peanut.tech — Scientific Hiring Execution

The Problem

Everyone is p-hacking.
Nobody's getting signal.

Companies move the goalposts mid-process. Candidates spray 200 applications until noise produces a hit. Both sides end up with false positives and zero trust.

🏢 Company-side p-hacking

Interviewers score on vibes. Rubrics shift after seeing candidates. The loudest voice in debrief wins. Unstructured interviews predict just 4% of actual job performance.

~4%

Variance in job performance explained by unstructured interviews

0.20

Predictive validity — barely better than flipping a coin

👤 Candidate-side p-hacking

50–200+ applications. Random prep overlap. Lucky interviewer match. One noisy "significant" hit. Companies end up hiring the luckiest mass-applier, not the most capable.

0.4%

Success rate per application submitted

83/wk

Applications by top-10% spray-and-pray candidates

The Hiring P-Hack Simulator

Drag the sliders. Watch Bayes' theorem destroy your interview pipeline. Then flip the protocol lock.

🔓 No protocol 🥜 Protocol lock

Applicants 200

% Actually Qualified 28%

Interview Noise (FP rate) 25%

True Positive Rate (power) 65%

How often a truly qualified candidate is correctly identified

Planned Hires 5

Cost of Unqualified Hire $300k

Recruiting + onboarding + severance + backfill + lost productivity

50%
Hire is qualified

True positives

False positives

Missed talent

Who passed the interview?

Qualified & passed

Unqualified & passed (noise)

Unqualified & rejected

Qualified & missed

Without protocol

2.5 expected unqualified hires

$750k

exposure to wasted spend

→

With protocol lock 🥜

1.3 expected unqualified hires

$390k

exposure to wasted spend

Value of Peanut $360k saved per hiring round

Model: Bayesian binary classification. PPV = π(1−β) / [π(1−β) + (1−π)α]. Protocol lock reduces false pass-through (α) by 50% and raises true positive rate from 65% → 90%. Unstructured interviews are noisy in both directions — they miss qualified candidates and pass unqualified ones. Grounded in meta-analytic evidence: structured interviews roughly double predictive validity vs unstructured (Schmidt & Hunter 1998; Sackett et al. 2022).

The Fix

Pre-register the protocol.
Lock it. Execute it.

Like scientific pre-registration kills p-hacking in research, Peanut kills it in hiring. Commit to your process before the first applicant lands.

Step 01

Pre-commit the Protocol

Lock competencies, rubric anchors, question banks, weights, and decision thresholds before a single application arrives. Immutable. Timestamped. No post-hoc changes.

Step 02

Execute with Integrity

LLM-assisted interviews that probe evidence, not trivia. Adaptive questions tied to what each candidate claimed — collaborative forensics, not interrogation.

Step 03

Signal, Not Silence

Auto-generated, rubric-based feedback every candidate can trust. Stack-rank with statistical controls. No ghosting — because the protocol won't let you.

🥜 Peanut Protocol — Senior Backend Engineer 🔒 Locked · Mar 2 2026 09:41 UTC

competency.k8s_troubleshoot weight: 0.20 · anchor: real-world failure handling + root-cause + mitigation ✓ sealed

competency.system_design weight: 0.25 · anchor: trade-off reasoning with scale evidence ✓ sealed

competency.cross_team weight: 0.20 · anchor: influence outcome without authority ✓ sealed

competency.execution weight: 0.20 · anchor: shipped under ambiguity with measurable impact ✓ sealed

competency.communication weight: 0.15 · anchor: concise escalation + stakeholder clarity ✓ sealed

decision.threshold ≥ 3.8 weighted avg · ≥ 3.0 every competency · committee unanimous ✓ sealed

anti_phack.irr_check inter-rater reliability auto-calculated · outlier alerts enabled ✓ sealed

For Candidates

Apply with 🥜

One profile. Every application. Real feedback, always. No more 50 CVs, no more ghosting, no more black holes.

📄

Upload once, apply everywhere

Your Peanut Profile replaces the PDF circus. Skills, evidence, projects — structured and rubric-ready.

🔍

Evidence-based interviews

Questions probe what you claimed — no trivia, no gotchas. Collaborative, dignified, adaptive to your background.

📊

Always get the "why"

Rubric-based feedback on every competency. Not "we went with someone else" — real data you can grow from.

🛡️

Peanut is the neutral referee

Feedback comes from the protocol, not the company. They committed to transparency the moment they chose Peanut.

🥜

Your Peanut Report

Senior Backend Eng · Acme Corp · Rank #3 of 12

K8s Troubleshoot

4.2

System Design

3.8

Cross-team

3.4

Execution

4.5

Communication

2.8

🥜 Peanut Feedback

Strong execution evidence (4.5) with depth in K8s failure recovery. Gap: cross-team influence lacked measurable outcomes — consider framing with stakeholder metrics. Communication scored below bar on escalation clarity.

Who it's for

Better talent. More dignity.
Less 🥜 TC.

For Companies

→ Pre-registered protocols that survive legal audits and bias reviews
→ 2× predictive validity over unstructured interviews
→ Auto inter-rater reliability checks — weak questions flagged automatically
→ One audit trail for every interview that ever happened
→ Higher offer acceptance because top talent trusts the process

For Candidates

→ Always know why — rubric-based feedback on every decision
→ One profile, no more rewriting 50 tailored CVs
→ Evidence-based interviews that respect your actual experience
→ No ghosting — the protocol forces transparency
→ Skill wins over luck. Period.

Stop p-hacking your hiring.

🏢 Company-side p-hacking

👤 Candidate-side p-hacking

The Hiring P-Hack Simulator

Pre-commit the Protocol

Execute with Integrity

Signal, Not Silence

Upload once, apply everywhere

Evidence-based interviews

Always get the "why"

Peanut is the neutral referee

For Companies

For Candidates

Kill the noise.