Humanity’s Last PRD
Latest drop: Gemini 4 Argon · 13 Oct
Tasks / Leadership

Hire a PM

Can the model design a hiring process that finds the right PM, and make the call on real candidates from the evidence?

Task type v1.1 · 2 tasksLast changed 6 Oct 2026 · ChangelogDifficulty

What AI gets right here, and what you’ll still have to catch

From 18 graded outputs by 9 models. 61% were usable with at most a quick edit.

Reliably right

  1. Produces the required deliverable100% pass
    The output provides a complete loop, scorecard, and decision plan in the requested form, under 900 words, usable by Amara with light edits.
    GPT-6.1 Sol · API · A loop for the first growth PM
  2. Tests what the last hire failed at100% pass
    The loop includes an influence simulation with the Head of Sales and a scorecard dimension on influence without authority, directly testing what the last hire failed at.
    GPT-6.1 Sol · API · A loop for the first growth PM
  3. Finds which rounds line up with performance100% pass
    It compares each round's scores with 12-month ratings, finding behavioral lines up and take-home, presentation, VP Engineering do not.
    GPT-6.1 Sol · API · What our interviews predict

Where it slips

  1. Uses the supplied evidence correctly56% pass
    The take-home correlation is stated as -0.57, but the data yields approximately -0.33; this fact about the current situation is incorrect.
    GPT-6 Astra · ChatGPT · What our interviews predict
  2. Avoids unsupported claims60% pass
    The claim 'A growth PM in a senior role is likely to be exactly that kind of person' is presented as fact without evidence and is not labelled as an assumption.
    Opus 5.5 · Claude · A loop for the first growth PM
  3. Defines good for this role first63% pass
    The memo does not explicitly name the competencies that matter most for this PM role and level, nor define what strong and weak look like before designing the interviews.
    Sonnet 5.5 · API · What our interviews predict

How it’s graded

Each check is built on how world-class product and growth leaders do this work, as they’ve described it on Lenny’s Podcast, and names the leader it comes from.

  1. Judges on evidence, not presence

    Does the output judge candidates, or design the process to judge them, on specific evidence of results they drove and how they worked with their team, rather than impressions such as presence, confidence or polish?

    Passes when Asks for or weighs concrete evidence: the metric a candidate moved, the part they played, how they shared credit, what they learned from a failure. Impressions are treated as weak signals.

    Product and growth leadersOn Lenny’s Podcast

  2. Defines good for this role first

    Does the output set out what good looks like for this specific role (the competencies that matter most at this level, and what strong and weak look like) before judging candidates or designing interviews?

    Passes when Names the competencies weighted for this role and level, with strong and weak signals, and the judgement or the process follows from them.

    Petra WilleLenny’s Podcast

  3. Keeps each judgement independent

    Does the output keep each interviewer's judgement independent (written down before the group discusses it), and guard against the loudest voice or the first speaker deciding?

    Passes when Requires or relies on written, independent evaluations before discussion, and discounts views that changed under group pressure.

    Product and growth leadersOn Lenny’s Podcast

Plus our standard checks

Uses the supplied evidence correctly · Addresses the actual decision · Respects explicit constraints · Identifies material uncertainty · Avoids unsupported claims · Produces the required deliverable · and 3 written for each task, which you’ll see in the tasks below.

The tasks

Read the brief, then compare up to three outputs, each with the LLM judge’s verdict on every check. Highlights mark what a PM had to fix.

The brief

You're helping Amara Osei, VP Product at Copperline, hire our first Growth PM (a senior role). She's shared her draft interview loop and asked you to redesign it. Write the loop you'd run (each round: who runs it, what it tests, how long it takes), the scorecard (what strong and weak look like for each thing we're testing), and how we'll make the decision at the end. No more than 900 words. What we know is below.

