Refolk / Guides
Reference documents for running your own job search: how to write the materials, apply at volume, interview, and negotiate, in a form you can work from rather than skim.
Skip ahead and let me write itDeciding what you are selling and getting it onto a page: the resume, the profile, the portfolio, the story.
You will write one positioning thesis and deliver a stage-matched version of it to every interviewer so every debrief note points the same direction.
Interviewers score independently before the debrief, so contradictory signals resolve in the direction of caution - one fixed thesis is your defense against a process you never see.
You will be able to grade your own name-search footprint pass or fail on every criterion a recruiter checks, and know which item to fix before you go active.
Roughly 70 percent of employers screen candidates via social media, and in a 2023 ResumeBuilder survey 85 percent of hiring managers said they had rejected someone over an online finding.
You can score every accomplishment on four dimensions and walk out with one headline differentiator, a ranked support set, and a hold pile.
You can run a recruiter's own searches for your target title, find where your profile ranks, and make the specific edits that move it onto the first results page.
In Refolk's index, a US search for the current title 'Product Manager' resolves to 65,849 people while 'Product Owner' resolves to 14,164 - near-disjoint pools, so a candidate titled only one way is invisible to a recruiter searching the other.
After reading, you can grade any resume pass or fail against named binary criteria, match a second reviewer's verdict, and cite the exact criterion each failure breaks.
You can flag every genuine AI tell in your resume, keep the false positives worth keeping, and rewrite each one while holding a 75 percent keyword match.
The signal is clustering and genericness, not any single word: in Refolk's index, results-driven appears in 68,248 US professional headlines and spearheaded in 661, so both are ordinary human vocabulary, not machine inventions.
You can take a scattered multi-industry history and rebuild it around one defensible thread, knowing which role to lead with, which to demote, and which to cut.
The initial resume scan is a rejection gate, not a read: eye-tracking measured 7.4 seconds, and 80% of it lands on six data points, so a generalist who buries the thread in prose loses before the prose is read.
You can score every section of your profile against the semantic AI first pass and decide, section by section, to deepen it, keep it, or trim it.
LinkedIn's AI-assisted search is now the default over Boolean, so repeating your target title adds no vector signal while contextual, quantified sentences do.
You will score each short-tenure role on your resume and route it to a grouped heading, a standalone entry with a context line, or off the page entirely.
You can take any resume bullet with no tracked metric and decide in under a minute whether to quantify it, which lever to reach for, and what figure you could defend to an interviewer.
You can score your exact situation across five dimensions and land on one of three Open-to-Work settings, knowing what each gains and risks.
You can take your resume, look up your target country, and change the exact fields that differ there before you submit.
You will produce a publishable first-person About section in one sitting: a fold-safe hook, a body from your history, a proof block, an embedded keyword bank, and a CTA.
You will pick the right recommenders, send asks that get a yes, and end with three to five diverse recommendations a recruiter reads as real proof.
You can score your resume against a single posting on four dimensions and land it in the 75-85% evidenced band instead of guessing at keyword density.
You can score any drafted headline across six weighted dimensions, rank your variants head to head, and know whether it will surface in recruiter search before you publish.
In Refolk's index of professional profiles, only 6.4% of US Data Scientists put "machine learning" in their headline and only 5.0% put "python," so adding one core skill leaps you past roughly 94% of the field on that exact query.
You can build one base summary block from your history and produce a tailored, defensible version for any posting in under fifteen minutes, without inventing numbers.
You can carry your resume from a band-low read to a next-band read where every uplevel claim survives a deep-dive probe and a LinkedIn cross-check.
The uplevel read is decided before your bullets are read: the Ladders eye-tracking study found 80% of initial scan time goes to six top-left items, so the band signal lives in your title and company block.
You will grade each resume element pass or fail against a fixed bar and tell whether it projects one consistent target band or leaks a lower-level signal.
You can take one of your own projects, cut it to read at your target level, and defend those cuts against the probes that follow.
For every recurring follow-up probe after a work-sample presentation, you can name what it tests, spot its trap, and answer to the selling point instead of describing the work.
A polished portfolio looks compelling for about 30 seconds; the real evaluation starts with the follow-up questions, where rehearsed fluency without decision depth is exposed.
Running a search as a process: where to find openings, how many to send, what to tailor, and how to keep track.
You can take any posting whose title differs from your own and land on one of three defensible moves, with a rule for when each survives a reference check.
