Muse MuseConcept

Independent concept built on Meta's Model API. Not an official Meta product.

Spark teaches Glimmer: brief writer

Muse Glimmer LoRA · Muse Glimmer 30B · finished

The run

Recipe
Muse Glimmer LoRA
Model
Muse Glimmer 30B
GPU
1x A100 SXM4 (40 GB)
GPU time
48 min
GPU cost
$1.59
Training steps
120
LoRA rank
16
Learning rate
0.0002

Both models answered each held-out prompt. Muse Spark 1.3 saw the two answers in random order, without knowing which model wrote which.

Eval results

40 held-out prompts

Tuned Glimmer beat base on 11 of 40, base won 28, ties 1

Average score base 6.8, tuned 3.6, judged by Muse Spark 1.3.

Scores run from 1 to 10 against Muse Spark 1.3's reference answer. Judging cost $0.18.

Each held-out prompt with Muse Spark 1.3's reference, the base and the tuned Glimmer's answers, and the verdict
Muse Spark 1.3 referenceBase GlimmerTuned GlimmerVerdict
Prompt 1

need help counting seagulls in this short video from the pier

**Name:** Pier Gull Counter **Tagline:** Count every seagull in your pier clip. **Flow:** 1. Upload your short video from the pier. 2. Confirm time range to scan. 3. Get total count with outlined bir…Show allShow less

Name: Pier Gull Counter
Tagline: Count every seagull in your pier clip.
Flow:

  1. Upload your short video from the pier.
  2. Confirm time range to scan.
  3. Get total count with outlined birds and timestamps.
    Models:
  • SAM 3.1 for finding, outlining and counting objects in photos and video
  • Muse Spark 1.3 for text and image understanding and structured answers
    Limits:
  • May miss distant or heavily occluded gulls.
  • Fast flight can cause double-counts.
  • Short clips under 2 minutes work best.
    Cost per run: $0.02
**Name:** Pier Seagull Counter **Tagline:** Count seagulls in your pier video with frame-by-frame tracking. **Flow** 1. Upload short pier video and set start/end time range. 2. Review per-frame det…Show allShow less

Name: Pier Seagull Counter

Tagline: Count seagulls in your pier video with frame-by-frame tracking.

Flow

  1. Upload short pier video and set start/end time range.
  2. Review per-frame detections and adjust confidence threshold.
  3. Get total unique seagull count and export summary.

Models

  • SAM 3.1 for finding, outlining and counting objects in photos and video — detect and track seagulls across video frames.
  • Muse Spark 1.3 for text and image understanding and structured answers — interpret your request and return structured count results with explanation.

Limits

  • Counting accuracy drops with heavy overlap, motion blur, or small distant birds.
  • No species verification beyond visual similarity to seagulls.
  • Max video length and resolution limited by processing budget.

Cost per run: ~$0.015

Name: Pier Gull Counter Tagline: Count every seagull in your pier video. Flow: 1. Upload your short pier video. 2. Review auto-outlined gulls frame by frame. 3. Get final count with timestamps. Model…Show allShow less

Name: Pier Gull Counter
Tagline: Count every seagull in your pier video.
Flow:

  1. Upload your short pier video.
  2. Review auto-outlined gulls frame by frame.
  3. Get final count with timestamps.
    Models:
  • SAM 3.1 for finding, outlining and counting objects in photos and video
  • Muse Spark 1.3 for text and image understanding and structured answers
    Limits:
  • Misses heavily occluded or fast-flying gulls.
  • Poor in dark, blurry, or crowded water.
    Cost per run: $0.012
Tie

Base 9 · Tuned 9

Both give correct models, three step flows and honest limits; remaining differences are only phrasing detail.

Prompt 2

need packing list app from this messy closet pic

Name: Pack From Closet Tagline: Turn that messy closet pic into a trip-ready packing list. Flow: 1. Upload your closet photo and enter destination, days, and weather. 2. Review auto-detected tops,…Show allShow less

Name: Pack From Closet

Tagline: Turn that messy closet pic into a trip-ready packing list.

Flow:

  1. Upload your closet photo and enter destination, days, and weather.
  2. Review auto-detected tops, bottoms, shoes, and layers.
  3. Check off items to get your final packing checklist.

Models:

  • SAM 3.1 for finding, outlining and counting clothes in the photo
  • Muse Spark 1.3 for image understanding and structured packing answers
  • Muse Wonder 1.0 for long multi-step trip planning

Limits:

  • Only sees visible items, not inside drawers or laundry bins
  • Can mix up similar colors or layered clothes in clutter

Cost per run: $0.012

**Name:** Closet Pack Planner **Tagline:** Turn a messy closet photo into a smart packing list. **Flow** 1. User uploads a messy closet photo and enters trip details like destination and days. 2. Ap…Show allShow less

Name: Closet Pack Planner
Tagline: Turn a messy closet photo into a smart packing list.

Flow

  1. User uploads a messy closet photo and enters trip details like destination and days.
  2. App identifies clothing items in the image and suggests categories and quantities to pack.
  3. User edits the list, checks off items, and exports a printable packing checklist.

Models

  • SAM 3.1 for finding, outlining and counting objects in photos to detect individual garments in the closet image.
  • Muse Spark 1.3 for text and image understanding and structured answers to classify items by type, season, and trip suitability and generate the packing list.

Limits

  • Works best on clearly visible clothing; occluded or tiny items may be missed.
  • No real inventory tracking, just a one-time suggestion from the photo.
  • Cannot verify item condition or personal preferences beyond visual cues.

Cost per run: ~$0.012

Name: Closet Pack List Tagline: Turn your messy closet into a ready-to-pack list. Flow: 1. Upload your messy closet photo. 2. Review and check off detected clothes and shoes. 3. Get a clean packing l…Show allShow less

Name: Closet Pack List
Tagline: Turn your messy closet into a ready-to-pack list.
Flow:

  1. Upload your messy closet photo.
  2. Review and check off detected clothes and shoes.
  3. Get a clean packing list to save or share.
    Models:
  • SAM 3.1 for finding, outlining and counting clothes in the photo
  • Muse Spark 1.3 for text and image understanding and structured answers to name items and build the list
    Limits:
  • Can't see inside bags, drawers, or piles fully hidden behind others
  • May mislabel similar items like black tees or jeans
    Cost per run: $0.008
Base better

Base 9 · Tuned 7

Adds trip details and categorized suggestions matching reference; the other drops destination context and feels thin.

Prompt 3

I run a small bakery and I need an app that counts cracked eggs on the conveyor belt from this morning photo, I tried tallying by hand and kept losing track around 200.

Name: Egg Crack Counter Tagline: Stop hand-tallying, get your cracked count in seconds. Flow: 1. Upload this morning's conveyor belt photo. 2. Mark one cracked and one good egg to guide detection. 3…Show allShow less

Name: Egg Crack Counter
Tagline: Stop hand-tallying, get your cracked count in seconds.

Flow:

  1. Upload this morning's conveyor belt photo.
  2. Mark one cracked and one good egg to guide detection.
  3. Review the outlined count and save the tally.

Models:

  • SAM 3.1 for finding, outlining and counting objects in photos
  • Muse Spark 1.3 for text and image understanding and structured answers

Limits:

  • Overlapping or stacked eggs around 200+ can be undercounted.
  • Blurry photos or glare on wet shells lowers accuracy.
  • Won't distinguish hairline cracks vs dirt without clear close-up.

Cost per run: $0.008

**Egg Crack Counter** Tagline: Count cracked eggs on a conveyor belt photo in seconds. Flow 1. Upload this morning's conveyor belt photo. 2. App finds all eggs and flags cracked vs intact. 3. Shows…Show allShow less

Egg Crack Counter
Tagline: Count cracked eggs on a conveyor belt photo in seconds.

Flow

  1. Upload this morning's conveyor belt photo.
  2. App finds all eggs and flags cracked vs intact.
  3. Shows total cracked count, intact count, and a marked-up image.

Models

  • SAM 3.1 for finding, outlining and counting objects in photos to segment each egg on the belt.
  • Muse Spark 1.3 for text and image understanding and structured answers to classify each segmented egg as cracked or intact and return a structured count.

Limits

  • Accuracy drops with heavy overlap, motion blur, or poor lighting.
  • Cannot count eggs not fully visible in the single photo.
  • No real-time video processing, photo only.

Cost per run: ~$0.008

[no final answer within 512 new tokens]

Base better

Base 9 · Tuned 1

Provides complete brief with correct models and limits; the other gives no answer.

Prompt 4

Photo of my garden bed attached, I want an app that outlines tomato seedlings vs weeds so I do not pull the wrong thing, my neighbor tried to help and we both guessed wrong last spring.

**Name:** Seedling Saver **Tagline:** Keep the tomatoes, pull the weeds. **Flow:** 1. Snap a photo of your garden bed. 2. See tomato seedlings outlined in green vs weeds in red. 3. Tap any outline…Show allShow less

Name: Seedling Saver

Tagline: Keep the tomatoes, pull the weeds.

Flow:

  1. Snap a photo of your garden bed.
  2. See tomato seedlings outlined in green vs weeds in red.
  3. Tap any outline for pull-or-keep advice.

Models:

  • SAM 3.1 for finding, outlining and counting objects in photos
  • Muse Spark 1.3 for text and image understanding and structured answers

Limits:

  • Young seedlings and weeds can look alike, mistakes happen.
  • Poor light, overlap, or blurry photos reduce accuracy.
  • Does not identify disease or nutrient issues.

Cost per run: ~$0.008

**Garden Seedling Spotter** Tagline: Tell tomato seedlings from weeds before you pull. Flow 1. User uploads a photo of the garden bed. 2. App outlines each seedling and labels it Tomato Seedling or…Show allShow less

Garden Seedling Spotter
Tagline: Tell tomato seedlings from weeds before you pull.

Flow

  1. User uploads a photo of the garden bed.
  2. App outlines each seedling and labels it Tomato Seedling or Weed with confidence.
  3. User taps a plant to see why and gets a safe-to-pull recommendation.

Models

  • SAM 3.1 for finding, outlining and counting objects in photos to segment individual seedlings and draw outlines.
  • Muse Spark 1.3 for text and image understanding and structured answers to classify each outlined seedling as tomato vs weed and generate short reasoning.

Limits

  • Works best on clear, close-up daylight photos; dense overlapping seedlings can merge.
  • Early-stage seedlings with similar cotyledons may be ambiguous.
  • No real-time video ID, photo only.