What the model was given6 items: About Copperline, The role, Amara's draft loop, The last two PM hires, From the recruiter, Who can interview
About CopperlineInvoicing and payments software for small accountancy firms. 60 people. Trial to paid conversion is 9%.
The roleOwns trial conversion and expansion revenue. Works with three engineers and a designer, reports to Amara, and has to ship experiments every week. Needs Sales and Marketing to change how they work, without managing them.
Amara's draft loop1. Recruiter screen. 2. Take-home: 'design a growth strategy for Copperline', a 10-page deck, five days to complete. 3. Presentation to a panel of six. 4. Culture fit with the CEO: 'would I grab a beer with them?' 5. Debrief: everyone discusses, then votes by show of hands. There's no scorecard.
The last two PM hiresOne left after five months: 'couldn't get Sales to change anything'. The other is doing well. Both sets of interview notes say mainly 'great presence' and 'very smart'.
From the recruiterOf the last 40 candidates given the take-home, 14 withdrew, saying they didn't have time. Nine of the 14 had young children or demanding current roles.
Who can interviewAmara, the Head of Engineering, a senior designer, the Head of Sales, a data analyst and the CEO. Each can give at most two hours to this hire. Amara wants an offer out within three weeks of the first interview.
What a strong answer doesThe answer key the graders mark against

A loop built on a scorecard written first: for a senior growth PM, experimentation and analytics, execution, and influence without authority (weighted up, since the last hire failed on it), with strong and weak signals for each. Rounds test the real job: a live diagnosis of Copperline's own trial funnel instead of a five-day generic deck (which loses candidates, mostly parents and people in demanding roles), a behavioural round on past experiments and metrics actually moved, and an influence round run by the Head of Sales. Replaces the 'grab a beer' culture fit with a structured values or collaboration conversation. Each interviewer writes their evaluation before the debrief, and the decision is made against the scorecard, not a show of hands. Fits each interviewer's two hours and the three-week target, and adds reference checks.

Critical failures (cap the score at 40)
  • Keeps a culture-fit test based on whether the interviewer would like a beer with the candidate
Task

Mid-level · v1.1

The outputs

9 setups answered this brief.

GPT-6.1 SolAPI98Usable with a quick editOpus 5.5Claude94Usable with a quick editGPT-6 AstraChatGPT92Usable with a quick editSonnet 5.5API90Usable with a quick editGPT-6 LunaAPI87Usable with a quick editGemini 3.8 FlashAPI87Usable with a quick editGemini 4 ArgonAPI85Usable with a quick editHaiku 5.5API83Usable with a quick editGemini 3.5 Flash-LiteGemini81Usable with a quick edit
Gemini 4 ArgonAPI

Usable with a quick edit9 of 12 checks passed

The output is a complete, actionable loop with clear evidence-based scorecard and decision process. A PM could adopt it with minimal edits, even though it omits explicit uncertainty acknowledgment.

Output

Here is the redesigned interview loop for Copperline’s first Growth PM.

We’ve removed the 10-page take-home—which was costing us 35% of our pipeline, disproportionately filtering out parents and busy top performers—and replaced the six-person panel, which burns interviewer hours and selects for "presence" over substance. We’ve also replaced the CEO’s "beer test" and the hand-raise vote to eliminate groupthink and bias.

This new loop fits your 3-week timeline, requires zero unpaid homework, and keeps every interviewer well under their 2-hour cap per candidate.

1. The Interview Loop

Stage 1: Recruiter Screen (30 mins) * Interviewer: Recruiter * What it tests: Role alignment, compensation, B2B SaaS growth experience, and baseline appetite for high-velocity experimentation.

Stage 2: Hiring Manager Screen (45 mins) * Interviewer: Amara (VP Product) * What it tests: Past Impact & Strategic Thinking. Deep dive into a past funnel or expansion initiative they owned end-to-end. Focuses on how they diagnose bottlenecks and drive commercial outcomes.