Verification providers confirm only your official title against HR records, so a lateral translation of a vague title is near-costless while a level jump is near-certain to surface.
You will locate the single lowest-converting stage in your own search from your real counts and match it to the correct fix.
You can take any posting and route it to one-click apply, the company portal, or a referral-first path using a repeatable four-dimension score.
You will build a set of alerts that surface fresh, non-duplicate roles within hours, route them into one queue, and keep them tuned with a repeatable two-week review.
Start with 5 to 10 alerts across two or three platforms; past roughly ten, inbox noise rises faster than the quality of matches.
You will sort each silent application into follow-up-now with a named recipient and date, hold, or write-off, instead of nudging all of them or none of them.
You will be able to grade one filled-out application pass or fail on field, attachment, and screening-question errors, the same way twice.
You will run one application sitting that drops dead postings before you invest, survives the form's three kill points, and ends with a counted number of confirmed submissions.
Completion collapses between the 5- and 15-minute marks: applications under 5 minutes complete at 12.47%, while those over 15 minutes complete at 3.61%, a 71% drop, so a hard 15-minute cap per form is the highest-leverage rule of the sitting.
You will grant or withhold submission consent per role, prevent duplicate submissions, and keep a log proving who was authorized to submit you where and when.
You will score any inbound recruiter message across five dimensions and decide to engage, qualify, or pass before you write a single reply.
You will score any auto-apply tool across six mechanics and reach a defensible verdict: run it, restrict it to standardized postings, or avoid it.
You can sort a recruiter message into legit or fraud using signals that still work after AI polish, and finish verification touching nothing the sender supplied.
Typos, grammar, and tone are dead tells: generative AI now produces recruiter outreach indistinguishable from a real talent acquisition professional's writing, so surface quality no longer discriminates.
You can take one past rejection, score it across five dimensions, and land on a verdict with a date and the one thing that has to change first.
The clock you owe a company depends on the stage it rejected you at, not a blanket three to six months: an auto-screen reject never reached a human, so no cooldown applies once you fix your keywords.
You can audit every channel that leaks your job search to your employer, close or mitigate each with a specific action, and name the residual risks that stay open.
The one leak with no platform mitigation is your employer's own hardware: under the ECPA, a company-owned device gives near-total visibility of files, traffic, and screen, and roughly three of four US employers monitor work devices.
You will be able to keep a search organised well enough to name the next action for every live application and know your true interview rate.
The mandatory core of any job application tracker is six fields: company, role, date applied, status, contact person, and follow-up date. Everything else is optional.
Read one posting and decide in under two minutes to apply now, apply within a set window, or deprioritize, with a defensible call on whether the early-review window has closed.
Almost half of all applications flow to openings posted in the past 48 hours, and the median posting duration in a 125M-application dataset was just seven days, so speed is real but role-dependent.
You will route any single posting to apply, ask the recruiter, or skip by reading its sponsorship language and cross-checking it against public filings in under a minute.
You can enter one set pay position correctly into every pay-field variant across a real batch, knowing which entries advance, which trigger a knockout, and which are illegal to ask.
The knockout is a ceiling test, not a floor test: an above-budget number auto-eliminates you silently, while a below-market one advances but caps your offer.
You will finish with one documented pay position - floor, target, stretch, and an exact entry per form-field type - that you apply identically across every posting in the batch.
You can decide, element by element, what to change on your resume for a posting and what to lock, in one row each, without rewriting the whole document.
The initial recruiter screen averages 7.4 seconds and fixates on a fixed set of points: name, current title and company, previous title and company, dates, and education, all in the top third.
You can build a fixed prep-hour budget across your live loops, resolve every calendar collision with the right move, and choose which loop to deliberately under-prep.
Every stage from the recruiter screen to the final round, and what each one is actually testing.
You can name each test in a screening link, state what it measures and how it is scored, and decide per module whether to prepare, answer straight, or decline.
You will locate your exact mid-process situation, send the right message, and know precisely what it signals to the recruiter and how it can be misread.
A second reschedule reads as flaky where the first read as life; the mechanism is coordination cost, and a two-hour panel is far harder to move than a fifteen-minute recruiter call.
You will be able to grade each live interview process pass or fail on your own upkeep and flag a stalling loop before it dies.
The stall signal is a Next Action Date you set and blew past, not silence from the company, because silence is ambiguous and a past-due action you owned is unambiguously your neglect.
You will carry one interview loop from first invite to final round while employed, spending one PTO day, keeping the search invisible, and holding a single reschedule in reserve.
