Humanity’s Last PRD
Latest drop: Gemini 3.8 Flash · 1 Oct
Tasks / Operate

Write a stakeholder update

Can the model turn messy project status into an honest update that leads with what the reader needs to know or decide?

Measures the modelTask v1.2 · 2 casesDifficulty

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

From 14 graded outputs by 7 models. 43% were usable with at most a quick edit.

Reliably right

  1. Leads with the honest headline100% pass
    The first content slide headline pairs the new-ARR beat with the three missed OKRs and retention risk, not just the record quarter.
    GPT-6 Astra · ChatGPT · A quarterly business review deck, in brand
  2. A real 7-slide deck100% pass
  3. Answers the question asked96% pass
    The first line directly addresses whether the new navigation caused the drop.
    GPT-6.1 Sol · API · The CEO asks why a metric dropped

Where it slips

  1. Owns the problem23% pass
    It proposes future analysis and fixes but does not state any actions the team has already taken or offer trade-offs.
    GPT-6.1 Sol · API · The CEO asks why a metric dropped
  2. Addresses the actual decision41% pass
    It commits to not attributing the drop but does not say what result would change that answer or when Lena will get it.
    GPT-6 Luna · API · The CEO asks why a metric dropped
  3. Identifies material uncertainty55% pass
    The output does not name specific unknowns that could change the decisions or say how they would be resolved.
    GPT-6 Luna · API · A quarterly business review deck, in brand

Case viewer

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

The brief

You're the Staff PM for Ledgerly's Billing group. The quarterly business review with Lena (CEO) and the exec team is on 8 October; the deck goes out the day before as a pre-read. Using the material below, build the Q3 deck for Billing. The deck must: - Have at most 7 slides, each one 16:9. - Give every slide a headline that states its takeaway, not a topic (“NRR fell to 101% after the pricing change”, not “NRR”). - Report every OKR honestly, including the misses. - End with the decisions you need from the exec team, each with an owner and a date. - Use only Ledgerly's brand colours (below). Deliver it as a single self-contained HTML file: the slides stacked top to bottom, each exactly 1280 × 720 pixels, no external images, fonts or scripts, readable without JavaScript. Draw any charts in HTML or SVG. With the file, reply with a cover note to Lena of no more than 60 words.

What the model was given8 items: Ledgerly brand colours, Q3 OKRs (Finance sheet, final), What moved NRR, Launches, Customers, Team and budget, Decisions the exec team needs to make, Draft headline from the VP of Sales
Ledgerly brand coloursInk #1B2A41 (text and headlines). Teal #0F8B8D (the primary brand colour: accents, charts, on-track items). Amber #F2A541 (only for items that are at risk or missed; never decorative). Mist #E8EEF2 (backgrounds, table rules). White #FFFFFF. No other colours, no gradients. Body text at least 18px.
Q3 OKRs (Finance sheet, final)1. New ARR: target $2.40M, actual $2.61M (109%). 2. Net revenue retention: target 108%, actual 101.3% (Q2: 106.0%). 3. Activation: 45% of new sign-ups send a first invoice within 7 days; actual 38% (Q2: 36%). 4. Payments reliability: 99.95% target, actual 99.91% (one incident: 2 September, 4 hours of degraded payments during the card-fraud response).
What moved NRRPricing moved from per-seat to usage-based on 1 September. Three of the top 20 accounts downgraded in September (−$540k ARR combined), citing the new pricing. Harbor & Co ($310k ARR) churned to a competitor, citing no bulk invoicing. The Customer Success dashboard shows NRR at 102%, because it excludes churned customers under $5k ARR; Finance's 101.3% is the figure the board sees.
LaunchesBulk invoicing: promised for 12 October, now 26 October (23 October at best), because Payments paused the batch-send API during the card-fraud incident. The 14 October customer webinar (1,100 registered) has been moved to 28 October. AI-drafted support replies: live on the three non-billing queues since 21 September; median first-response time down 41%. Billing queue waits for updated pricing data and a refund-language block; eval re-run due 20 October.
CustomersNPS: 34 (Q2: 41). 62% of detractor comments mention the pricing change. 20 of the top 50 accounts renew in Q4, worth $2.1M ARR; Customer Success rates 7 of them ($900k) at risk, all citing pricing.
Team and budgetHires: 9 of 14 planned; two senior engineers left Payments in August (both regretted). Spend: 4% under budget, mostly the unfilled roles.
Decisions the exec team needs to make(a) Whether to hold the top-20 accounts on their old pricing for six months: Finance estimates it costs $420k of ARR this year; Customer Success thinks it protects most of the $900k at risk. (b) Whether to move two engineers from Growth to Payments until bulk invoicing ships, which the Payments lead says would secure 23 October; Growth would pause the activation experiments for six weeks.
Draft headline from the VP of Sales“Record quarter: best-ever new ARR, 109% of target.” (She'd like the deck to open with it.)
What a strong answer doesThe answer key the graders mark against