Stage 3: Live Growth Jam (60 mins) * Interviewers: Data Analyst & Senior Designer (plus 15 mins prep) * What it tests: Experimentation Velocity & Squad Collaboration. No prep required. We share a mock snapshot of Copperline’s 9% trial-to-paid funnel during the call. The candidate works live with the Analyst and Designer to diagnose drop-offs, form hypotheses, and scope two scrappy experiments to ship next week.

Stage 4: Go-To-Market Alignment (45 mins) * Interviewer: Head of Sales * What it tests: Cross-Functional Influence. Directly addresses why our last PM left. Behavioral questions and a live roleplay on getting Sales and Marketing to change workflows without direct authority.

Stage 5: Delivery & Ambition (60 mins — two 30-min back-to-backs) * Part A: Head of Engineering (30 mins). Tests Pragmatic Execution: balancing tech debt with weekly experiment shipping, and how they work with engineers. * Part B: CEO (30 mins). Tests Commercial Empathy & Ownership: passion for small accountancy firms, comfort with startup ambiguity, and executive alignment.

Time commitment per finalist: Amara (1.25 hrs), Analyst/Designer (1.5 hrs), Sales (1.25 hrs), Eng/CEO (1 hr)—including the 30-minute debrief.

---

2. The Scorecard

Interviewers will grade only their assigned competencies on a 1–4 scale (1 = Strong No, 2 = No, 3 = Yes, 4 = Strong Yes), citing specific evidence—never "great presence" or "very smart."

1. Experimentation Velocity & Rigor (Tested in: Live Jam, Eng) * Strong: Comfortably forms hypotheses from imperfect data. ruthlessly scopes MVPs down to ship within a week. Focuses on learning velocity over perfection. * Weak: Wants months of research before shipping. Designs massive features instead of tests. Confuses shipping features with moving metrics.

2. Cross-Functional Influence (Tested in: GTM Alignment, HM Screen) * Strong: Speaks the language of Sales (quota, pipeline, commission) and Marketing. Co-creates solutions, anticipates resistance, and uses data and quick wins to earn trust. Has clear examples of changing GTM behavior. * Weak: Expects Sales/Marketing to comply "because the data says so." Complains about stubborn stakeholders or escalates to the VP immediately when blocked.

3. Funnel & Expansion Strategy (Tested in: HM Screen, Live Jam) * Strong: Connects micro-experiments to macro revenue goals. Clearly articulates levers for moving a 9% conversion rate and driving expansion in a B2B workflow tool. * Weak: Throws random growth hacks at the wall. Focuses purely on top-of-funnel acquisition or B2C viral loops that don’t apply to accountancy firms.

4. Squad Collaboration & Execution (Tested in: Live Jam, Eng) * Strong: Treats Design, Data, and Eng as problem-solving partners, not ticket-takers. Welcomes pushback during the Live Jam and adapts their ideas. * Weak: Dictates exact UI solutions to the designer; ignores technical constraints raised by engineering; talks over teammates.

5. Commercial & Customer Empathy (Tested in: CEO, HM Screen) * Strong: Genuinely curious about how small accountancy firms work, make money, and experience friction with invoicing/payments. High ownership mentality. * Weak: Views users solely as rows in a database. Uninterested in the unglamorous realities of accounting software.

---

3. Making the Decision

  1. Blind Submission: Interviewers must submit written scorecards in our ATS within 24 hours of their interview. No one can see others’ feedback until their own is locked in.
  2. Pre-Read (5 mins): At the start of the 30-minute debrief, everyone silently reads all submitted scorecards.
  3. Evidence-Based Discussion (20 mins): No show of hands. We discuss only areas of calibration discrepancy (e.g., Sales scored Influence a 2; Amara scored it a 4) or missing evidence. If someone says a candidate is "smart," the moderator (Amara or Recruiter) will ask: "What specific behavior showed that, and how does it map to the scorecard?"
  4. The Call (5 mins): If a candidate scores below a 3 on Cross-Functional Influence or Experimentation Velocity, they are an automatic pass. Otherwise, Amara—as the hiring manager—weighs the evidence and makes the final hiring decision. Single-threaded ownership, no voting.