You will be able to score any "where else are you interviewing" moment and pick name, signal, or decline, knowing what each does to your timeline and exposure.
You will run pay through every gate of one loop knowing what to say, deflect, and record so you finish with the role's range confirmed and your number protected.
You will finish with 6 to 10 rehearsed STAR stories and a coverage matrix proving every competency your target loop tests has at least one strong story mapped to it.
You will locate any post-loop stage or recruiter phrase in a lookup table, read what it proves about hire, level, and timeline, and know your single move.
Passing the committee approves the person, not the job; at Google a team-match packet expires in about 8 weeks if no manager with open headcount claims you, and committee approval is valid for roughly 12 months.
You can score a quiet hiring process into still in, stalled, or out, and take the one correct move for each read instead of guessing from the silence.
You will be able to score any interview round to one of four verdicts - use openly, use and disclose, prep-only, or do not use - with a disclosure line ready to send.
You will name the protected topic behind an off-limits question, answer the lawful concern instead of stonewalling, and record the exchange for later.
You can move through a profitability case in the interviewer's expected sequence and back out of the two wrong turns that sink most candidates.
You will be able to name every signal an AI-monitored interview collects, spot which of your honest habits trips each one, and set up your room, disclosures, and answers so a clean interview reads clean.
An AI-monitored interview never observes cheating; it observes correlates of cheating, then a threshold turns a glance or a pause into an accusation.
You can run a 45 to 60 minute system design round to a fixed clock: scope, estimate peak QPS and storage, name a datastore trade-off, and reverse a bad choice.
You can build a prep plan sized to your runway and run the live round in a fixed sequence that surfaces the behaviors interviewers actually score.
You will turn a roster of names into a concern map, cover every interviewer's lens with a story matrix, work the room, and send follow-ups that survive comparison.
You can grade any pay answer, from recruiter screen to final round, as pass or fail against a fixed rubric before you say it.
You will run a live AI voice screen start to finish: set up audio, open tight, answer adaptive follow-ups with one concrete example each, and avoid the delivery patterns that lower a competency score.
The live AI voice screen runs short: documented ranges cluster from 2 to 20 minutes, so a short specific answer literally scores higher than a long vague one.
You will be able to name what every question in a recruiter phone screen is collecting, and answer to the hidden criterion instead of the surface words.
You can grade each resume line pass or fail on whether it survives a live probe, and name exactly what to add or soften on the ones that fail.
Reading an offer, comparing two, and negotiating without torching the relationship.
You can turn a panic week into a day-by-day plan: which process to push, which to drop, and what to send each one on which day, until you hold one decision.
You will be able to grade a specific offer against fixed criteria and get one of three verdicts: keep searching, wind down selectively, or shut down fully.
You will route each offer ask - base, sign-on, title, level, remote, equity - to the one party who can grant it, and name what each cannot.
You will score each live interview to accelerate, keep alive, or drop, and get a defensible call on whether any pending bet beats the offer in hand.
You will be able to diagnose a mis-leveled offer, run the leveling conversation, and leave with a higher level in writing or a dated, criteria-bound review clause.
You will grade each compensation term in an offer letter as PASS or FAIL and know the exact field to request in writing before you sign.
A compensation term is binding only when it states a number or a rule a court would not have to guess about; grade everything else FAIL until a number or rule replaces the placeholder.
You will be able to read any repayment clause, compute your exposure at each departure month, spot the carve-outs that protect you, and draft the exact edit to request.
You will time and deliver your resignation so the new offer is locked in writing before you give notice, your notice length is defensible, and your references survive.
You can carry an option or RSU grant through ownership, dilution, the liquidation stack, and taxes to a single expected-value range you can compare to cash.
You will score each restrictive covenant in your offer on restrictiveness and enforceability where you work, then decide to sign, negotiate, or walk with named clauses to push on.
You will send a dated extension request sized to your role level, keep it separate from any comp talk, and run a clean recovery if the recruiter says no.
You will convert an hourly contract-to-hire rate into a benefits-adjusted annual number and reach a defensible take, counter, or decline call.
You can convert any offer's benefits summary into one expected annual dollar figure sized to your own usage and compare two offers on that basis.
You will tally what you forfeit by leaving, convert it into a defensible after-tax sign-on number, and send a scripted make-whole counter to the recruiter at the right moment.
Roughly 30 to 40 percent of a sign-on bonus disappears to combined federal, FICA, and state withholding before it reaches your account, so a gross-for-gross match under-asks every time.
