rules 2, 6-12 are Andrew-specific. Keep the pattern, replace the
content: pin YOUR core projects, YOUR exact titles, YOUR links,
and any correction you find yourself making twice.
Build a capstone for that role at that company
This is the step almost nobody does, which is exactly why it works. Have Claude re-read the JD for the skills the team actually cares about, then build a small project in about a week of evenings that proves those skills. You walk into the interview with a demo instead of adjectives. Andrew's Salesforce capstone ideas came straight out of this prompt.
The capstone prompt we made
Read this job description: [paste the JD]. Draft three capstone project ideas I could build in about a week of evenings. Each idea must map to skills the posting actually asks for. For each one give me: the JD skills it proves, what it is in two sentences, and the demo moment I would show in an interview. Then help me pick one and turn it into a working plan with milestones.
Real example: Andrew's Salesforce capstone ideas
Artifact: working capstone with a demo momentCreate your portfolio page
One clean page: who you are, what you have done, the capstone front and center, how to reach you. It does not get you the job. It is the reason someone forwards your email. Build it with the portfolio prompt, iterate in plain English until it looks like you.
The portfolio prompt we made
Create a professional personal portfolio page as a single HTML file. Use clean, modern styling. Include: Header: [Full Name], [Professional Title/Tagline] About: [2-3 sentence professional summary] Experience Highlights: [3-5 key accomplishments with context] Skills & Expertise: [core competencies organized by category] What I Bring: [what makes you valuable, your unique combination of experience, perspective, and capabilities] Contact: [Email, LinkedIn, or however you want to be reached] Design notes: professional color palette, readable fonts, good spacing. No stock photos. Clean and confident. Save as portfolio.html. Then iterate: "change the palette to navy and white", "add a section for the capstone project with a screenshot", "make it less corporate, more personal".
Hosting it: go to here.now and publish the HTML file, you get a live URL at a here.now address in seconds. If you want a different URL name than the random one you are given, add a domain website name in here.now's site settings, either a custom slug or your own domain.
Artifact: live portfolio URL, like Andrew's: andrewyu.appCompile your battlecard
Everything you know about this company and this role, compressed onto one page you can review the morning of the interview: the panel and their LinkedIn highlights, the product, likely questions mapped to your stories, your questions for them. Andrew's real Salesforce battlecard is below, use its structure as your reference.
Reference: Salesforce battlecard
Artifact: one-page battlecardThrow it all into NotebookLM
Everything the playbook produced now converges. Create a notebook and add every source below, then let NotebookLM turn your whole pursuit into study material about you.
Generate an Audio Overview and listen on a walk, a Video Overview for the visual version, and quizzes to drill the gaps. Then the best part: use Interactive mode to drop into the podcast as a guest. Two AI hosts discuss how your background maps to the role, and you interrupt, ask questions, and defend your answers out loud. It is the closest thing to a mock interview you can run at midnight.
The NotebookLM prompt we made (Andrew's real Salesforce one, adapt the bracketed parts)
You are my interview-prep coach for the [Company] [Role] position. Use the sources I have added: the job description, my tailored resume, and my STAR stories. Build me a study guide that: 1. Lists the top eight requirements from the job description, in the team's own words. 2. For each requirement, names the strongest matching evidence from my resume and the STAR story that proves it. Keep each story tied to the company where it happened. 3. Flags any requirement where my evidence is thin, so I know where to prepare a forward-looking answer instead of a past-tense one. Then draft likely interview questions: - Five behavioral questions this panel would ask, each tagged with the STAR story I should answer it with. - Three technical questions about the core skills in the posting. - One curveball about handling disagreement, with the story to use. For the Audio Overview: focus the conversation on how my background maps to this specific role. Spend the most time where the match is strongest, and close with the gaps I should address out loud in the interview rather than hope nobody notices.
See it built: the real prompt file · the NotebookLM prep demo video
Artifact: a rehearsal room that knows your whole pursuitAndrew's real artifacts
Every file referenced above, in one place. These are the actual artifacts from the pursuit that ended in a signed offer, use them as templates for your own.
Hang in there
Job search is hard. Some weeks it will feel like nothing is moving, and that is normal, not a verdict on you.
And for those of you who are not from tech but are part of our church: you are more prepared for this moment than you think. We listen to messages. We write reflections. We read books. In other words, we are trained to keep track of narratives more than most, by virtue of being English majors at heart and people who fight for our own attention. In the age of AI, where the work is managing agents and keeping track of what is going on across various threads, that is the ultimate superpower.
And as you prepare with AI, prepare with a goal: be the most prepared candidate in the room, in the best position to get the job compared to everyone else. With AI, you can be. Whether that is a resume tailored to the exact job, a capstone project that shows you can do the work before you start, or a portfolio you built as a non-coder to highlight your AI projects, every step in this playbook exists to put you in that position.
You are unemployed at the right time to learn AI, more so than those who are currently working and cannot keep up. This is the opportune time, the kairos, for you to take advantage of it, get a job quickly, and be deployed quickly to serve.
Andrew Yu
Read: LeCun studied LLMs. I studied C.S. Lewis. Both of us learned the limits of language.
Follow along: substack.com/@andrewyu3