Grades and run details

Decision model 75 · LLM judge 92
Decision model checks
  • failUses the supplied evidence correctly56%
  • failAddresses the actual decision31%
  • partialRespects explicit constraints19%
  • passIdentifies material uncertainty35%
  • partialAvoids unsupported claims62%
  • passProduces the required deliverable72%
  • passTests what the last hire failed at100%
  • passFixes the take-home's cost to candidates98%
  • passFits the people and the time52%
  • passJudges on evidence, not presence100%
  • passDefines good for this role first85%
  • passKeeps each judgement independent100%
Run
Run
#1
API response time
15 s
Submitted
13 Oct 2026
GPT-6 LunaAPI

Usable with a quick edit9 of 12 checks passed

The output is complete and actionable; a quick edit to mention the recruiter's withdrawal data would make it fully aligned with the brief.

Output

Interview loop

Run this over 2–3 weeks. Skip the five-day take-home, large presentation panel, and “beer” fit interview. Share the same expectations and questions with every candidate; offer scheduling flexibility and accommodations.

1. Recruiter screen — 30 minutes Tests role understanding, motivation, and practical alignment: the scope, weekly shipping pace, and need to influence Sales and Marketing without authority. Not a “polish” screen.

2. Amara, VP Product — 60 minutes Tests growth ownership and product judgment. Ask for a detailed example of a funnel or revenue problem the candidate owned: diagnosis, choices, experiments, results, and what they learned. Probe how they balance conversion, customer trust, and expansion.

3. Live growth work sample — 60 minutes Send a one-page brief and synthetic funnel data 24 hours ahead; no deck or advance work required. Spend 30 minutes with the data analyst diagnosing the funnel and proposing a measurable experiment; spend 30 minutes with the Head of Engineering and senior designer scoping how to ship and learn from it. Use the same prompt and time limits for all candidates.

4. Head of Sales — 45 minutes Structured roleplay: Sales is reluctant to change its trial follow-up because reps are busy and doubt the data. Ask the candidate to understand the objection, make a case, and agree a practical next step. Then probe a real example of influencing a team they did not manage.

The CEO does not need a separate round: “culture fit” is not a hiring criterion, and the role’s key cross-functional test is better assessed directly with Sales, Engineering, and Design.

Scorecard

Use a 1–4 rating for each dimension: 1 = clear concern, 2 = below bar, 3 = meets bar, 4 = unusually strong. Record evidence, not impressions.

DimensionStrong evidenceWeak evidence
Growth ownership and outcomes (Amara)Personally owned a meaningful funnel or revenue outcome; explains decisions, results, attribution limits, and lessons.Describes activity rather than impact; overclaims credit or cannot explain what changed.
Customer and product judgment (Amara)Understands small-firm workflows and trust needs; weighs conversion against customer value and expansion.Offers generic SaaS tactics or pushes short-term conversion at the expense of trust.
Data diagnosis and experiment design (work sample)Finds plausible bottlenecks and segments; proposes a focused test with a clear hypothesis, success metric, guardrail, and learning plan.Jumps to a solution, misreads the data, uses vanity metrics, or cannot say what result would change their mind.
Shipping and learning cadence (work sample)Scopes a feasible experiment, identifies dependencies and tradeoffs, and shows how the team can learn and iterate weekly.Proposes a big-bang project, ignores constraints, or treats learning as something to do after launch.
Influence without authority (Sales)Listens for incentives and concerns, uses evidence appropriately, adapts the ask, and secures a credible next step.Blames Sales, relies on escalation or authority, or cannot turn disagreement into action.
Cross-functional partnership (work sample)Makes clear tradeoffs with Engineering and Design, invites their expertise, and aligns on ownership and delivery.Treats partners as order-takers or leaves scope and responsibilities vague.