You will time the remote ask, propose a structured operating model with named trades, and get the exact arrangement written into the offer letter as enforceable language.
After reading, you can decode any relocation offer line by line, mark each component's tax and gross-up status, and state your clawback exposure by month before you sign.
Since the 2017 Tax Cuts and Jobs Act, most employer-paid relocation is taxable wages: a $10,000 benefit without a gross-up can lose $3,000 or more to tax, so the real ask is the gross-up, not the headline number.
You will be able to grade your offer conversation complete or incomplete against fixed criteria and know exactly which items to confirm in writing before you sign.
You will score any resume line on check-likelihood and mismatch-severity, then decide to leave, document, soften, or correct it before the check runs.
A standard employment verification reliably returns only three fields - employer, job title, and start/end dates - so almost all of your verification exposure sits on three or four resume lines, not the whole document.
After reading, you can assemble, for each past employer, the documents and live verifiers a screener accepts, and pre-explain every mismatch before consent is signed.
Career changes, layoffs, gaps, re-entry, and the searches that do not go to plan.
You will score any out-of-field posting on four fixed dimensions and land on apply-direct, bridge, or skip instead of guessing per listing.
You will find where your field-switch search leaks, rule out materials versus targeting versus an over-ambitious jump, and re-scope to a bridge role that converts.
You can run a return-to-work interview from screen invite to follow-up note with a timed break statement, dated currency proof, and a competency-to-STAR map.
You can name every public and employer-provided reemployment program, state its eligibility gate and whether it is mandatory, and decide in one sitting which to enroll in and which to skip.
After reading, you can go through a severance offer line by line and state what each component is worth, whether it is negotiable, and how it can quietly cost you.
You will score any transferable-skill claim on four dimensions and rule it lead, mention, or drop before a domain expert can probe it and find the gap.
You will turn a stale prior salary into one re-entry target figure plus a confidence band, scored across six measurable dimensions of your break.
You will rebuild your own multi-year-gap resume from a stale draft into a sent, ATS-parseable version with a placed break entry and refreshed bullets.
You will score a specific stopgap opportunity across six dimensions, land a Take, Time-Box, or Pass verdict, and write the exit trigger for any bridge you accept.
States zero your partial unemployment benefit at a hard ceiling: New York pays 0% at 31+ hours, and over roughly 32 hours a week you can be deemed fully employed, so a stopgap sized just under the cap can pay more total than one just over it.
You will audit your resume signal by signal, neutralize each age tell without lying or leaving a gap, and answer the screener who notices.
The measured penalty is large and well established: one correspondence study of over 40,000 applications found callbacks about 35% lower for older workers, and it is steeper for women.
You can run the return-to-a-former-employer path end to end: clear the eligibility gate, time the outreach, reactivate a warm ally, and negotiate a leveled offer.
Before you send a single message, learn how you were coded at separation: the 'Do Not Rehire' flag lives in one company's system, and it auto-triggers a rejection when the applicant system cross-references your old employee ID.
You will render your founder, owner, consultant, or freelance history as a hireable, verifiable resume entry that defuses the flight-risk and overqualified flags.
You will run a multi-week field-switch campaign paced to your measured yield, routing each posting by tier and channel, and reading your pipeline to know when to widen, narrow, or add referrals.
You will run a career-break return end to end: name the gap, build a returnship pipeline, apply in parallel with regular jobs, and vet a program before you accept it.
After reading, you can run the first 72 hours after an accepted offer is pulled in the right order, starting from the "did I already resign?" fork.
You can look up any element of your service record and write the exact civilian resume line, plus the honest clearance status wording that neither oversells nor reads as dishonest.
After reading, you can assemble and send a written severance counter with specific terms and numbers, ordered by likelihood of yes, and time it inside your legal window so you never sign under pressure.
An individual ADEA waiver must give you at least 21 days to consider it; a group termination program must give at least 45 days, and the 7-day revocation after signing cannot be shortened by agreement.
You can grade your field-switch search pass or fail against five criteria, name the one that fails, and fix it before the first application goes out.
You will run a full re-entry application cycle: refresh a stale profile, route each posting into one of three lanes, fill the gap-forcing form fields, and submit at a sustainable weekly cadence.
You can turn your existing work history into a career-change resume through a fixed eight-stage build that ends in a draft passing ATS parsing and recruiter scrutiny.