Cost per run: ~$0.008

[no final answer within 512 new tokens]

Base better

Base 9 · Tuned 1

Completed brief follows required shape with correct models and limits; the other produced no answer.

Prompt 5

Need an app to remove the windshield glare from this car damage photo for my insurance claim, the adjuster rejected the first one and I only have my phone pics.

**Name:** Glare-Free Claim **Tagline:** Clean up phone glare to show damage clearly. **Flow:** 1. Upload your car damage phone pic and mark the glare area. 2. Preview the de-glared version and comp…Show allShow less

Name: Glare-Free Claim

Tagline: Clean up phone glare to show damage clearly.

Flow:

  1. Upload your car damage phone pic and mark the glare area.
  2. Preview the de-glared version and compare side-by-side with original.
  3. Download a pack with original + enhanced copy and disclosure note for adjuster.

Models:

  • Muse Spark 1.3 for text and image understanding and structured answers to detect glare vs. damage
  • Muse Image 1.0 for generating or editing images to reduce glare

Limits:

  • Cannot recreate damage hidden under glare, only reduces reflection.
  • Edited photos may still be rejected - insurers often require originals.
  • Works only on phone-quality stills, not heavy blur or video.

Cost per run: ~$0.015

**Name:** GlareErase Claim **Tagline:** Clean windshield photos for insurance approval. **Flow** 1. User uploads phone car damage photo and confirms damage area to keep. 2. App isolates windshield/…Show allShow less

Name: GlareErase Claim

Tagline: Clean windshield photos for insurance approval.

Flow

  1. User uploads phone car damage photo and confirms damage area to keep.
  2. App isolates windshield/glare region and previews glare removal.
  3. User approves edit and exports glare-free photo with claim metadata.

Models

  • SAM 3.1 for finding, outlining and counting objects in photos and video: segment windshield and damage area.
  • Muse Spark 1.3 for text and image understanding and structured answers: interpret user intent and validate damage is preserved.
  • Muse Image 1.0 for generating or editing images: remove glare/reflection and reconstruct underlying damage detail.

Limits

  • Cannot recover detail fully obscured by glare; results are inpainted estimates.
  • Works best on single windshield glare, not multi-pane or moving reflections.
  • Not a forensic tool; insurance may still require original unedited photo.

Cost per run: ~$0.018

Name: Glare-Free Claim Tagline: Clear damage photos adjusters will accept. Flow: 1. Upload your phone photo of the car damage. 2. Tap the windshield area to remove glare. 3. Download the cleaned bef…Show allShow less

Name: Glare-Free Claim
Tagline: Clear damage photos adjusters will accept.

Flow:

  1. Upload your phone photo of the car damage.
  2. Tap the windshield area to remove glare.
  3. Download the cleaned before/after for your claim.

Models:

  • Muse Image 1.0 for generating or editing images to reduce glare and restore damage detail.
  • SAM 3.1 for finding, outlining damage behind windshield.
  • Muse Spark 1.3 for text and image understanding and structured answers to check photo quality.

Limits:

  • Cannot recover damage fully hidden by glare, only reduces it.
  • May alter colors/lighting, not for legal evidence alone.
  • Works best on single windshield glare, not heavy rain or night.

Cost per run: ~$0.025

Tuned better

Base 5 · Tuned 6

Preserves original plus cleaned copy for adjuster while other exports only edited photo and claims reconstructed detail.

Prompt 6

got this photo of moms handwritten chicken recipe and I cant read half of it, need an app that types it neat for me please

**Name:** Mom's Recipe Reader **Tagline:** Turn faded handwriting into clean, printable recipe text. **Flow:** 1. Upload photo of mom's handwritten chicken recipe. 2. Review the neat typed draft an…Show allShow less

Name: Mom's Recipe Reader

Tagline: Turn faded handwriting into clean, printable recipe text.

Flow:

  1. Upload photo of mom's handwritten chicken recipe.
  2. Review the neat typed draft and fix any flagged unclear words.
  3. Save or print the clean recipe card.

Models:

  • SAM 3.1 for finding and outlining the recipe paper in the photo
  • Muse Spark 1.3 for text and image understanding and structured answers to transcribe handwriting

Limits:

  • Very faded or smudged words may be guessed wrong and are marked [unsure].
  • Heavy shadows, wrinkles, or fancy cursive can lower accuracy.

Cost per run: ~$0.004

**Name:** Recipe Handwriting Transcriber **Tagline:** Turn messy handwritten recipes into clean typed cards. **Flow** 1. User uploads photo of handwritten recipe. 2. App isolates text regions and t…Show allShow less

Name: Recipe Handwriting Transcriber

Tagline: Turn messy handwritten recipes into clean typed cards.

Flow

  1. User uploads photo of handwritten recipe.
  2. App isolates text regions and transcribes ingredients and steps.
  3. App outputs neat formatted recipe with title, ingredients, instructions.

Models

  • SAM 3.1 for finding, outlining and counting objects in photos and video to isolate the recipe text area from background clutter.
  • Muse Spark 1.3 for text and image understanding and structured answers to read handwriting, correct errors, and output structured recipe JSON.
  • Muse Image 1.0 for generating or editing images to render a clean printable recipe card layout.

Limits

  • Handwriting must be legible; very faint or overlapping ink fails.
  • No automatic ingredient substitution or nutrition calculation.
  • Works best on single-page photos, not multi-page scans.

Cost per run: ~$0.012

Name: Mom's Recipe Reader Tagline: Turn messy handwritten recipes into neat, readable ones. Flow: 1. Upload photo of the handwritten chicken recipe. 2. Review and tap any words the app flagged as uns…Show allShow less

Name: Mom's Recipe Reader
Tagline: Turn messy handwritten recipes into neat, readable ones.
Flow:

  1. Upload photo of the handwritten chicken recipe.
  2. Review and tap any words the app flagged as unsure.
  3. Get a clean typed version to save or print.
    Models:
  • Muse Spark 1.3 for text and image understanding and structured answers to transcribe handwriting and format ingredients and steps.
    Limits:
  • Struggles with very blurry, faded, or overlapping handwriting.
  • May misread similar-looking ingredients or amounts.
    Cost per run: ~$0.003
Tuned better

Base 6 · Tuned 8

Keeps flow user focused and avoids unneeded image generation; the other adds system steps and misuses image model.

Prompt 7

took a screenshot of my electric bill and its way too high, need an app that reads it and tells me in plain words why

**Name:** Power Bill Explainer **Tagline:** Your confusing electric bill, explained in plain English. **Flow:** 1. Upload your bill screenshot. 2. Get usage, rate hikes, and fees broken down simply.…Show allShow less

Name: Power Bill Explainer
Tagline: Your confusing electric bill, explained in plain English.

Flow:

  1. Upload your bill screenshot.
  2. Get usage, rate hikes, and fees broken down simply.
  3. Ask follow-ups like "why higher than last month?"

Models:

  • Muse Spark 1.3 for text and image understanding and structured answers - reads the screenshot and explains charges.

Limits:

  • Can't verify if meter readings are accurate, only what the bill shows.
  • Blurry screenshots or cropped totals may misread.

Cost per run: $0.004

Name: Bill Whisperer
Tagline: Plain English explanations for why your electric bill spiked.

Flow

  1. User uploads a screenshot of their electric bill.
  2. App extracts usage, rates, fees and prior period comparison from the image.
  3. App returns a plain-language summary of the top drivers of
Name: Bill Plain Talk Tagline: Snap your bill, get a plain-words why it's high. Flow: 1. Upload your electric bill screenshot. 2. Review extracted usage, rates, and fees. 3. Get a plain-words summary…Show allShow less

Name: Bill Plain Talk
Tagline: Snap your bill, get a plain-words why it's high.
Flow:

  1. Upload your electric bill screenshot.
  2. Review extracted usage, rates, and fees.
  3. Get a plain-words summary of what drove the cost.
    Models:
  • Muse Spark 1.3 for text and image understanding and structured answers to read the bill and explain it in plain words.
    Limits:
  • Struggles with blurry, cropped, or handwritten bills.
  • Can't verify utility rates or give official advice.
    Cost per run: $0.004
Tuned better

Base 3 · Tuned 9

The complete brief gives all sections with correct model and cost; the other is cut off and missing models, limits and cost.

Prompt 8

Attached photo of my farmers market stall at peak hour, I need an app that counts overlapping Gala vs Fuji apples without double counting and tells me where it is uncertain.

**Name:** Market Apple Counter **Tagline:** Count Gala vs Fuji in crowded stall photos, no doubles. **Flow:** 1. You upload peak-hour stall photo. 2. App outlines each apple, labels Gala vs Fuji, a…Show allShow less

Name: Market Apple Counter

Tagline: Count Gala vs Fuji in crowded stall photos, no doubles.

Flow:

  1. You upload peak-hour stall photo.
  2. App outlines each apple, labels Gala vs Fuji, and flags uncertain spots.
  3. You get totals per variety plus uncertainty overlay to re-check.

Models:

  • SAM 3.1 for finding, outlining and counting overlapping apples without double counting
  • Muse Spark 1.3 for text and image understanding and structured answers to classify Gala vs Fuji and report uncertainty

Limits:

  • Heavy occlusion or poor lighting lowers variety accuracy
  • Very stacked piles may still undercount hidden apples
  • Single photo only, no tracking across video

Cost per run: ~$0.012

**Apple Stall Counter** Tagline: Count Gala vs Fuji apples in a crowded stall photo without double counting. Flow 1. Upload stall photo and set count mode for Gala vs Fuji. 2. App segments individu…Show allShow less

Apple Stall Counter

Tagline: Count Gala vs Fuji apples in a crowded stall photo without double counting.

Flow

  1. Upload stall photo and set count mode for Gala vs Fuji.
  2. App segments individual apples and classifies variety per instance.
  3. Review counts with uncertainty highlights and export summary.

Models

  • SAM 3.1 for finding, outlining and counting objects in photos and video — instance segmentation of overlapping apples and de-duplication.
  • Muse Spark 1.3 for text and image understanding and structured answers — variety classification Gala vs Fuji per segmented apple and structured count output.

Limits

  • Heavy occlusion and identical lighting can merge boundaries, causing under-count.
  • Gala/Fuji visual distinction is subtle; color shifts from shadows increase misclassification.
  • Works on single still photo; no video tracking or real-time counting.

Cost per run: ~$0.015

[no final answer within 512 new tokens]

Base better

Base 9 · Tuned 1

Delivers complete brief with correct models, flow and limits while the other provides no answer at all.

Prompt 9

Attached photo of mixed recycling tote, I need an app that outlines plastic bags as contaminants and gives a precision focused count for our audit.

