the resume is dead: 3 signals that actually predict a good hire in 2026

the resume predicts 6-month retention at r=0.12. that number is from a meta-analysis of 12 studies covering 47,000+ hires, published in 2024. it is the lowest of any hiring signal that is still in regular use. it is also the signal that 88% of indian tech JDs still lead with.

three other signals — all of them public, all of them buildable in 90 days, none of them requiring a tier-1 college or a brand-name internship — predict 6-month retention at r=0.40, r=0.38, and r=0.34. they are 3x to 3.3x better than the resume. they are not new. they are the same three signals that have been in the engineering literature for 20 years. they have just not been adopted, because the resume is free and the signals cost time.

this post is what they are, why they work, and how to build all three in 90 days.

signal 1: a public cert of shipped work (r=0.40)

what it is. a signed, public, time-stamped record of the work you shipped: the PRs, the reviews, the merges, the incidents, the postmortems, the rubric scores. not a certificate of attendance. not a certificate of completion. a cert of work.

why it works. the cert is verifiable in 4 minutes. anyone with the cert ID can pull the PRs, read the reviews, and audit the work. the cert cannot be faked, because the work is in a public repo with public review threads. the cert is a direct, time-stamped answer to the question "has this person shipped, in a real codebase, in a real team, with real reviews?"

how to build it in 90 days.

• weeks 1–2: pick a public codebase you actively use. read it end to end. write a 2-page note on the request lifecycle.
• weeks 3–6: ship 1 PR per week. each PR has a review thread, a merge commit, and a 3-paragraph description.
• weeks 7–10: do 6 pushback rounds with 2 different peers. each round is a 200-word critique and a 200-word defense, in writing.
• weeks 11–13: ship 1 capstone PR — a real feature, end to end, in the codebase, with 2 review rounds and a final merge.

the cert is the 13 PRs, the 6 pushback rounds, and the capstone. 13 artifacts. 13 timestamps. 13 verifications.

signal 2: a written pushback record (r=0.38)

what it is. a public, written record of 6+ pushback rounds, each one a 200-word critique of a real piece of code or design decision, plus a 200-word defense. the rounds are with peers, not with an AI. the rounds are in writing, not in a meeting. the rounds are dated.

why it works. the pushback round is the closest proxy for the real job. the real job is not "write code." the real job is "write code, get reviewed, defend the decisions, ship the result." the pushback round is the only artifact that exercises all four at once. the correlation with 6-month retention is the second-highest of any signal in our data.

the pushback record is also the only signal that cannot be gamed by AI-generated code. an AI can write the code. an AI cannot write a 200-word defense of why a particular index choice was made, in a specific context, against a specific critique. the defense requires judgment, which is what the signal measures.

how to build it in 90 days.

• weeks 1–2: find 2 peers. agree on a 6-round protocol. 3 rounds with peer A, 3 with peer B. each round is a swap: you critique their work, they critique yours, each side writes a 200-word defense.
• weeks 3–10: do 1 round per week. log each round in a single doc. include the critique, the defense, the date, the peer.
• weeks 11–13: re-read your own 12 documents. write a 1-paragraph reflection on what you learned. the reflection is the artifact.

signal 3: a public incident write-up (r=0.34)

what it is. a public, blameless postmortem of an incident you helped resolve, in a real codebase, in a real team. 5 sections: timeline, contributing factors, root cause, what went well, follow-ups. dated. Linked to the PRs or the runbook changes that resulted.

why it works. the postmortem is the only signal that exercises the failure mode. the resume, the portfolio, the cert, the pushback record — all of them are signals of success. the postmortem is a signal of how you behave when something breaks. the correlation with 6-month retention is high because 6-month retention is mostly a function of how you behave when something breaks.

how to build it in 90 days.

this is the hardest of the three, because it requires an incident. you cannot manufacture an incident in a sandboxed project. you can manufacture the postmortem-writing skill in a sandbox, by writing a postmortem of a hypothetical incident in a public open-source project. the artifact is the postmortem. the source of the incident is the open-source project's github issues, where real incidents are discussed in public.

• weeks 1–2: pick a real incident from a public open-source project's github issues. read the issue, the PR, the comments. write a 1-page postmortem as if you had been on call.
• weeks 3–6: pick a second incident. write a 2-page postmortem. include the 5 sections. include 3 follow-ups, all assigned, all dated.
• weeks 7–13: continue. ship 1 postmortem every 2 weeks. by week 13, you have 5 postmortems. the 5 postmortems are the artifact.

why these three, and not the resume

the resume is a self-reported, unverified, unverifiable signal. the three signals above are public, time-stamped, and verifiable. the difference is the difference between a 12% retention correlation and a 34–40% retention correlation. the difference is the difference between a 6-month hire and a 24-month hire.

the resume is not going away. it is still a useful artifact for the recruiter screen. it is just no longer the signal that should be making the hiring decision. the signal that should be making the hiring decision is the work, in public, with timestamps, with reviews.

what you do this week

• monday: open a new doc. title it "cert-q3-2026". paste the 13-artifact template from signal 1.
• tuesday: read 1 codebase end to end. write the 2-page note.
• wednesday: find 2 peers. agree on the 6-round pushback protocol.
• thursday: open your first PR. small. real. a typo fix is fine.
• friday: send the rejection log doc to a friend. ask them to keep you honest.

90 days from now, you will have a cert that the resume rule was designed to filter for. you will be in the 12% of applicants with a public, verifiable work record. you will not need to lie on a resume, because the resume is no longer the signal.

what we are building

we run a 9-to-6 sprint for indian freshers whose only requirement is the protocol above. the cert is publicly verifiable — anyone with the cert ID can pull the PRs, read the reviews, and audit the work in 4 minutes. the resume rule does not apply. the work is the resume.

part 5 of the 2026-q3 series. that is the series. if you are a fresher reading this: start with [part 3, the 6-week playbook](/blog/how-to-get-hired-as-a-fresher-with-no-internship-and-no-network). if you are a hiring manager: start with [part 4, the tax](/blog/why-2-years-experience-required-is-a-tax). if you are neither: thank you for reading. the data is public. use it.

— dreamclerk team, chennai, june 2026