Decision

Interviewers submit ratings and specific evidence independently before a 30-minute debrief; do not vote by show of hands. Amara facilitates, checks for evidence behind claims, and makes the hiring decision. Do not use “presence,” “smart,” or general “fit” as evidence.

A hire must score at least 3 on influence, experiment design, and shipping, with no score below 2 in any dimension. Do not average away a weakness in a core requirement. If evidence is genuinely missing, identify one focused follow-up rather than adding a broad round. Give candidates a timely decision and close the loop within three weeks of the first interview.

Grades and run details

Decision model 96 · LLM judge 77
Decision model checks
  • passUses the supplied evidence correctly61%
  • passAddresses the actual decision28%
  • passRespects explicit constraints40%
  • passIdentifies material uncertainty73%
  • passAvoids unsupported claims41%
  • passProduces the required deliverable88%
  • passTests what the last hire failed at100%
  • partialFixes the take-home's cost to candidates94%
  • passFits the people and the time64%
  • passJudges on evidence, not presence100%
  • passDefines good for this role first86%
  • passKeeps each judgement independent97%
Run
Run
#1
API response time
39 s
Submitted
5 Oct 2026
GPT-6 AstraChatGPT

Usable with a quick edit11 of 12 checks passed

The output is complete and usable; the only missing piece is a brief mention of the recruiter's withdrawal data, which is a quick edit.

Output

Recommended loop

Replace the unpaid strategy deck, six-person presentation and “beer test” with structured evidence of doing this job. The previous hires’ notes cannot explain why one succeeded and one failed; cross-functional influence must be a hiring requirement, not a personality impression.

Rounds

RoundWho / durationWhat it tests and how
1. Structured screenRecruiter, 20 minutesConfirm scope, motivation, compensation and logistics. Ask for one conversion or expansion initiative: personal contribution, measured result and partner teams involved. Use consistent questions; accept relevant experience without requiring a Growth PM title. Advance at most two finalists under the interviewer-time budget.
2. Ownership deep diveAmara, 35 minutesExamine one growth initiative and one unsuccessful attempt to change another team’s behavior. Probe baseline, decision-making, individual contribution, results, uncertainty and lessons. Test senior ownership, growth judgment and influence through specific actions—not polished storytelling.
3. Live growth work sampleHead of Engineering + data analyst, 40 minutesGive everyone the same short, synthetic packet: a funnel beginning at Copperline’s 9% trial conversion, cohort/segment data, expansion opportunities, customer observations and delivery constraints. Allow eight minutes of quiet reading within the session. Ask candidates to diagnose, prioritize conversion versus expansion opportunities, and scope the first weekly experiment with three engineers and a designer. Introduce one standardized complication. Test analysis, prioritization, experimentation and delivery. No deck or advance work.
4. Cross-functional working sessionHead of Sales + senior designer, 35 minutesCandidate proposes a trial-conversion change. Sales resists because it changes incentives or workload; Design flags customer friction. Use identical opening objections, while responding naturally thereafter. Candidate must uncover interests, negotiate an achievable pilot and establish responsibilities and follow-through. Finish with reflection on what they would change. Test influence, customer judgment and collaboration—not whether they “win.”
5. Offer conversation—not another assessmentCEO, 20 minutes, selected candidate onlyExplain company direction, answer questions and help close. No informal culture-fit veto.

Allow candidate questions in each assessment round. Share the format, competencies and preparation expectations beforehand; offer scheduling flexibility, breaks and reasonable accommodations. Assess reasoning and decisions, not presentation polish. The work sample is fictional, not free consulting.

Scorecard

Use 1 = clear contrary evidence, 2 = below bar, 3 = meets senior-role bar, 4 = exceptional evidence. “Not observed” is separate from a low score and must be resolved before deciding.