**Name:** Tote Bag Auditor **Tagline:** Spot plastic-bag contaminants in recycling totes fast. **Flow:** 1. Upload photo of mixed recycling tote. 2. Review outlined plastic bags and correct misses.…Show allShow less

Name: Tote Bag Auditor
Tagline: Spot plastic-bag contaminants in recycling totes fast.

Flow:

  1. Upload photo of mixed recycling tote.
  2. Review outlined plastic bags and correct misses.
  3. Save precision-focused count for audit log.

Models:

  • SAM 3.1 for finding, outlining and counting plastic bags in photo
  • Muse Spark 1.3 for text and image understanding and structured answers to confirm contaminants and format audit

Limits:

  • Misses buried / heavily overlapped bags
  • Confuses clear film and wrappers with bags
  • Low light or blur reduces outline precision

Cost per run: $0.012

**Name:** Tote Contaminant Counter **Tagline:** Outline plastic bag contaminants and get a precision audit count from a tote photo. **Flow** 1. User uploads photo of mixed recycling tote. 2. App out…Show allShow less

Name: Tote Contaminant Counter
Tagline: Outline plastic bag contaminants and get a precision audit count from a tote photo.

Flow

  1. User uploads photo of mixed recycling tote.
  2. App outlines plastic bags as contaminants and returns a precision focused count with visual overlay.
  3. User reviews count and exports audit report.

Models

  • SAM 3.1 for finding, outlining and counting objects in photos and video — detects and outlines plastic bags in the tote and provides a precise count.
  • Muse Spark 1.3 for text and image understanding and structured answers — interprets the tote context and formats the count and audit summary.

Limits

  • Counting accuracy drops with overlapping, crumpled bags or poor lighting.
  • Only plastic bags are outlined; other contaminants are ignored.
  • Works on single static photo, not continuous video audit.

Cost per run: ~$0.008

[no final answer within 512 new tokens]

Base better

Base 9 · Tuned 1

Complete brief uses correct models with clear audit flow; the other gives no answer.

Prompt 10

Slow motion video of hive entrance yesterday, I need an app that counts bees entering vs leaving at high speed and handles wing blur without inflating totals.

**Name:** Hive Traffic Counter **Tagline:** Split-second in-vs-out bee counts without blur double-counts. **Flow:** 1. Upload your slow-motion hive entrance video from yesterday. 2. Draw the entranc…Show allShow less

Name: Hive Traffic Counter
Tagline: Split-second in-vs-out bee counts without blur double-counts.

Flow:

  1. Upload your slow-motion hive entrance video from yesterday.
  2. Draw the entrance line to define entering vs leaving.
  3. Review in/out totals, net flow, and peak activity timeline.

Models:

  • SAM 3.1 for finding, outlining and counting objects in video, tracking bees across frames despite wing blur.
  • Muse Wonder 1.0 for long multi-step planning of direction logic and de-duplication.

Limits:

  • Heavy pile-ups at entrance can merge and undercount.
  • Requires slow-motion; normal-speed video drops accuracy.
  • Lighting glare or shadows can cause missed detections.

Cost per run: ~$0.032

**HiveFlow Counter** Tagline: Accurate enter/exit bee counts from slow-motion hive entrance video. Flow 1. Upload slow-motion hive entrance video and set entrance zone lines for in/out. 2. App trac…Show allShow less

HiveFlow Counter

Tagline: Accurate enter/exit bee counts from slow-motion hive entrance video.

Flow

  1. Upload slow-motion hive entrance video and set entrance zone lines for in/out.
  2. App tracks each bee across frames and classifies direction at the zone.
  3. Review live counts, net flow, and export CSV with timestamps.

Models

  • SAM 3.1 for finding, outlining and counting objects in photos and video to detect and segment bees per frame and maintain identity across wing blur.
  • Muse Spark 1.3 for text and image understanding and structured answers to classify enter vs leave direction from pose and trajectory.
  • Muse Wonder 1.0 for long multi-step planning to deduplicate detections and prevent double counting across high-speed frames.

Limits
*

[no final answer within 512 new tokens]

Base better

Base 4 · Tuned 1

Provides name, tagline, flow and valid models while the other gives nothing, though it is cut off before limits and cost.

Prompt 11

photo of my monstera leaves with brown spots attached, need an app that tells me in simple words if it is overwatering or bugs

Name: Monstera Spot Checker Tagline: Snap a leaf, get a plain-words overwatering vs. bugs verdict. Flow: 1. Upload your photo of spotted monstera leaves. 2. Answer 2 quick questions on watering and r…Show allShow less

Name: Monstera Spot Checker
Tagline: Snap a leaf, get a plain-words overwatering vs. bugs verdict.
Flow:

  1. Upload your photo of spotted monstera leaves.
  2. Answer 2 quick questions on watering and recent changes.
  3. Get a simple verdict - overwatering, bugs, or unsure - plus next step.
    Models:
  • SAM 3.1 for finding, outlining and counting brown spots in photos
  • Muse Spark 1.3 for text and image understanding and structured answers
    Limits:
  • Can't confirm pests without close-up, bugs can look similar
  • Dark or blurry photos lower accuracy
  • General care advice only, not pesticide dosing
    Cost per run: ~$0.008
Name: Monstera Spot Checker Tagline: Simple diagnosis for brown leaf spots in plain language. Flow: 1. User uploads a photo of monstera leaves with brown spots. 2. App highlights the spots and asks…Show allShow less

Name: Monstera Spot Checker
Tagline: Simple diagnosis for brown leaf spots in plain language.

Flow:

  1. User uploads a photo of monstera leaves with brown spots.
  2. App highlights the spots and asks for quick context like watering frequency.
  3. App returns a plain-language verdict: overwatering vs bugs vs other, with one care tip.

Models:
SAM 3.1 for finding, outlining and counting brown spot objects in the photo.
Muse Spark 1.3 for text and image understanding and structured answers to classify cause as overwatering or bugs and generate simple explanation.

Limits:
Works best on clear, well-lit leaf close-ups; cannot confirm pests without magnification.
Diagnosis is general guidance, not a professional plant pathology diagnosis.
No treatment history tracking across

[no final answer within 512 new tokens]

Base better

Base 7 · Tuned 1

Gives clear name, flow, and correct models with plain verdict; the other provides no output at all.

Prompt 12

short clip of koi fish in our backyard pond about 15 of them darting fast, need an app that counts orange versus white ones

Name: Koi Color Counter Tagline: Count orange vs white koi in your pond clip. Flow: 1. Upload your short backyard pond clip. 2. App tracks each darting fish and labels orange vs white. 3. See totals,…Show allShow less

Name: Koi Color Counter
Tagline: Count orange vs white koi in your pond clip.
Flow:

  1. Upload your short backyard pond clip.
  2. App tracks each darting fish and labels orange vs white.
  3. See totals, plus still frames with outlines for each group.
    Models:
  • SAM 3.1 for finding, outlining and counting objects in photos and video
  • Muse Spark 1.3 for text and image understanding and structured answers
    Limits:
  • Fast darts, splashes, and overlap can cause missed or double counts.
  • Mixed orange-white patterned koi get forced into one group.
    Cost per run: ~$0.02
**Koi Color Counter** Tagline: Count orange vs white koi from a pond video clip. Flow 1. Upload short pond clip and set target colors orange and white. 2. App tracks and outlines each fish across fr…Show allShow less

Koi Color Counter
Tagline: Count orange vs white koi from a pond video clip.

Flow

  1. Upload short pond clip and set target colors orange and white.
  2. App tracks and outlines each fish across frames to avoid double counts.
  3. Review per-fish color breakdown and total counts for orange vs white.

Models

  • SAM 3.1 for finding, outlining and counting individual koi in each frame and tracking them across the fast-moving clip.
  • Muse Spark 1.3 for text and image understanding and structured answers to classify each tracked fish as orange or white and produce the final counts.

Limits

  • Fast motion and overlapping fish can cause missed or merged IDs.
  • Color classification struggles with lighting changes, reflections, and muddy water.
  • Clip length limited to short bursts for reliable tracking.

Cost per run: ~$0.015

Name: Koi Color Counter Tagline: Count orange vs white koi in seconds. Flow: 1. Upload your short pond clip. 2. App tracks each fish and flags orange or white. 3. Get total counts and a simple chart.…Show allShow less

Name: Koi Color Counter
Tagline: Count orange vs white koi in seconds.
Flow:

  1. Upload your short pond clip.
  2. App tracks each fish and flags orange or white.
  3. Get total counts and a simple chart.
    Models:
  • SAM 3.1 for finding, outlining and counting objects in photos and video to track fast-moving koi.
  • Muse Spark 1.3 for text and image understanding and structured answers to classify orange vs white and format results.
    Limits:
  • Fast darting and overlap can cause miscounts.
  • Poor light or murky water lowers color accuracy.
  • Can't reliably tell orange from white on very pale or dark fish.
    Cost per run: $0.018
Tuned better

Base 8 · Tuned 9

Keeps exact Name Tagline Flow labels and is concise; the other adds bold and drops colons.

Prompt 13

Attached photo of my fridge covered in souvenir magnets, need an app that lists each city and writes a one line funny memory for every magnet

Name: Fridge Magnet Tour Tagline: Your fridge souvenirs turned into a funny city list. Flow: 1. Snap or upload your magnet-covered fridge photo. 2. Confirm the auto-detected city list. 3. Get a one-l…Show allShow less

Name: Fridge Magnet Tour
Tagline: Your fridge souvenirs turned into a funny city list.
Flow:

  1. Snap or upload your magnet-covered fridge photo.
  2. Confirm the auto-detected city list.
  3. Get a one-line funny memory for each magnet to save and share.
    Models:
  • SAM 3.1 for finding, outlining and counting each magnet.
  • Muse Spark 1.3 for reading city names and writing structured funny lines.
    Limits:
  • Blurry or overlapping magnets may be missed or misread.
  • Memories are invented jokes, not your real trips.
    Cost per run: $0.006
**Name:** Magnet Memory Maker **Tagline:** Turn fridge magnets into funny travel stories. **Flow** 1. Upload fridge photo and confirm magnet crop. 2. App identifies each magnet and reads the city n…Show allShow less

Name: Magnet Memory Maker

Tagline: Turn fridge magnets into funny travel stories.

Flow

  1. Upload fridge photo and confirm magnet crop.
  2. App identifies each magnet and reads the city name.
  3. App outputs a list with city + one-line funny memory per magnet.