You will grade your own readiness for a career-change interview against pass/fail criteria and name exactly which ones you still fail before the loop starts.
Where the hiring actually is: by role, by city, by industry, and what the numbers mean for your search.
You will turn one role's posted ranges across four metros into a net-pay table that separates cost of labor from cost of living and drops unreliable ranges.
Base pay drops about 11% from a Tier 1 hub to a Tier 2 metro and 16% to Tier 3, but new-hire equity drops far more, about 29% and 36% respectively, so relocation cuts total comp more than the base number implies.
After this you can grade a compiled market pay estimate pass or fail on six criteria, so two people grading the same figures reach the same verdict.
You can take any quoted total, split it into guaranteed floor, at-risk variable, and discounted equity, and compare two market segments without being fooled by the bigger headline.
You will be able to score any trending job title across six demand signals and land on a defensible verdict: retarget, bolt on, or ignore.
In Refolk's index, 59,784 US people list prompt engineering as a skill against only 530 holding the title, a ratio of roughly 113 to 1, so the skill outlived the standalone job.
You will grade a specific target against a deduplicated opening count, refresh rate, level depth, pay floor, and fit gate, and return pass, adjust, or reject.
You can take one open posting, score it on named signals, and reach an apply-now, verify-first, or skip verdict without discarding a real opening.
The most checkable single signal is the requisition ID, not posting age: a genuinely reopened role gets a new req number, so the same internal ID reposted repeatedly is near-proof of a pipeline listing.
You will produce a triangulated low, median, and stretch pay range for your role, level, and metro, each figure traced to a dated source and adjusted for lag and geography.
An un-aged government median can be up to four years stale, understating a job with 3% annual raises by roughly 9 to 12 percent, which is exactly the gap the aging step closes.
You will produce a ranked shortlist of three to five metros for your exact role, each scored on opening density, competition, and cost-adjusted pay, plus a move-or-stay breakeven.
You can score a named company's remote policy stability across weighted public signals and land on a low, medium, or high reversal-risk verdict tied to an action.
You will be able to read a bootcamp's outcomes page, name what each number proves, flag the inflation tactic inside it, and decide whether to trust it.
You will score one named certification against your own target postings and get a payback period in months, then decide to pursue it, skip it, or have an employer fund it.
The confounder-adjusted value of a job-relevant credential is roughly a 6% wage premium per Brookings, not the 23 to 33% vendor premiums, and accumulation returns fall close to zero for experienced workers.
You will be able to pull the public downside-risk record on a named employer in about an hour and end with a defensible stable, watch, or avoid read.
You will discount any applicant count into a true-competitor estimate and convert it into one action: apply, deprioritize, referral-only, or widen.
You will score your occupation's automation exposure from public task data and reach one defensible call: hold, reskill in place, or retrain.
You will turn three public wage trackers into a personalized net pay-change range and a go/no-go verdict for leaving your current job.
The switcher-stayer gap ranges from 0.5 to 3.0 percentage points depending on the tracker, so quote all three as a range rather than picking the widest one as the premium.
You can take a shortlist of industries, score each for your occupation on four dimensions, and sort them into target, watch, or skip.
You will run a fixed eight-step sequence that turns labor-market data and aggregate posting demand for your occupation into a resume built for the market.
Two federal signals disagree on which skills matter: problem solving ranks top by growth-correlation (7.77) but only mid by weighted score change (0.23), so a market-built resume has to carry both the important-now list and the growing list.
You will grade any live interview loop as on-market, elevated, or outlier across round count, timeline, panel size, and unpaid work, with thresholds two people would apply the same way.
You will be able to classify any "remote" posting as fully remote, hybrid, or bait, before you spend an hour applying.
Fully remote means no required office attendance ever; any location limit is a payroll or legal constraint, not an in-person requirement.
Put this to work
Drop a resume or a LinkedIn URL. I rank the live openings against it, rewrite the resume and write a cover letter for the best of them, and fill in the employer's form when you press the button. You read, you decide what goes out.
01Drop your resume
A PDF or a LinkedIn URL. About a minute, once.
02I rank the openings
Every weekday morning, the live catalog scored against your history. Up to 20 worth your time, not two hundred links.
03Each one is written up
Resume rewritten for the posting, a cover letter, a fit score. Press send, or let me fill in the form.
500 free credits on sign-up. No card. Nothing is sent until you say so.
Every guide is also served as plain Markdown at /candidates/guides/<slug>.md, and the index at /candidates/guides/index.md.