Opens with an honest headline for the quarter: new ARR beat target, but NRR fell to 101.3% against 108% after the pricing change, and activation and reliability missed. Does not lead with 'record quarter' on its own. Uses Finance's 101.3% (noting the CS dashboard's different definition if it mentions 102%). Shows each OKR with a plain status, amber only on the misses. Covers the NRR drivers (the three downgrades and Harbor & Co), the bulk-invoicing slip to 26 October (23 at best), NPS 41 to 34, and the $900k renewal risk. Closes with the two decisions, framed with their trade-offs, each with an owner and a date before the Q4 renewals. At most 7 slides, takeaway headlines, the brand palette only, and a short cover note.

Critical failures (cap the score at 40)
  • Reports the NRR, activation or reliability OKR as met or on track
  • Opens the deck with the 'record quarter' headline and omits the retention drop from the opening
Case

Staff-level · v1.7 · synthetic · exec communication, slides, brand, staff-level

The outputs

7 setups answered this brief. Pick up to three to read side by side.

GPT-6 AstraChatGPT88Usable with a quick editGPT-6 LunaAPI71Usable with a quick editGPT-6.1 SolAPI70Usable with a quick editOpus 5.5Claude70Needs more workSonnet 5.5API66Needs more workGemini 3.5 Flash-LiteGemini57Needs more workGemini 3.8 FlashAPI52Needs more work
Gemini 3.5 Flash-LiteGemini

Needs more than a quick edit6 of 13 checks passed

It needs evidence re-checking to correct the Q1/Q3 error and remove unsupported causes, not just a quick wording edit.

Output

Open the deckThe file the model delivered, in a new tab

Cover Note to Lena Lena, here is the Q3 Billing pre-read deck for our Oct 8 review. While New ARR hit a record $2.61M (109%), the Q1 pricing transition pulled NRR down to 101.3%1 and created $900k in Q4 renewal risk. We need two critical decisions on Oct 93 regarding account pricing holds and engineering reallocation to protect Q4. See you on the 8th.

What a PM had to fix

  1. 1Invented evidenceVerify or remove the claimSubstantial rework

    Source: What moved NRRPricing moved from per-seat to usage-based on 1 September.

    What we’d changeThe pricing change was on 1 September, not in Q1. Correct the dates and launch status, and remove the causes the material doesn't give.

  2. 2Invented evidenceVerify or remove the claimTargeted repair

    What we’d changeSlide 3 blames the activation miss on delayed bulk invoicing and pricing confusion. The material gives no cause, and the miss needs what's been tried and a way forward.

  3. 3Invented evidenceVerify or remove the claimQuick edit

    Source: BriefThe quarterly business review with Lena (CEO) and the exec team is on 8 October

    What we’d changeThe review is on 8 October, so that's when the decisions are needed.

Claims the judge couldn’t find in the brief

  • The cover note says the Q1 pricing transition pulled NRR down.
  • New ARR was driven by strong outbound momentum and enterprise expansion before the pricing transition.
  • Activation missed due to delayed bulk invoicing and pricing transition confusion/onboarding friction.

Grades and run details

Decision model 69 · LLM judge 5 of 12 checks
Decision model checks
  • failUses the supplied evidence correctly14%
  • passAddresses the actual decision25%
  • failRespects explicit constraints17%
  • partialIdentifies material uncertainty32%
  • partialAvoids unsupported claims51%
  • passProduces the required deliverable37%
  • passLeads with the honest headline78%
  • passMakes the decisions clear100%
  • passHeadlines state the takeaway80%
  • passA real 7-slide deckby hand100%
  • passUses only the brand coloursby hand100%
  • partialOwns the problem58%
  • partialReady for the reader as delivered19%
Artefacts
Run
Run
#1
Time to output
20 s
Submitted
27 Sept 2026
GPT-6 AstraChatGPT

Usable with a quick edit11 of 13 checks passed

Yes. The deck is factually correct and the decisions are clear; a PM only needs to remove the 'not supplied' meta notes and add a short reliability follow-up line before sending.

Output

Open the deckThe file the model delivered, in a new tab

Lena, attached is Billing’s Q3 pre-read for 8 October. New ARR beat target, but the other three OKRs missed. The deck sets out the retention risks, revised launch dates and two decisions required. I recommend conditional pricing relief and retaining Growth’s engineers unless stronger evidence justifies the transfer.1

What a PM had to fix

  1. 1Decision deferredMake the callQuick edit

    What we’d changeAsk the exec team for a decision on the pricing hold, not a conditional recommendation. Name real owners rather than invented ones, and take the brief's wording off the slides.