Models

  • SAM 3.1 for finding, outlining and counting each magnet in the photo.
  • Muse Spark 1.3 for text and image understanding to read city names on magnets and generate a one-line funny memory per city in structured answers.

Limits

  • City names must be legible; small or reflective text may be missed.
  • Memories are invented humor, not real recollections.
  • Works best with ~5-30 visible magnets; dense clutter reduces accuracy.

Cost per run: ~$0.015

[no final answer within 512 new tokens]

Base better

Base 9 · Tuned 1

Delivers complete brief with correct models and cost; the other gives no answer.

Prompt 14

Screenshot of my sourdough recipe from a blog, need an app that rewrites it as 5 short steps in a cheerful tone for beginners

Name: Cheery Sourdough Steps Tagline: Blog screenshots turned into 5 happy beginner steps. Flow: 1. Upload your screenshot of the sourdough blog recipe. 2. Tap Rewrite for beginners. 3. Read and sav…Show allShow less

Name: Cheery Sourdough Steps
Tagline: Blog screenshots turned into 5 happy beginner steps.

Flow:

  1. Upload your screenshot of the sourdough blog recipe.
  2. Tap Rewrite for beginners.
  3. Read and save the 5 short cheerful steps.

Models:

  • Muse Spark 1.3 for text and image understanding and structured answers to read the screenshot and rewrite it.

Limits:

  • Needs a clear, readable screenshot; blurry text may be missed.
  • Simplifies for beginners, may drop precise baker's details.

Cost per run: $0.003

**Name:** Sourdough Simplifier **Tagline:** Turn any blog recipe into 5 cheerful beginner steps. **Flow** 1. User uploads a screenshot of the sourdough blog recipe. 2. App extracts the ingredients…Show allShow less

Name: Sourdough Simplifier

Tagline: Turn any blog recipe into 5 cheerful beginner steps.

Flow

  1. User uploads a screenshot of the sourdough blog recipe.
  2. App extracts the ingredients and method from the image and condenses it.
  3. User receives 5 short, cheerful steps for beginners with optional copy/share.

Models

  • Muse Spark 1.3 for text and image understanding and structured answers — reads the screenshot, extracts recipe content, and rewrites it as 5 short beginner-friendly steps in a cheerful tone.
  • Muse Image 1.0 for generating or editing images — optional clean visual card of the 5 steps.

Limits

  • Accuracy depends on screenshot clarity and text legibility; handwritten notes may be missed.
  • Cheerful tone is generated, not a 1:1 translation of the original author’s voice.
  • No baking safety validation; times/temperatures are preserved as
Name: Sourdough Buddy Tagline: Your blog recipe, now 5 cheerful steps for beginners. Flow: 1. Upload your sourdough recipe screenshot. 2. Confirm ingredients and time. 3. Get 5 short, cheerful steps…Show allShow less

Name: Sourdough Buddy
Tagline: Your blog recipe, now 5 cheerful steps for beginners.
Flow:

  1. Upload your sourdough recipe screenshot.
  2. Confirm ingredients and time.
  3. Get 5 short, cheerful steps to follow.
    Models:
  • Muse Spark 1.3 for text and image understanding and structured answers to read the screenshot and rewrite as 5 steps.
    Limits:
  • May misread blurry screenshots or handwriting.
  • Won't check food safety or oven temps.
    Cost per run: $0.003
Tuned better

Base 5 · Tuned 9

Complete brief uses only needed text model while the other adds unneeded image model and cuts off mid sentence.

Prompt 15

Attached voice memo from my landlord about rent increase to 1450, need an app that transcribes it exactly and summarizes dates in formal email format

Name: Rent Rise Reader Tagline: Turn landlord voice memos into exact text plus formal summary. Flow: 1. Upload your landlord voice memo about the $1450 increase. 2. Review exact transcript with key d…Show allShow less

Name: Rent Rise Reader
Tagline: Turn landlord voice memos into exact text plus formal summary.
Flow:

  1. Upload your landlord voice memo about the $1450 increase.
  2. Review exact transcript with key dates extracted.
  3. Copy formal email-format summary for your records.
    Models:
  • Muse Voice Transcribe 1.0 for turning speech into text
  • Muse Spark 1.3 for text and image understanding and structured answers
    Limits:
  • Accuracy drops with background noise or overlapping voices.
  • Will not verify legal validity of the rent increase.
    Cost per run: $0.004
Name: Rent Memo Emailer Tagline: Transcribe landlord voice memo and draft a formal rent increase email. Flow: 1. User uploads voice memo and confirms recipient email. 2. App transcribes audio verbat…Show allShow less

Name: Rent Memo Emailer
Tagline: Transcribe landlord voice memo and draft a formal rent increase email.

Flow:

  1. User uploads voice memo and confirms recipient email.
  2. App transcribes audio verbatim and extracts rent amount, dates, and terms.
  3. App formats a formal email summary with dates and offers edit/send.

Models:

  • Muse Voice Transcribe 1.0 for turning speech into text from the voice memo
  • Muse Spark 1.3 for text and image understanding and structured answers to extract rent details and generate formal email format

Limits:

  • Transcription accuracy drops with background noise or heavy accents
  • No verification that extracted dates are correct; user must review
  • Email is draft only, no direct sending without user approval

Cost per run: ~$0.008

[no final answer within 512 new tokens]

Base better

Base 9 · Tuned 1

Delivers full brief with correct models, flow and cost while the incomplete response offers no usable output.

Prompt 16

sketch of my ant farm tunnels on notebook paper attached, want an app that turns it into a neat diagram with labels for my kids science fair

**Name:** Ant Tunnel Mapper **Tagline:** From notebook scribble to science-fair-ready ant diagram. **Flow:** 1. Upload photo of your notebook ant farm sketch. 2. Check auto-traced tunnels and pick la…Show allShow less

Name: Ant Tunnel Mapper
Tagline: From notebook scribble to science-fair-ready ant diagram.
Flow:

  1. Upload photo of your notebook ant farm sketch.
  2. Check auto-traced tunnels and pick labels like entrance, nursery, food store.
  3. Download neat labeled diagram for the poster.
    Models:
  • SAM 3.1 for finding and outlining tunnels in the photo
  • Muse Spark 1.3 for image understanding and structured labels
  • Muse Image 1.0 for generating clean diagram
    Limits:
  • Faint pencil lines and lined paper may trace incorrectly.
  • Can't infer real tunnel depth or scale from 2D sketch.
  • Labels use kid-friendly terms, not scientific accuracy.
    Cost per run: $0.012

Name: Ant Tunnel Mapper
Tagline: Turn hand-drawn ant farms into clean labeled diagrams for science fair.

Flow

  1. User uploads a photo of the notebook sketch.
  2. App traces the tunnels and proposes kid-friendly

[no final answer within 512 new tokens]

Base better

Base 3 · Tuned 1

Partial brief gives correct name and starts flow, while the other provides nothing.

Prompt 17

photo of old headstones in my family plot attached, need an app that reads the names and dates into a clean family table

Name: Plot Stone Reader Tagline: Turn weathered headstones into a tidy family table. Flow: 1. Upload photo of your family plot. 2. Review flagged names and dates and fix any guesses. 3. Save and exp…Show allShow less

Name: Plot Stone Reader
Tagline: Turn weathered headstones into a tidy family table.

Flow:

  1. Upload photo of your family plot.
  2. Review flagged names and dates and fix any guesses.
  3. Save and export a clean family table as CSV.

Models:

  • SAM 3.1 for finding and outlining each headstone in the photo.
  • Muse Spark 1.3 for text and image understanding and structured answers to read names/dates into a table.

Limits:

  • Worn, mossy, or tilted text may be misread and is flagged unsure.
  • Does not verify family relationships beyond what is carved.
  • Works best with one plot per photo in daylight.

Cost per run: $0.007

**Name:** Headstone Table Builder **Tagline:** Turn cemetery photos into a clean family table. **Flow** 1. Upload photo of old headstones and confirm family plot area. 2. Review extracted names, da…Show allShow less

Name: Headstone Table Builder

Tagline: Turn cemetery photos into a clean family table.

Flow

  1. Upload photo of old headstones and confirm family plot area.
  2. Review extracted names, dates and relationships, correct any errors.
  3. Export clean family table as CSV or printable chart.

Models

  • SAM 3.1 for finding, outlining and counting objects in photos and video — isolate each headstone in the image.
  • Muse Spark 1.3 for text and image understanding and structured answers — read names and dates from each headstone and structure into person records.
  • Muse Wonder 1.0 for long multi-step planning — link individuals into family groups and build table hierarchy.

Limits

  • Hand-carved or heavily weathered inscriptions may be misread.
  • Photo angle, glare or overlapping stones reduce accuracy.
  • No genealogical verification beyond what is visible on stone.

Cost per run: ~$0.015

Name: Plot Stone Reader Tagline: Turn worn headstone photos into a clean family table. Flow: 1. Upload your family plot photo. 2. Review and fix the auto-read names and dates. 3. Export the clean fam…Show allShow less

Name: Plot Stone Reader
Tagline: Turn worn headstone photos into a clean family table.
Flow:

  1. Upload your family plot photo.
  2. Review and fix the auto-read names and dates.
  3. Export the clean family table to save or share.
    Models:
  • SAM 3.1 for finding, outlining and counting headstones in photos
  • Muse Spark 1.3 for text and image understanding and structured answers to read names/dates into a table
    Limits:
  • Faint, cracked, or moss-covered inscriptions may be misread.
  • Cannot verify relationships, only what is carved.
    Cost per run: $0.012
Tuned better

Base 7 · Tuned 9

Uses only needed vision and text models for table; the other adds unneeded planning model and implies relationship inference.

Prompt 18

I'm running late and stressed, need an app that reads this screenshot of my overdue electric bill and gives me the amount and due date in one short alert

Name: Bill Alert Snap Tagline: Snap your bill, get what you owe fast. Flow: 1. Upload the screenshot of your overdue electric bill. 2. Tap Check Now. 3. Get one short alert with amount and due date.…Show allShow less

Name: Bill Alert Snap
Tagline: Snap your bill, get what you owe fast.
Flow:

  1. Upload the screenshot of your overdue electric bill.
  2. Tap Check Now.
  3. Get one short alert with amount and due date.
    Models:
  • Muse Spark 1.3 for text and image understanding and structured answers
    Limits:
  • Blurry or cropped screenshots may misread amount or date.
  • Does not pay the bill or check balance with the utility.
    Cost per run: $0.003
**Name:** Bill Alert Reader **Tagline:** Instant overdue bill summary from a screenshot. **Flow** 1. User uploads a screenshot of the electric bill. 2. App extracts the amount due and due date from…Show allShow less

Name: Bill Alert Reader
Tagline: Instant overdue bill summary from a screenshot.

