PORSCHE 991.2 / BLIND APEX 15

911 Spectrum

Stop moving sliders. Make choices and let the car reveal your priorities.

Relative comparisonSelected-set normalization
Differences intentionally exaggeratedMouse, tap, or arrow keys
A radar chart comparing selected Porsche 911 variants.
Guided comparison

One metric. Every car. No clutter.

Pin a reference, cycle the photography, or let the story play itself.

THE COMPARISON

Metric

01 / 11

Performance map

Where is the smart compromise?

Every direction means better. The frontier contains cars that no rival beats on both axes.

PARETO FRONTIERHOVER · TAP · CLICK TO PIN
All positions use the currently selected cars as the comparison set.
Preference stress test

Will your choice survive changing tastes?

The current priority mix is tested against 4,096 nearby preference worlds in a dedicated worker.

Building the first forecast…

The simulation worker is warming up.

—
CALCULATINGLEADER WIN SHARE
03

Outcome tape

Each sliver is a representative preference world. A striped sliver means the leaders tied.

04

Garage forecast

WIN SHARE · RANK RANGE
05

Smallest verdict flips

CLICK TO APPLY

The lightest single priority move that lets each challenger pass the current leader.

4,096 WORLDS · DETERMINISTIC MODELRe-roll changes the sampled worlds, never the underlying vehicle data.
Weighted similarity atlas

What is genuinely closest to this 911?

The anchor stays at the center. Alternatives move outward as their weighted character becomes more distinct.

WEIGHT MODEL
Building the similarity atlas…

Reading the active garage and priority mix.

06

Similarity ledger

TOP SOURCES OF DISTANCE
CURRENT TUNED PRIORITIESDistances use selected-set normalization; tied specifications create no artificial separation.
Persistent comparison journal

Keep the decision—not just the winner.

Save complete garages, priorities, maps, forecasts, themes, and constellation anchors locally on this device.

0 SAVED SETUPSStored privately in this browser with IndexedDB.
OFFLINE SHELL · PREPARINGPWA · INSTALLABLECTRL / ⌘ + K · COMMANDS
07

Saved setups

A / B ASSIGNS THE DIFF
08

Decision diff

→
LOCAL-FIRST · NO ACCOUNT · NO SERVER UPLOAD
Quick command

Go anywhere. Change anything.

↑↓ NAVIGATEENTER RUNSESC CLOSES
Cross-generation evidence

Five Carreras. Fourteen years. One bloodline.

Factory specifications are kept separate from the 991.2 variant radar, so a model-year comparison never pretends different generations share the same subjective evidence.

THE TURBO PIVOT
2017 · 991.2370 HP
3.0 L twin-turbo flat-six7-speed PDK

09

Evolution trace

DIFFERENCES EXAGGERATED
Reading the lineage…

Selected-set min/max normalization makes small real differences legible. Equal values stay neutral.

CONFIGURATION RULEBase Carrera coupés · PDK · Sport Chrono acceleration where published · Factory curb/unladen definitions vary by market
FACTORY SOURCES
PHOTO CREDITS ↗
Pairwise evolution scanner

Put any two generations under the glass.

The image seam is tactile; the verdict is evidence-based. Directional metrics call gains and regressions, while displacement remains contextual.

PHOTOGRAPHY
DRAG · TAP · ← →
10

Then / now reveal

EXTERIOR VIEW
BEFORE2011 · 997.2345 HP
AFTER2025 · 992.2388 HP
11

Mutation ledger

CLICK A ROW TO PIN IT
14-YEAR ENGINEERING JUMPReading the generations…

0GAINS
0REGRESSIONS
0CONTEXT / TIES
Positive and negative directions follow the same inverse-metric rules as the radar. Equal values never receive a false winner.PAIR STATE · URL SYNCED
Complete garage cross-examination

Every rivalry. No buried trade-offs.

Each unique matchup appears once. Open one to inspect all eleven metrics with selected-set colors, inverse-aware winners, explicit ties, and proportionally exaggerated gaps.

SCOPE
SELECTED GARAGE0 CARS · 0 MATCHUPS
12

All-pairs field

EACH PAIR APPEARS ONCE
13

Rivalry portrait

LEFT—
RIGHT—
DIFFERENCES ARE EXAGGERATED IN THE BATTLE TAPEEXTERIOR PHOTOGRAPHY
14

Battle tape

HOVER FOR SERIOUS + FUN CONTEXT
READING THE GARAGESelect a rivalry to begin.
0LEFT WINS
0TIES
0RIGHT WINS
Pair scores count category wins, not raw magnitude. Ties are neutral; inverse categories reward lower values.RIVALRY STATE · URL SYNCED
Exhaustive collection optimizer

Draft a garage, not another compromise.

Every valid combination is scored. Coverage rewards the best car for each metric; complementarity rewards lineups whose members contribute different strengths and mechanical characters.

GARAGE SIZE
REQUIRED TRAITS
36COMBINATIONS SEARCHED
36PASS THE BRIEF
20FIT THE BUDGET
EXHAUSTIVENO RANDOM SAMPLING
15

Winning lineup

SEARCHING EVERY COMBINATIONBuilding the garage…

$0k SPENT$350k REMAINS
#1 OF 20 VALID GARAGES
16

Alternative garages

CLICK ANY LINEUP TO INSPECT IT
17

Coverage ownership

BEST MEMBER PER METRIC · GLOBAL POOL NORMALIZATION
COVER EVERY ROADBalanced collection

0.0COVERAGE
0.0COMPLEMENTARITY
0.0DRAFT SCORE
Prices use the dataset’s original U.S. MSRP figures; taxes, options, and market premiums are intentionally excluded.DRAFT STATE · URL SYNCED
Adaptive preference decoder

Twelve honest choices. No weight sliders. One unnervingly personal 911 ranking.

The learner asks the most informative unseen matchup next, fits a transparent category fingerprint, and keeps the vehicle identities masked until you decide to reveal them.

CARD IDENTITY
PHOTOGRAPHY
HISTORY
QUESTION 01 OF 12Which brief would you rather live with?Exact values stay visible. Names and recognizable photography do not.
0%MODEL CONFIDENCE
0ANSWERED
36UNSEEN PAIRS
18

Current taste test

ADAPTIVE PAIR · MAXIMUM INFORMATION GAIN
Masked Porsche 911 candidate A
CANDIDATE A · IDENTITY MASKEDCANDIDATE AONE OF NINE 991.2 VARIANTS
Masked Porsche 911 candidate B
CANDIDATE B · IDENTITY MASKEDCANDIDATE BONE OF NINE 991.2 VARIANTS
20

Your inferred podium

0.0POINTS BETWEEN #1 AND #2
0ACTIVE PREFERENCE SIGNALS
0NON-SKIPPED DECISIONS

Confidence reflects choice consistency and sample depth—not certainty about your soul.

21

Decision tape

LATEST FIRST · URL-SYNCED
NO TAPE YETYour first choice will appear here with its effect on the learner.
The model uses normalized category directions and regularized logistic learning. A tie creates no fake preference; a skip teaches nothing and never repeats the pair.BLIND APEX · URL SYNCED
Priority mixer

Build your own brief

PRESET SYNCED
01

Priority channels

0 mutes · 5 dominates
02

Advantage ledger

VS
Tuned shortlist

Your priorities, ranked

LIVE MODEL