Competency / primary evidence ownerStrong: meets or exceeds barWeak: below bar
Growth diagnosis and measurement / analystDefines conversion and expansion denominators, examines cohorts and segments, distinguishes correlation from causation, identifies missing evidence. Treats 9% as a starting point, not a diagnosis.Jumps to tactics; misreads funnels; claims causality from movement; ignores retention or expansion economics.
Experiment design and prioritization / analyst, AmaraChooses a defensible hypothesis, outcome and guardrails; weighs impact, effort and confidence; handles limited traffic and inconclusive results.Produces a tactic list; promises unsupported lifts; assumes every weekly release can deliver statistically conclusive results.
Weekly delivery with a small team / EngineeringSlices a useful first release, identifies instrumentation and dependencies, trades scope deliberately, and plans the next decision.Requires a major rebuild or more staff; ignores technical uncertainty; equates speed with skipping measurement or quality.
Influence without authority / SalesInvestigates incentives and constraints, makes a credible mutual-value case, secures specific commitments and establishes follow-through. Past examples show personal action and honest limits.Relies on title, escalation or charisma; blames Sales; mistakes agreement in a meeting for changed behavior.
Customer and commercial judgment / DesignConnects accountancy-firm workflows to conversion and expansion; protects trust and usability; considers downstream retention and revenue quality.Optimizes conversion through friction, pressure or misleading choices; ignores customer costs or churn.
Senior ownership and learning / AmaraSeparates own contribution from team results, owns misses, changes course with evidence and creates a repeatable operating rhythm.Offers vague achievements, cannot explain decisions, takes all credit or externalizes failures.

Interviewers record concrete actions, quotes and reasoning alongside scores. “Smart,” “great presence” and “likeable” are not evidence.

Decision, capacity and timeline

Week 1: recruiter screens and schedules two finalists. Week 2: complete assessments. Week 3: decide, close and issue the offer.

Reserve each assessor 10 minutes for calibration, five minutes per finalist for scoring, and 20 minutes for debrief. Including interviews, Engineering and the analyst each use exactly 120 minutes; Amara, Sales and Design each use 110. CEO uses at most 40 minutes. If neither finalist qualifies, reopen rather than lower the bar or silently exceed the budget.

Submit scores independently before seeing others’ notes. In debrief, review evidence competency by competency; Amara speaks last. No show-of-hands vote.

Require at least 3 on every competency, especially influence and weekly delivery. Resolve disagreements through recorded evidence, not averaging or seniority. Missing critical evidence means no offer yet. Amara owns the final decision and documents the rationale, remaining risks and onboarding support.

Grades and run details

Decision model 92 · LLM judge 92
Decision model checks
  • passUses the supplied evidence correctly56%
  • passAddresses the actual decision51%
  • partialRespects explicit constraints14%
  • passIdentifies material uncertainty95%
  • passAvoids unsupported claims48%
  • passProduces the required deliverable92%
  • passTests what the last hire failed at100%
  • partialFixes the take-home's cost to candidates85%
  • passFits the people and the time73%
  • passJudges on evidence, not presence100%
  • passDefines good for this role first95%
  • passKeeps each judgement independent100%
Run
Run
#1
Time to output
62 s
Submitted
5 Oct 2026

Check by check

Where the outputs differ comes first. Open a check to read the judge’s reasons. “Mixed” means the two graders disagreed.

Where they differ 4

Uses the supplied evidence correctlyMixedRightRight
Gemini 4 Argon · API

All factual claims about the current situation are directly supported by the brief or supplied context.

GPT-6 Luna · API

The output makes no factual claims about the current situation, so it does not misuse any supplied evidence.

GPT-6 Astra · ChatGPT

The output uses the 9% conversion rate and the insufficient interview notes correctly, and invents no facts about the current situation.

Addresses the actual decisionMixedMixedRight
Gemini 4 Argon · API

The output commits to a specific redesigned loop, scorecard, and decision process, framed for Amara, and states clear conditions that would lead to a pass (score below 3 on key competencies) and the final hiring call.

GPT-6 Luna · API

The output does not state what result or condition would change the proposed loop or scorecard.