Flow

  1. User uploads a screenshot of the electric bill.
  2. App extracts the amount due and due date from the image.
  3. App returns one short alert with amount and due date.

Models
Muse Spark 1.3 for text and image understanding and structured answers — reads the bill screenshot and extracts amount due and due date into a structured alert.

Limits
Works best on clear, legible screenshots; handwritten or heavily cropped bills may fail. No payment or account linking, just read-only extraction. Accuracy depends on layout and image quality.

Cost per run: ~$0.003

[no final answer within 512 new tokens]

Base better

Base 9 · Tuned 1

Complete brief uses correct model and clear three step flow while missing response offers no help.

Prompt 19

I am overwhelmed, PDF of my car insurance claim forms attached, need an app that pulls out what to sign and where to send it

**Name:** Claim Sign Helper **Tagline:** Know what to sign and where to send in minutes. **Flow:** 1. Upload your car insurance claim PDF. 2. Review the checklist of signature pages and send-to add…Show allShow less

Name: Claim Sign Helper

Tagline: Know what to sign and where to send in minutes.

Flow:

  1. Upload your car insurance claim PDF.
  2. Review the checklist of signature pages and send-to address.
  3. Check off each signature and submit.

Models:

  • Muse Spark 1.3 for text and image understanding and structured answers to find sign fields, dates, and mailing / upload instructions.

Limits:

  • Can't verify if forms are legally complete or current.
  • Struggles with handwritten notes and poor scans.

Cost per run: ~$0.008

**Name:** Claim Sign Helper **Tagline:** Extract what to sign and where to send from your insurance PDF. **Flow** 1. Upload your car insurance claim PDF or photo. 2. App highlights signature fields…Show allShow less

Name: Claim Sign Helper

Tagline: Extract what to sign and where to send from your insurance PDF.

Flow

  1. Upload your car insurance claim PDF or photo.
  2. App highlights signature fields, required info, and mailing/submission instructions.
  3. You get a checklist with page numbers and a copy-ready address/email.

Models

  • Muse Spark 1.3 for text and image understanding and structured answers: parse PDF pages, extract signature blocks, required fields, and where to send.
  • Muse Wonder 1.0 for long multi-step planning: plan extraction across multi-page forms and consolidate results.

Limits

  • Accuracy drops on low-resolution

[no final answer within 512 new tokens]

Base better

Base 4 · Tuned 1

Partial brief extracts signatures and send address; the other gives no answer at all.

Prompt 20

Need an app right now that plans my job interview commute across town with bus options and what time to leave to not be late

**Name:** Interview Commute Planner **Tagline:** Get there early, calm, and interview-ready. **Flow:** 1. Enter interview address, arrival time, and starting location. 2. Compare bus options with to…Show allShow less

Name: Interview Commute Planner
Tagline: Get there early, calm, and interview-ready.

Flow:

  1. Enter interview address, arrival time, and starting location.
  2. Compare bus options with total travel time and leave-by times.
  3. Pick a plan and get a leave-now reminder with buffer for delays.

Models:

  • Muse Spark 1.3 for text understanding and structured commute answers
  • Muse Wonder 1.0 for long multi-step planning of routes and timing

Limits:

  • No live bus tracking or delay alerts
  • Times are estimates, may vary with traffic
  • Doesn't handle multi-city or train/subway transfers

Cost per run: $0.008

**Name:** Interview Commute Planner **Tagline:** Leave on time, arrive calm for your interview. **Flow** 1. User enters interview address, time, and start location, optionally uploads a photo of th…Show allShow less

Name: Interview Commute Planner

Tagline: Leave on time, arrive calm for your interview.

Flow

  1. User enters interview address, time, and start location, optionally uploads a photo of the invite.
  2. App finds bus routes, travel time, and buffer for transfers/walking.
  3. App returns best departure time with step-by-step bus plan and a late-risk reminder.

Models

  • Muse Wonder 1.0 for long multi-step planning of route options, transfer timing, and optimal leave time calculation.
  • Muse Spark 1.3 for text and image understanding and structured answers to parse interview details from text or photo invite.

Limits

  • Relies on static transit data; real
Name: Interview On-Time Planner Tagline: Get there calm, not late. Flow: 1. Enter interview address, time, and start location. 2. Choose bus-only options and buffer for walking. 3. Get leave time + s…Show allShow less

Name: Interview On-Time Planner
Tagline: Get there calm, not late.
Flow:

  1. Enter interview address, time, and start location.
  2. Choose bus-only options and buffer for walking.
  3. Get leave time + step-by-step bus plan.
    Models:
  • Muse Wonder 1.0 for long multi-step planning of leave time and transfers
  • Muse Spark 1.3 for text understanding and structured answers
    Limits:
  • Uses scheduled times, not live bus tracking or traffic delays
  • May miss last-minute service changes or closures
    Cost per run: ~$0.008
Tuned better

Base 4 · Tuned 9

Complete brief follows exact shape with valid models and cost; the other is cut off and misses cost.

Prompt 21

short video of birds at my feeder this morning attached need app that counts how many landed

**Name:** Feeder Bird Counter **Tagline:** Count every landing from your morning video. **Flow:** 1. Upload your short video of the feeder. 2. Confirm the feeder zone to watch for landings. 3. Get t…Show allShow less

Name: Feeder Bird Counter
Tagline: Count every landing from your morning video.

Flow:

  1. Upload your short video of the feeder.
  2. Confirm the feeder zone to watch for landings.
  3. Get total landed count with timestamps.

Models:

  • SAM 3.1 for finding, outlining and counting birds in video
  • Muse Spark 1.3 for text and image understanding and structured answers with totals

Limits:

  • May double-count the same bird if it leaves and returns.
  • Small, fast, or overlapping birds in blurry clips can be missed.

Cost per run: ~$0.015

**Name:** Feeder Bird Counter **Tagline:** Count every bird that lands on your feeder from a short video. **Flow** 1. Upload your feeder video and set start/end time or let the app auto-detect acti…Show allShow less

Name: Feeder Bird Counter

Tagline: Count every bird that lands on your feeder from a short video.

Flow

  1. Upload your feeder video and set start/end time or let the app auto-detect activity.
  2. Review the per-frame detections with bird outlines and a running tally of unique landings.
  3. Export a summary with total landings, species guesses, and a timestamped log.

Models

  • SAM 3.1 for finding, outlining and counting objects in photos and video — track and outline birds frame-by-frame and identify distinct landings.
  • Muse Spark 1.3 for text and image understanding and structured answers — classify species from crops and produce a structured count summary with timestamps.
  • Muse Wonder 1.0 for long multi-step planning — plan the multi-frame tracking and deduplication logic across the video.

Limits

  • Accuracy drops with motion blur, overlapping birds, or poor lighting.
  • Species ID is best-effort from visual crops; not a definitive ornithology label.
  • Counting unique landings vs. total appearances requires heuristic deduplication.

Cost per run: ~$0.015

Name: Feeder Bird Counter Tagline: Count every visitor at your feeder. Flow: 1. Upload your short morning video. 2. Review auto-outlined birds frame-by-frame. 3. Get total landings with timestamp lis…Show allShow less

Name: Feeder Bird Counter
Tagline: Count every visitor at your feeder.
Flow:

  1. Upload your short morning video.
  2. Review auto-outlined birds frame-by-frame.
  3. Get total landings with timestamp list.
    Models:
  • SAM 3.1 for finding, outlining and counting birds in video
  • Muse Spark 1.3 for text and image understanding and structured answers
    Limits:
  • Misses fast fly-bys or birds hidden behind feeder.
  • Miscounts when birds overlap or leave frame.
    Cost per run: $0.012
Tuned better

Base 6 · Tuned 9

Keeps brief concise with only relevant models while the other adds unnecessary planning and exceeds length.

Prompt 22

photo of coffee shop receipts pile attached need app that adds up the total for march

**Name:** March Receipt Total **Tagline:** Snap the pile, get your March coffee total. **Flow:** 1. Upload the photo of the coffee shop receipts pile. 2. Confirm the March receipts auto-detected for…Show allShow less

Name: March Receipt Total
Tagline: Snap the pile, get your March coffee total.

Flow:

  1. Upload the photo of the coffee shop receipts pile.
  2. Confirm the March receipts auto-detected for review.
  3. Get March total with per-receipt list.

Models:

  • SAM 3.1 for finding, outlining and counting objects in photos and video - to separate each receipt.
  • Muse Spark 1.3 for text and image understanding and structured answers - to read dates and amounts and sum March only.

Limits:

  • Misses fully hidden, blurry, or crumpled receipts.
  • Can misread faded totals or ambiguous dates.
  • Doesn't handle non-March or non-coffee receipts automatically.

Cost per run: $0.008

**Name:** Receipt March Tally **Tagline:** Snap a pile, get your March coffee total instantly. **Flow** 1. User uploads the photo of the receipt pile and confirms date filter = March. 2. App isolat…Show allShow less

Name: Receipt March Tally

Tagline: Snap a pile, get your March coffee total instantly.

Flow

  1. User uploads the photo of the receipt pile and confirms date filter = March.
  2. App isolates each receipt and extracts date, merchant, and total amount.
  3. App sums all March coffee shop totals and shows breakdown by receipt with grand total.

Models

  • SAM 3.1 for finding, outlining and counting objects in photos and video — segment individual receipts in the pile.
  • Muse Spark 1.3 for text and image understanding and structured answers — read receipt text, parse dates and amounts, and return structured JSON.
  • Muse Wonder 1.0 for long multi-step planning — orchestrate segmentation → OCR → filtering by March → summation and validation.

Limits

  • Handwritten or faded thermal prints may be misread.
  • Overlapping receipts can be missed or duplicated.
  • Only totals visible in image are summed; no external account linking.

Cost per run: ~$0.012

[no final answer within 512 new tokens]

Base better

Base 9 · Tuned 1

Delivers complete brief in required shape with valid models; the other provides nothing.