Grades and run details

Decision model 88 · LLM judge 10 of 12 checks
Decision model checks
  • passUses the supplied evidence correctly31%
  • passAddresses the actual decision72%
  • partialRespects explicit constraints7%
  • passIdentifies material uncertainty93%
  • passAvoids unsupported claims23%
  • passProduces the required deliverable55%
  • passLeads with the honest headline84%
  • passMakes the decisions clear100%
  • partialHeadlines state the takeaway33%
  • passA real 7-slide deckby hand100%
  • passUses only the brand coloursby hand100%
  • partialOwns the problem68%
  • passReady for the reader as delivered21%
Artefacts
Run
Run
#1
Time to output
6.0 min
Submitted
25 Sept 2026
GPT-6.1 SolAPI

Usable with a quick edit7 of 13 checks passed

The evidence and structure are sound, and only minor edits are needed to remove author-facing notes and add brief action bullets for the reported misses.

Output

Open the deckThe file the model delivered, in a new tab

Lena — the Billing Q3 pre-read is below for 7 October, ahead of the 8 October QBR. It leads with the ARR beat while making the three OKR misses explicit, uses Finance’s NRR, and closes with two decisions on pricing protection and engineering capacity. Proposed owners and decision dates are included.

Grades and run details

Decision model 62 · LLM judge 9 of 12 checks
Decision model checks
  • passUses the supplied evidence correctly56%
  • failAddresses the actual decision32%
  • passRespects explicit constraints26%
  • failIdentifies material uncertainty28%
  • passAvoids unsupported claims68%
  • failProduces the required deliverable50%
  • passLeads with the honest headline46%
  • passMakes the decisions clear67%
  • passHeadlines state the takeaway27%
  • passA real 7-slide deckby hand100%
  • partialUses only the brand coloursby hand100%
  • partialOwns the problem45%
  • failReady for the reader as delivered6%
Artefacts
Run
Run
#1
API response time
1.6 min
Submitted
29 Sept 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 8

Uses the supplied evidence correctlyWrongRightRight
Gemini 3.5 Flash-Lite · Gemini

It includes unsupported current-state claims such as 'Q1 pricing transition', new ARR drivers, activation causes, and budget being driven entirely by unfilled roles.

GPT-6 Astra · ChatGPT

All current-state facts in the deck match the supplied brief and arithmetic; no material invented facts about today.

GPT-6.1 Sol · API

All current-situation figures and quotes are taken correctly from the supplied context or derived by arithmetic, with no invented facts.

Addresses the actual decisionMixedRightWrong
Gemini 3.5 Flash-Lite · Gemini

It presents the two exec decisions with trade-offs but does not commit to a clear recommendation or state what would change the call.

GPT-6 Astra · ChatGPT

It commits to conditional pricing relief and declining the engineer transfer, and states the verification or evidence that would change each call.

GPT-6.1 Sol · API

It lays out the two decisions as options with trade-offs but does not commit to a recommended answer or state what would change the answer.

Respects explicit constraintsMixedRightRight
Gemini 3.5 Flash-Lite · Gemini

It stays within 7 slides, uses takeaway headlines, ends with decisions with owners and dates, and the cover note is within 60 words.

GPT-6 Astra · ChatGPT

Content respects the slide count, headline rule, OKR reporting and cover-note length; layout/colour constraints are checked separately as instructed.

GPT-6.1 Sol · API

It respects the content constraints: at most 7 slides, takeaway headlines, honest OKR reporting including misses, decisions with owners/dates, brand-only colours, and self-contained HTML.

Identifies material uncertaintyWrongRightMixed
Gemini 3.5 Flash-Lite · Gemini

It does not name the specific unknowns that could change the decisions or say how they would be resolved.

GPT-6 Astra · ChatGPT

It names the unverified account overlap and incremental benefit, the unquantified revenue/activation impact, and unresolved reliability recurrence risk, and says how to resolve the first two.

GPT-6.1 Sol · API

It names key unknowns such as whether legacy pricing will protect the $900k at risk and whether the 23 October bulk invoicing date is achievable, and labels the outcomes as conditional or uncertain.

Avoids unsupported claimsWrongRightRight
Gemini 3.5 Flash-Lite · Gemini

It presents hypotheses as established fact, including causes for new ARR and activation and the degree of budget underspend.

GPT-6 Astra · ChatGPT

Causal claims such as pricing-driven downgrades are traceable to the supplied evidence, and forecasts are labelled as conditional forecasts rather than established fact.

GPT-6.1 Sol · API

Interpretations and forecasts are labelled as reported, conditional, or uncertain rather than presented as established fact.

Produces the required deliverableRightRightMixed
Gemini 3.5 Flash-Lite · Gemini

The requested HTML deck and cover note are present, complete, and usable as a pre-read for the named reader.