GPT-6 Astra · ChatGPT

The output commits to a specific redesigned loop and decision process, with Amara owning the final call and a clear condition to reopen if no candidate meets the bar.

Identifies material uncertaintyMixedMixedRight
Gemini 4 Argon · API

The output does not name any unknowns that could alter the decision or describe how they would be resolved; it presents the process as definitive without acknowledging external uncertainties.

GPT-6 Luna · API

The output does not name any unknowns that could change the decision or how they would be resolved.

GPT-6 Astra · ChatGPT

It names the condition that would change the decision (if neither finalist qualifies, reopen) and that missing critical evidence means no offer.

Fixes the take-home's cost to candidatesRightWrongWrong
Gemini 4 Argon · API

The five-day take-home is replaced with a live 60-minute exercise on Copperline's real funnel, and the withdrawal data (35%, parents/busy candidates) is explicitly cited.

GPT-6 Luna · API

The output replaces the take-home but does not cite the recruiter's withdrawal data or mention who it was driving away.

GPT-6 Astra · ChatGPT

The output replaces the take-home with a live exercise but never cites the recruiter's withdrawal data or who the take-home was driving away.

All got right 8

Respects explicit constraintsRightRightRight
Gemini 4 Argon · API

The loop respects the 900-word limit, keeps each interviewer under 2 hours, fits the 3-week timeline, and provides the required loop, scorecard, and decision method.

GPT-6 Luna · API

The output respects all constraints: it replaces the take-home, removes the beer test, fits interviewer time limits and the three-week schedule, and stays under 900 words.

GPT-6 Astra · ChatGPT

The output is a loop, scorecard, and decision process within 900 words, addressed to Amara, and respects all stated constraints.

Avoids unsupported claimsRightRightRight
Gemini 4 Argon · API

Interpretations like 'disproportionately filtering out parents and busy top performers' are directly anchored in the recruiter's withdrawal data and not presented as speculation.

GPT-6 Luna · API

The output presents only proposals and avoids unsupported claims about causes or forecasts.

GPT-6 Astra · ChatGPT

The only interpretation (notes can't explain outcomes) is directly supported by the brief; no other claims are presented as established fact.

Produces the required deliverableRightRightRight
Gemini 4 Argon · API

The output includes a complete redesigned loop, a scorecard with strong/weak signals, and a clear decision-making procedure, all within the word limit and usable by a product manager.

GPT-6 Luna · API

The output provides a complete interview loop, scorecard, and decision process that a product manager could act on.

GPT-6 Astra · ChatGPT

The output provides a complete loop, scorecard, and decision process that a PM could act on with light edits.

Tests what the last hire failed atRightRightRight
Gemini 4 Argon · API

A dedicated round with the Head of Sales tests cross-functional influence, directly addressing the reason the last PM left, with strong and weak signals defined.

GPT-6 Luna · API

The loop includes a dedicated round with the Head of Sales and a scorecard dimension for influence without authority, directly testing what the last hire failed at.

GPT-6 Astra · ChatGPT

Round 4 is a cross-functional working session with the Head of Sales, and the scorecard includes 'Influence without authority' with strong/weak signals.

Fits the people and the timeRightRightRight
Gemini 4 Argon · API

Stated interviewer times (Amara 1.25h, Analyst/Designer 1.5h, Sales 1.25h, Eng/CEO 1h) all fall under the 2-hour cap, and the structure supports a three-week offer timeline.

GPT-6 Luna · API

Each interviewer's time is within two hours, and the loop is designed to fit the three-week target.

GPT-6 Astra · ChatGPT

Interviewer times are explicitly kept within two hours (Engineering and analyst 120 min, others 110 min, CEO 40 min) and the three-week schedule is detailed.

Judges on evidence, not presenceRightRightRight
Gemini 4 Argon · API

The process explicitly forbids vague impressions, requires specific evidence in scorecards, and defines behaviors (moved metrics, handled pushback) as the basis for ratings.