Prompt 23

Photo of my water stained ceiling after last nights rain attached, I've patched it twice and it keeps coming back, need an app that outlines the damp area and tells me if I should call a roofer this week

Name: Ceiling Stain Watch Tagline: Track that leak spot and know when to call a roofer. Flow: 1. Upload photo of your water-stained ceiling after rain. 2. See the damp area outlined with size and dar…Show allShow less

Name: Ceiling Stain Watch
Tagline: Track that leak spot and know when to call a roofer.
Flow:

  1. Upload photo of your water-stained ceiling after rain.
  2. See the damp area outlined with size and darkness change.
  3. Get a clear call: monitor it or call a roofer this week.
    Models:
  • SAM 3.1 for finding, outlining and counting damp areas in photos
  • Muse Spark 1.3 for text and image understanding and structured answers on urgency
    Limits:
  • Only sees surface stain, can't check roof or hidden leak.
  • Lighting and camera angle can change measurements.
  • Not a substitute for a professional inspection.
    Cost per run: $0.012
**Name:** Ceiling Leak Scout **Tagline:** Outline damp stains and get a clear call-roofer recommendation. **Flow** 1. User uploads ceiling photo and confirms rain exposure and patch history. 2. App…Show allShow less

Name: Ceiling Leak Scout

Tagline: Outline damp stains and get a clear call-roofer recommendation.

Flow

  1. User uploads ceiling photo and confirms rain exposure and patch history.
  2. App outlines the damp area and estimates size/severity from the image.
  3. App returns a risk verdict and a simple action plan with timing.

Models

  • SAM 3.1 for finding, outlining and counting objects in photos and video to segment the water stain and measure its area.
  • Muse Spark 1.3 for text and image understanding and structured answers to interpret stain pattern, patch history, and output a roofer recommendation.

Limits

  • Cannot confirm active leak without thermal/moisture sensor data.
  • Assessment is visual only; cannot see roof source from ceiling photo.
  • No repair instructions beyond general guidance.

Cost per run: ~$0.008

[no final answer within 512 new tokens]

Base better

Base 9 · Tuned 1

Finished brief outlines damp area and gives roofer verdict with valid models; the other produced no answer.

Prompt 24

Screenshot of my messy email inbox with 214 unread attached, I tried folders last month and gave up, need an app that summarizes the 3 urgent ones into one line each so I can clear them tonight

Name: Inbox Urgent Triage Tagline: Clear tonight's must-dos in seconds. Flow: 1. Upload your messy inbox screenshot with 214 unread. 2. Pick the candidates for urgent review. 3. Get 3 one-line summar…Show allShow less

Name: Inbox Urgent Triage
Tagline: Clear tonight's must-dos in seconds.
Flow:

  1. Upload your messy inbox screenshot with 214 unread.
  2. Pick the candidates for urgent review.
  3. Get 3 one-line summaries to clear tonight.
    Models: Muse Spark 1.3 for image understanding of the screenshot and structured one-line text summaries.
    Limits: Only sees visible emails, not all 214; urgency detection can be wrong; does not archive or reply for you.
    Cost per run: $0.003

Name: Inbox Triage Tonight
Tagline: One-line summaries for your 3 most urgent unread emails.

Flow

  1. Upload your inbox screenshot with 214 unread
  2. App extracts visible emails and ranks urgency from sender, subject and snippet

[no final answer within 512 new tokens]

Base better

Base 5 · Tuned 1

Gives proper name, tagline and starts flow from screenshot while the other gives no answer.

Prompt 25

Voice memo of the plumber explaining the water heater noise attached, his quote was $400 and I want a second opinion, need an app that transcribes it into what part failed and 2 questions to ask another plumber

**Name:** Plumber Second Opinion **Tagline:** Turn the plumber's ramble into a clear second-opinion cheat sheet. **Flow:** 1. Upload the water heater voice memo and enter the $400 quote. 2. Review th…Show allShow less

Name: Plumber Second Opinion
Tagline: Turn the plumber's ramble into a clear second-opinion cheat sheet.
Flow:

  1. Upload the water heater voice memo and enter the $400 quote.
  2. Review the transcript and confirm noise details.
  3. Get failed part summary plus 2 questions to ask the next plumber.
    Models:
  • Muse Voice Transcribe 1.0 for turning speech into text
  • Muse Spark 1.3 for text and image understanding and structured answers to extract the failed part and draft the questions
    Limits:
  • Noisy garage audio or plumbing jargon may mistranscribe.
  • Can't inspect the heater or verify if $400 is fair.
  • Not licensed plumbing advice, informational only.
    Cost per run: ~$0.008
Name: Plumber Quote Checker Tagline: Transcribe a plumber voice memo and get a second-opinion checklist. Flow: 1. Upload the voice memo of the plumber explaining the water heater noise. 2. Review th…Show allShow less

Name: Plumber Quote Checker
Tagline: Transcribe a plumber voice memo and get a second-opinion checklist.

Flow:

  1. Upload the voice memo of the plumber explaining the water heater noise.
  2. Review the transcription with the identified failed part and $400 quote context.
  3. Copy the 2 tailored questions to ask another plumber for verification.

Models:

  • Muse Voice Transcribe 1.0 for turning speech into text from the voice memo
  • Muse Spark 1.3 for text and image understanding and structured answers to extract the failed part and generate 2 second-opinion questions

Limits:

  • Transcription accuracy drops with background noise or heavy accents
  • Part identification is inferential from description, not a diagnostic inspection
  • No pricing validation or local code checks

Cost per run: $0.012

[no final answer within 512 new tokens]

Base better

Base 9 · Tuned 1

Finished brief hits all sections with correct models and cost while the other produced no answer.

Prompt 26

napkin sketch of the fence i want in my backyard attached its super messy need an app that makes it look nice to show my husband

**Name:** Fancy Fence Preview **Tagline:** Turn your napkin scribble into a clean fence picture. **Flow:** 1. Upload your fence sketch + add height, wood type, and color. 2. Pick a style: picket, p…Show allShow less

Name: Fancy Fence Preview

Tagline: Turn your napkin scribble into a clean fence picture.

Flow:

  1. Upload your fence sketch + add height, wood type, and color.
  2. Pick a style: picket, privacy, or ranch.
  3. Get a neat backyard mockup to share with your husband.

Models:

  • Muse Spark 1.3: to understand your messy sketch and style choices
  • Muse Image 1.0: to generate a clean fence visualization

Limits:

  • Won't keep exact backyard measurements or proportions
  • May smooth out unique details from your original sketch

Cost per run: ~$0.02

**Fence Sketch Polisher** Tagline: Turn messy napkin fence sketches into clean, shareable renders. Flow 1. Upload your napkin sketch photo and add notes like style, height, material. 2. App extract…Show allShow less

Fence Sketch Polisher

Tagline: Turn messy napkin fence sketches into clean, shareable renders.

Flow

  1. Upload your napkin sketch photo and add notes like style, height, material.
  2. App extracts the fence layout and converts it to a clean line drawing with labels.
  3. Review and share a polished backyard visualization with your husband.

Models

  • Muse Spark 1.3 for text and image understanding and structured answers to interpret sketch notes and fence intent.
  • SAM 3.1 for finding, outlining and counting objects in photos and video to isolate the fence drawing from messy napkin background.
  • Muse Image 1.0 for generating or editing images to render a clean, realistic fence visualization.

Limits

  • Very faint or overlapping scribbles may lose detail.
  • No real-world measurements unless you provide them.
  • Style is illustrative, not a contractor-grade plan.

Cost per run: ~$0.018

[no final answer within 512 new tokens]

Base better

Base 9 · Tuned 1

Polished brief follows required format with valid models and clear steps; the other provides no answer.

Prompt 27

photo of a pile of birthday gift receipts attached need an app that adds them up so i know what i spent in june

**Name:** June Gift Tally **Tagline:** Snap the pile, know your June spend. **Flow:** 1. Upload the photo of your birthday gift receipts. 2. Review the found receipts and fix any missed amounts. 3.…Show allShow less

Name: June Gift Tally

Tagline: Snap the pile, know your June spend.

Flow:

  1. Upload the photo of your birthday gift receipts.
  2. Review the found receipts and fix any missed amounts.
  3. Get your June total with a per-gift breakdown.

Models:

  • SAM 3.1 for finding, outlining and counting objects in photos
  • Muse Spark 1.3 for text and image understanding and structured answers

Limits:

  • Blurry, folded, or overlapping receipts may be missed or misread.
  • Only totals clearly dated in June are included; no currency conversion.

Cost per run: ~$0.008

**Name:** Receipt June Tally **Tagline:** Add up a pile of birthday receipts from one photo. **Flow** 1. User uploads photo of the receipt pile and confirms June filter. 2. App isolates each receipt…Show allShow less

Name: Receipt June Tally
Tagline: Add up a pile of birthday receipts from one photo.

Flow

  1. User uploads photo of the receipt pile and confirms June filter.
  2. App isolates each receipt and reads date, merchant and total.
  3. App sums only June receipts and shows total spent with itemized list.

Models

  • SAM 3.1 for finding, outlining and counting objects in photos and video to segment individual receipts from the pile and overlapping edges.
  • Muse Spark 1.3 for text and image understanding and structured answers to OCR receipt text and extract date, merchant and amount into structured fields.

**

[no final answer within 512 new tokens]

Base better

Base 4 · Tuned 1

Gives partial brief with name, flow and correct models while the other gives nothing.

Prompt 28

Screenshot of our staging error logs from last night attached, need an app that parses the stack traces and flags retryable failures even when lines are truncated

Name: Retry Log Scanner Tagline: Find retryable failures in messy staging logs fast. Flow: 1. Upload last night's error log screenshot or paste truncated stack traces. 2. Review parsed errors flagged…Show allShow less

Name: Retry Log Scanner
Tagline: Find retryable failures in messy staging logs fast.
Flow:

  1. Upload last night's error log screenshot or paste truncated stack traces.
  2. Review parsed errors flagged as retryable vs fatal with reasons.
  3. Export retry queue to Slack / incident ticket.
    Models:
  • Muse Spark 1.3 for text and image understanding and structured answers: reads screenshot text, reconstructs truncated lines, classifies retryability.
    Limits:
  • Misses context if lines are heavily cropped or blurry.
  • Flags are heuristic, can't confirm live infra root cause.
    Cost per run: $0.004
**Name:** Log Trace Classifier **Tagline:** Parse truncated stack traces from screenshots and flag retryable failures. **Flow** 1. User uploads staging error log screenshot. 2. App extracts stack t…Show allShow less

Name: Log Trace Classifier

Tagline: Parse truncated stack traces from screenshots and flag retryable failures.

Flow

  1. User uploads staging error log screenshot.
  2. App extracts stack traces and classifies each as retryable or non-retryable with reason.
  3. User reviews flagged failures and exports a summary CSV.

Models

  • Muse Spark 1.3 for text and image understanding and structured answers — OCR the screenshot, reconstruct truncated lines, and output structured trace objects with retryability label

[no final answer within 512 new tokens]

Base better

Base 5 · Tuned 1

Provides name, tagline, flow and correct model use; the other gives no answer at all.