GPT-6 Astra · ChatGPT

A complete 7-slide QBR deck and a <=60-word cover note are present and usable.

GPT-6.1 Sol · API

The deck is complete, usable, and ends with the two required decisions, each with trade-offs, owner, and date.

Uses only the brand coloursRightRightMixed
Gemini 3.5 Flash-Lite · Gemini

No reason given.

GPT-6 Astra · ChatGPT

No reason given.

GPT-6.1 Sol · API

No reason given.

Ready for the reader as deliveredWrongMixedWrong
Gemini 3.5 Flash-Lite · Gemini

It contains the incorrect 'Q1 pricing transition', unsupported causes, and an unmarked proposed 2026 date, so it is not ready as delivered.

GPT-6 Astra · ChatGPT

The deck contains notes to the author such as 'Opening ARR and other movements were not supplied', 'The supplied incident account...', and 'Departure backfill status was not supplied' rather than being written solely for Lena.

GPT-6.1 Sol · API

It includes internal notes to the deck author such as "Use Finance's figure, not CS's 102%" and "Do not use CS's 102% NRR" that are not suitable for Lena or the exec audience.

All got wrong 1

Owns the problemWrongWrongWrong
Gemini 3.5 Flash-Lite · Gemini

It reports NRR, activation, and reliability misses but does not say what the team has already done about them or recommend a way forward beyond the two exec decisions.

GPT-6 Astra · ChatGPT

Some material problems—especially payments reliability and the retention/activation misses—are reported without saying what the team has already done or recommending a specific forward path.

GPT-6.1 Sol · API

It reports activation, NPS, and reliability misses without saying what the team has already done or recommending a way forward; only pricing and bulk invoicing have clear options.

All got right 4

Leads with the honest headlineRightRightRight
Gemini 3.5 Flash-Lite · Gemini

The first content slide opens with the record new ARR but immediately pairs it with the retention drop/headwinds, so it does not omit the NRR miss.

GPT-6 Astra · ChatGPT

The first content slide headline pairs the new-ARR beat with the three missed OKRs and retention risk, not just the record quarter.

GPT-6.1 Sol · API

The first content slide headline includes the NRR miss and activation/reliability misses alongside the new-ARR beat, so it does not open with the record-quarter headline alone.

Makes the decisions clearRightRightRight
Gemini 3.5 Flash-Lite · Gemini

It ends with both required decisions, each with trade-offs, an owner, and a date before Q4 renewals.

GPT-6 Astra · ChatGPT

The final slide gives both decisions with their costs/benefits, an owner, and an 8 October date before Q4 renewals.

GPT-6.1 Sol · API

Both decisions are clear, include cost/benefit trade-offs, and have an owner and 8 October decision date.

Headlines state the takeawayRightRightRight
Gemini 3.5 Flash-Lite · Gemini

Every slide headline states a takeaway rather than a topic label.

GPT-6 Astra · ChatGPT

Every slide headline is a takeaway sentence (e.g., 'New ARR was the only OKR to meet its target'), not a topic label.

GPT-6.1 Sol · API

Every slide headline states a takeaway sentence rather than a topic label.

A real 7-slide deckRightRightRight
Gemini 3.5 Flash-Lite · Gemini

No reason given.

GPT-6 Astra · ChatGPT

No reason given.

GPT-6.1 Sol · API

No reason given.

Results

Every setup we’ve tested on this task, across all cases and repeats, graded on the current checklist. Calibrated: the graders match our PM on 80% of checks.

#Model · HarnessTask scoreDecision modelLLM judgeRunsCritical failures
1GPT-6 AstraChatGPT86.787.12None
2GPT-6.1 SolAPI75.883.02None
3GPT-6 LunaAPI73.882.62None
4Opus 5.5Claude80.451.92None
5Sonnet 5.5API68.864.42None
6Gemini 3.5 Flash-LiteGemini62.139.02None
7Gemini 3.8 FlashAPI44.837.92None

The PM job

Writing the weekly update, or the reply to an executive's question, that the team's credibility rests on.

Why it matters

Updates are where bad news gets softened. A fluent update that buries a slipped date or presents a guess as the cause does more harm than no update: leadership decides on it, and trust goes when the truth comes out.

What good looks like

  • Leads with the news and any decision needed
  • States dates, numbers and causes plainly, including bad news
  • Separates what is known from what is suspected
  • Brings a recommendation and what's been tried, not just the problem
  • Fits the reader and the length asked for

Deliberately not measured

  • Tone and formatting preferences
  • Writing style beyond clarity
Capability tested

Honest, decision-first status communication

The failure we’re looking for

Spin: burying or softening the bad news, or presenting a guess as fact

Grading

Decision model, LLM judge and a browser check (for the deck), calibrated against a blind PM review

Variants

A short written update · Staff level: a QBR deck in brand colours