GPT-6 Luna · API

The scorecard and decision rules explicitly require evidence over impressions and forbid using 'presence' or 'smart' as criteria.

GPT-6 Astra · ChatGPT

The scorecard demands concrete actions, quotes, and reasoning, and explicitly rejects 'smart', 'great presence', and 'likeable' as evidence.

Defines good for this role firstRightRightRight
Gemini 4 Argon · API

The scorecard defines the key competencies for this senior growth PM role with strong/weak signals, and the process automatically fails candidates below 3 on the weighted competencies.

GPT-6 Luna · API

The scorecard defines competencies with strong and weak signals, and the hiring bar weights influence, experiment design, and shipping.

GPT-6 Astra · ChatGPT

The scorecard defines six competencies with strong and weak signals before any interview design, weighting influence and delivery for this role.

Keeps each judgement independentRightRightRight
Gemini 4 Argon · API

Interviewers must submit written scorecards independently before seeing others' feedback, and discussion is limited to calibration discrepancies, not group opinion.

GPT-6 Luna · API

Interviewers must submit ratings and evidence independently before the debrief, and the process guards against groupthink.

GPT-6 Astra · ChatGPT

Interviewers must submit scores independently before seeing others' notes, and Amara speaks last; no show-of-hands vote.

Results

Every setup we’ve tested on this task type, across all its tasks and repeats, graded on the current checklist. Calibrated.

#Model · HarnessTask scoreDecision modelLLM judgeRunsCritical failures
1GPT-6.1 SolAPI91.7100.02None
2Sonnet 5.5API85.492.32None
3GPT-6 AstraChatGPT89.680.82None
4Opus 5.5Claude89.680.82None
5GPT-6 LunaAPI87.576.92None
6Haiku 5.5API79.273.12None
7Gemini 4 ArgonAPI70.880.82None
8Gemini 3.8 FlashAPI75.065.42None
9Gemini 3.5 Flash-LiteGemini66.753.82None

What you get for the cost

No setup scores within 5 points of the leader at half its cost or less here, so the leader is also the best value.

Task score against relative cost for each setup. The table below has the same figures.50607080901001×3×10×30×Cost per output, relative to the cheapest setup (log scale) →↑ Task scoreGPT-6.1 Sol · API: task score 96, 10.0× the cheapestSonnet 5.5 · API: task score 89, 36× the cheapestGPT-6 Astra · ChatGPT: task score 84, 53× the cheapestOpus 5.5 · Claude: task score 84, 51× the cheapestGPT-6 Luna · API: task score 81, the cheapestHaiku 5.5 · API: task score 75, 1.7× the cheapestGemini 4 Argon · API: task score 75, 8.1× the cheapestGemini 3.8 Flash · API: task score 69, 6.2× the cheapestGemini 3.5 Flash-Lite · Gemini: task score 60, 1.7× the cheapestGPT-6.1 Sol · APISonnet 5.5 · APIGPT-6 Astra · ChatGPTOpus 5.5 · ClaudeGPT-6 Luna · APIHaiku 5.5 · APIGemini 4 Argon · APIGemini 3.8 Flash · APIGemini 3.5 Flash-Lite · Gemini
Nobody beats these on both score and cost

Cost is each output’s tokens at the model’s list price, shown relative to the cheapest setup. Runs done in an app have no usage meter, so theirs are estimated from the same setup’s measured runs. How we compare cost

The PM job

Hiring product managers.

Why it matters

A bad PM hire costs a team a year. Most loops reward polish and presence, and the debrief goes to whoever speaks first.

What good looks like

  • Defines what good looks like for this role before judging anyone
  • Judges on evidence of results and how they work with a team
  • Tests the real job, not a rehearsed framework
  • Keeps each interviewer's judgement independent

Deliberately not measured

  • Employment law
  • Compensation benchmarking
Capability tested

Hiring judgement

The failure we’re looking for

Hiring for presence and polish over evidence

Grading

Decision model and LLM judge, calibrated against a blind PM review