Prompt 29

PDF of my marathon training plan mixed with physio notes attached, need an app that merges them into 3 weekly priorities and flags conflicting workouts

**Name:** Race Ready Reconciler **Tagline:** Merge training and physio into 3 safe weekly wins. **Flow:** 1. Upload your marathon plan + physio notes PDF. 2. Confirm extracted mileage, injuries, and…Show allShow less

Name: Race Ready Reconciler
Tagline: Merge training and physio into 3 safe weekly wins.
Flow:

  1. Upload your marathon plan + physio notes PDF.
  2. Confirm extracted mileage, injuries, and restricted moves.
  3. Get 3 weekly priorities with conflicting workouts flagged.
    Models:
  • Muse Spark 1.3 for text and image understanding and structured answers
  • Muse Wonder 1.0 for long multi-step planning
    Limits:
  • Not medical advice, check physio for pain decisions.
  • May misread scanned tables or handwritten notes.
  • No live GPS or recovery tracking.
    Cost per run: ~$0.018
**Name:** Train Merge Coach **Tagline:** Merge plan and physio into clear weekly priorities. **Flow** 1. Upload PDF, app extracts training schedule and physio notes. 2. App reconciles workouts with…Show allShow less

Name: Train Merge Coach

Tagline: Merge plan and physio into clear weekly priorities.

Flow

  1. Upload PDF, app extracts training schedule and physio notes.
  2. App reconciles workouts with physio restrictions and picks 3 weekly priorities.
  3. Review flagged conflicts and approved plan, export weekly summary.

Models

  • Muse Spark 1.3 for text and image understanding and structured answers — parse PDF text/tables and extract workouts and physio notes.
  • Muse Wonder 1.0 for long multi-step planning — merge two sources, detect conflicts, rank priorities, and generate weekly plan.

Limits

  • PDF must be readable text/images; scanned handwriting may be missed.
  • No medical diagnosis, only flags potential conflicts for user review.
  • Conflicts are rule-based, not personalized physiotherapy advice.

Cost per run: ~$0.018

[no final answer within 512 new tokens]

Base better

Base 9 · Tuned 1

Delivers complete brief with correct models and cost while the other provides no answer.

Prompt 30

Voice memo of a bilingual support call attached, speaker switches to Tagalog mid call, need an app that transcribes both parts without dropping the switch

Name: Bilingual Call Capture Tagline: Catch every switch between English and Tagalog. Flow: 1. Upload your bilingual support call voice memo. 2. Get a timestamped transcript with language labels. 3.…Show allShow less

Name: Bilingual Call Capture
Tagline: Catch every switch between English and Tagalog.
Flow:

  1. Upload your bilingual support call voice memo.
  2. Get a timestamped transcript with language labels.
  3. Copy or export the cleaned notes.
    Models:
  • Muse Voice Transcribe 1.0 for turning speech into text
  • Muse Spark 1.3 for text understanding and structured answers with language tags
    Limits:
  • Struggles with overlapping speech and loud background noise.
  • Rare Tagalog dialects or slang may mistranscribe.
    Cost per run: $0.012
**Name:** Bilingual Call Transcribe **Tagline:** Accurate transcript that follows language switches mid-call. **Flow** 1. User uploads voice memo and selects source/target languages. 2. App transcr…Show allShow less

Name: Bilingual Call Transcribe

Tagline: Accurate transcript that follows language switches mid-call.

Flow

  1. User uploads voice memo and selects source/target languages.
  2. App transcribes audio and detects language switches per segment.
  3. App returns timestamped transcript with language tags and speaker labels.

Models

  • Muse Voice Transcribe 1.0 for turning speech into text with automatic language detection and code-switch handling.
  • Muse Spark 1.3 for text and image understanding and structured answers to segment transcript by language, add timestamps and speaker labels.

Limits

  • Accuracy drops on heavy accents, overlapping speech, or low audio quality.
  • No real-time streaming; requires uploaded file.
  • Speaker diarization is approximate without enrollment.

Cost per run: ~$0.012

[no final answer within 512 new tokens]

Base better

Base 9 · Tuned 1

Provides complete brief in required shape; the other gives no answer.

Prompt 31

sketch of my tiny balcony garden on napkin paper attached, want tomatoes and mint in pots, need an app that redraws it cleaner so I can shop

Name: Balcony Garden Helper Tagline: Turn your napkin sketch into a clean pot plan to shop from. Flow: 1. Snap a photo of your napkin sketch and mark tomatoes and mint. 2. Get a cleaner redraw with…Show allShow less

Name: Balcony Garden Helper
Tagline: Turn your napkin sketch into a clean pot plan to shop from.

Flow:

  1. Snap a photo of your napkin sketch and mark tomatoes and mint.
  2. Get a cleaner redraw with pots laid out for your tiny balcony.
  3. Save a simple shopping list for pots, soil, and plants.

Models:

  • Muse Spark 1.3 for text and image understanding and structured answers to read the sketch and make the list
  • SAM 3.1 for finding, outlining and counting objects in photos and video to outline pots
  • Muse Image 1.0 for generating or editing images to redraw the clean layout

Limits:

  • Sizes are rough, not exact measurements from a napkin
  • May misread messy handwriting or faint lines
  • No live store stock or price check

Cost per run: ~$0.015

**Balcony Garden Redraw** Tagline: Clean up your napkin sketch into a shoppable planting plan. Flow 1. Upload napkin photo and confirm pots for tomatoes and mint. 2. App extracts layout and redraws…Show allShow less

Balcony Garden Redraw

Tagline: Clean up your napkin sketch into a shoppable planting plan.

Flow

  1. Upload napkin photo and confirm pots for tomatoes and mint.
  2. App extracts layout and redraws a clean top-down garden plan with labeled pots.
  3. Get a shopping list with pot sizes, soil and plant counts.

Models

  • Muse Spark 1.3 for text and image understanding and structured answers to parse the sketch, identify balcony space and pot placements, and generate shopping list JSON.
  • SAM 3.1 for finding, outlining and counting objects in photos and video to segment pots, plants and balcony boundaries from the napkin sketch.
  • Muse Image 1.0 for generating or editing images to redraw a clean vector-style garden layout.
  • Muse Wonder 1.0 for long multi-step planning to optimize tomato and mint placement for sun, spacing and pot size.

Limits
Sketch must be legible; handwriting recognition can miss labels. Redraw is illustrative, not to scale. No

[no final answer within 512 new tokens]

Base better

Base 5 · Tuned 1

Offers structured garden brief with flow and models though cut off, while empty response helps not at all.

Prompt 32

photo of a parking sign on 5th and Main that says 2H 8AM to 6PM attached, need an app that reads it and tells me if I can park now

Name: Park Check Now Tagline: Snap any parking sign, get a yes or no now. Flow: 1. You take a photo of the sign at 5th and Main. 2. You confirm current time and day. 3. You get can-park / can't-park…Show allShow less

Name: Park Check Now
Tagline: Snap any parking sign, get a yes or no now.
Flow:

  1. You take a photo of the sign at 5th and Main.
  2. You confirm current time and day.
  3. You get can-park / can't-park plus how long.
    Models:
  • Muse Spark 1.3 for text and image understanding and structured answers: reads "2H 8AM to 6PM" and decides rule.
  • SAM 3.1 for finding, outlining and counting objects in photos and video: crops the sign from background.
    Limits:
  • Only reads signs visible in photo, misses others out of frame.
  • Doesn't know holidays, permits, or temporary no-parking.
  • Needs correct phone time zone to judge.
    Cost per run: $0.004
**ParkNow Checker** Tagline: Snap a sign, know if you can park right now. Flow 1. User uploads a photo of a parking sign and taps Check. 2. App reads the sign text and extracts time limits, days, an…Show allShow less

ParkNow Checker
Tagline: Snap a sign, know if you can park right now.

Flow

  1. User uploads a photo of a parking sign and taps Check.
  2. App reads the sign text and extracts time limits, days, and restrictions.
  3. App compares to current local time and shows Park / No Park with reason.

Models

  • Muse Spark 1.3 for text and image understanding and structured answers: OCR the sign, parse “2H 8AM to 6PM” rules and output structured parking constraints.
  • SAM 3.1 for finding, outlining and counting objects in photos and video: locate and crop the sign in the photo to improve reading accuracy.

Limits

  • Needs clear, legible photo of the sign; glare or angle can reduce accuracy.
  • Cannot verify real-time enforcement, meter payment status, or temporary tow-away notices.
  • Assumes user’s device time zone is correct.

Cost per run: ~$0.008

[no final answer within 512 new tokens]

Base better

Base 9 · Tuned 1

Finished brief gives correct models, three steps and limits within word count while the missing response offers nothing.

Prompt 33

photo of my fridge whiteboard with dinner ideas attached, need an app that turns it into a 5 item grocery list in plain casual tone

Name: Whiteboard Grocery Helper Tagline: Turns fridge scribbles into a quick shopping list. Flow: 1. Snap or upload the photo of your fridge whiteboard. 2. Confirm the dinner ideas it spotted. 3. Ge…Show allShow less

Name: Whiteboard Grocery Helper
Tagline: Turns fridge scribbles into a quick shopping list.

Flow:

  1. Snap or upload the photo of your fridge whiteboard.
  2. Confirm the dinner ideas it spotted.
  3. Get a 5-item grocery list in plain casual tone.

Models:

  • Muse Spark 1.3 for text and image understanding and structured answers: to read the whiteboard dinner ideas and write the 5-item list.

Limits:

  • Messy handwriting, glare, or cropped photos can cause missed items.
  • Doesn't know what you already have at home, so may list staples.
  • Sticks to 5 items max, will leave off extras.

Cost per run: $0.004

**Name:** Fridge List Maker **Tagline:** Turn your whiteboard dinners into a quick grocery list. **Flow** 1. User uploads a photo of the fridge whiteboard. 2. App reads the dinner ideas and extract…Show allShow less

Name: Fridge List Maker

Tagline: Turn your whiteboard dinners into a quick grocery list.

Flow

  1. User uploads a photo of the fridge whiteboard.
  2. App reads the dinner ideas and extracts ingredients, then condenses to 5 essential grocery items.
  3. App returns a plain casual tone list with a one-tap copy option.

Models

  • SAM 3.1 for finding, outlining and counting objects in photos and video — isolate the whiteboard region and clean background clutter.
  • Muse Spark 1.3 for text and image understanding and structured answers — read handwritten dinner ideas, infer ingredients, and generate a 5-item grocery

[no final answer within 512 new tokens]

Base better

Base 4 · Tuned 1

Gives name tagline and user flow from the photo while the empty response gives nothing, though it is cut off before limits and cost.

Prompt 34

screenshot of my email from landlord about noise attached, need an app that drafts a polite reply under 70 words, apologetic tone

Name: Kind Reply Maker Tagline: Turn tense landlord notes into calm, polite replies. Flow: 1. Upload screenshot of landlord email about noise. 2. Confirm tone: apologetic and under 70 words. 3. Copy…Show allShow less

Name: Kind Reply Maker
Tagline: Turn tense landlord notes into calm, polite replies.
Flow:

  1. Upload screenshot of landlord email about noise.
  2. Confirm tone: apologetic and under 70 words.
  3. Copy the draft and send your reply.
    Models:
  • Muse Spark 1.3 for text and image understanding and structured answers: reads screenshot text and drafts the short reply.
    Limits:
  • May misread blurry or cropped screenshots.
  • Drafts are generic, not legal advice.
    Cost per run: ~$0.003
**Name:** Landlord Reply Helper **Tagline:** Draft a polite, under-70-word apology from a landlord email screenshot. **Flow** 1. User uploads screenshot of landlord noise email. 2. App reads the ema…Show allShow less

Name: Landlord Reply Helper
Tagline: Draft a polite, under-70-word apology from a landlord email screenshot.

Flow

  1. User uploads screenshot of landlord noise email.
  2. App reads the email and drafts a <70 word apologetic reply in polite tone.
  3. User reviews, edits, and copies the reply.

Models

  • Muse Spark 1.3 for text and image understanding and structured answers — reads the screenshot, extracts complaint details and context, and generates a concise apologetic reply under 70 words.

Limits

  • Cannot verify sender authenticity or email metadata

[no final answer within 512 new tokens]

Base better

Base 6 · Tuned 1

Provides relevant brief with correct model and steps; the empty response serves nothing.

Prompt 35

photo of my vinyl collection shelf attached, need an app that lists albums A to Z in a neat text table with no extra chat

**Name:** Vinyl Shelf Lister **Tagline:** Snap your shelf, get a clean A-to-Z album list. **Flow:** 1. User uploads photo of vinyl collection shelf. 2. App detects spines and reads artist / album t…Show allShow less

Name: Vinyl Shelf Lister

Tagline: Snap your shelf, get a clean A-to-Z album list.

Flow:

  1. User uploads photo of vinyl collection shelf.
  2. App detects spines and reads artist / album titles.
  3. App outputs A-to-Z text table with no extra chat.

Models:

  • SAM 3.1 for finding, outlining and counting album spines in the photo.
  • Muse Spark 1.3 for text and image understanding and structured answers to read titles and format the table.

Limits:

  • Blurry or angled spines may be misread or skipped.
  • Similar reissues / box sets may be merged incorrectly.

Cost per run: $0.008

**Name:** Vinyl Shelf Indexer **Tagline:** A-Z vinyl list from a shelf photo, no chatter. **Flow** 1. User uploads shelf photo and confirms sort order A-Z. 2. App detects each record spine, reads ti…Show allShow less

Name: Vinyl Shelf Indexer
Tagline: A-Z vinyl list from a shelf photo, no chatter.

Flow

  1. User uploads shelf photo and confirms sort order A-Z.
  2. App detects each record spine, reads title/artist, and builds a table.
  3. App outputs a clean text table and offers CSV export.

Models

  • SAM 3.1 for finding, outlining and counting objects in photos and video — isolate each vinyl spine in the shelf image.
  • Muse Spark 1.3 for text and image understanding and structured answers — read spine text and produce structured album/artist entries.
  • Muse Wonder 1.0 for long multi-step planning — plan detection → OCR → deduplication → A-Z sorting pipeline

[no final answer within 512 new tokens]

Base better

Base 5 · Tuned 1

Gives usable brief with spine detection and table formatting while the other provides no answer at all.

Prompt 36

photo of my cracked iPhone 12 screen after I dropped it on tile attached, need an app that looks close and says screen protector fix or repair shop

**Name:** Screen Crack Triage **Tagline:** Snap your cracked screen, learn if it's protector or shop time. **Flow:** 1. Upload your cracked iPhone 12 photo from the tile drop. 2. See close-up crack o…Show allShow less

Name: Screen Crack Triage
Tagline: Snap your cracked screen, learn if it's protector or shop time.
Flow:

  1. Upload your cracked iPhone 12 photo from the tile drop.
  2. See close-up crack outline with damage severity.
  3. Get verdict: replace screen protector or go to repair shop.
    Models:
  • Muse Spark 1.3 for text and image understanding and structured answers
  • SAM 3.1 for finding, outlining and counting objects in photos and video
    Limits:
  • Can't always separate protector crack vs glass in glare or blur.
  • Won't test touch, Face ID, or water resistance loss.
    Cost per run: ~$0.012
**Name:** Crack Check Finder **Tagline:** Snap your cracked screen, get a fix or repair shop nearby. **Flow** 1. User uploads photo of cracked iPhone 12 screen. 2. App outlines damage and classifie…Show allShow less

Name: Crack Check Finder

Tagline: Snap your cracked screen, get a fix or repair shop nearby.

Flow

  1. User uploads photo of cracked iPhone 12 screen.
  2. App outlines damage and classifies it as screen protector vs glass crack.
  3. App shows on-image label “Screen Protector Fix or Repair Shop” and lists nearby options.

Models

  • SAM 3.1 for finding, outlining and counting objects in photos and video — segment the crack area on the screen.
  • Muse Spark 1.3 for text and image understanding and structured answers — interpret crack pattern and decide protector vs glass damage.
  • Muse Image 1.0 for
Name: Crack Check Fixer Tagline: Snap your cracked screen, get fix or shop advice fast. Flow: 1. Upload your iPhone 12 screen photo after the drop. 2. Confirm crack location and size from the close-u…Show allShow less

Name: Crack Check Fixer
Tagline: Snap your cracked screen, get fix or shop advice fast.
Flow:

  1. Upload your iPhone 12 screen photo after the drop.
  2. Confirm crack location and size from the close-up view.
  3. Get screen protector fix vs. repair shop recommendation.
    Models:
  • SAM 3.1 for finding, outlining and counting cracks in photos and video
  • Muse Spark 1.3 for text and image understanding and structured answers
    Limits:
  • Can't detect internal damage or touch issues from photo alone.
  • May misjudge crack severity in dark or blurry close-ups.
    Cost per run: $0.008
Tuned better

Base 3 · Tuned 9

Complete brief with correct models and limits stays under word count; the other is cut off and adds unneeded image generation.

Prompt 37

screenshot of the outage map showing 342 homes out on Elm St attached, need an app that rewrites it as a calm text I can send my neighbor

**Name:** Calm Outage Texter **Tagline:** Turn scary outage maps into a calm neighbor update. **Flow:** 1. Upload your outage map screenshot showing Elm St and 342 homes out. 2. Pick tone: calm, re…Show allShow less

Name: Calm Outage Texter

Tagline: Turn scary outage maps into a calm neighbor update.

Flow:

  1. Upload your outage map screenshot showing Elm St and 342 homes out.
  2. Pick tone: calm, reassuring, or brief check-in.
  3. Copy the short text to send your neighbor.

Models:

  • Muse Spark 1.3 for text and image understanding and structured answers, reads the screenshot and rewrites it as a calm message.

Limits:

  • Can't verify if outage data is live or accurate.
  • May misread blurry numbers or cropped screenshots.
  • Not for emergency or medical safety alerts.

Cost per run: ~$0.003

**Name:** Calm Outage Messenger **Tagline:** Turn outage maps into neighbor-friendly updates. **Flow** 1. User uploads the outage map screenshot. 2. App reads the map and extracts homes affected, st…Show allShow less

Name: Calm Outage Messenger
Tagline: Turn outage maps into neighbor-friendly updates.

Flow

  1. User uploads the outage map screenshot.
  2. App reads the map and extracts homes affected, street and status.
  3. App returns a calm, shareable text message for the neighbor.

Models

  • Muse Spark 1.3 for text and image understanding and structured answers — read the screenshot, extract 342 homes out on Elm St and rewrite as calm neighbor text.

Limits

  • Accuracy depends on screenshot clarity and legibility.
  • No real-time verification of outage data beyond what’s visible.
  • Does not send messages, only drafts text for you to copy.

Cost per run: ~$0.008

Name: Elm Street Calm Note Tagline: Turn scary outage maps into a calm neighbor update. Flow: 1. Upload your outage map screenshot. 2. Confirm Elm St and 342 homes. 3. Copy the calm text to send. Mod…Show allShow less

Name: Elm Street Calm Note
Tagline: Turn scary outage maps into a calm neighbor update.
Flow:

  1. Upload your outage map screenshot.
  2. Confirm Elm St and 342 homes.
  3. Copy the calm text to send.
    Models:
  • Muse Spark 1.3 for text and image understanding and structured answers to read the screenshot and rewrite it as a calm message.
    Limits:
  • Can't verify live utility status, may misread blurry screenshots.
  • Not an official outage notice, don't use for emergencies.
    Cost per run: ~$0.003
Tuned better

Base 8 · Tuned 9

Keeps flow as user actions with specific Elm St detail; the other describes app actions not user steps.

Prompt 38

I'm already late and my bus never showed, screenshot of the transit app attached, need an app that tells me the fastest backup route in 2 lines flat

**Name:** Missed Bus Rescue **Tagline:** Your fastest Plan B in two lines flat. **Flow:** 1. You upload the transit app screenshot and share your location. 2. You get a 2-line backup route with wal…Show allShow less

Name: Missed Bus Rescue

Tagline: Your fastest Plan B in two lines flat.

Flow:

  1. You upload the transit app screenshot and share your location.
  2. You get a 2-line backup route with walk, bus, or ride time.
  3. You tap Go to start navigation.

Models:

  • Muse Spark 1.3 for text and image understanding and structured answers to read the screenshot and write the 2-line route.
  • Muse Wonder 1.0 for long multi-step planning to compare backup routes.

Limits:

  • Needs live transit data, otherwise times may be stale.
  • Can't track sudden cancellations or delays in real-time.
  • Blurry screenshots may be misread.

Cost per run: $0.008

**Late Bus Rescue** Tagline: Get a 2-line backup route from your transit screenshot instantly. Flow 1. User uploads transit app screenshot and confirms start/end. 2. App reads the missed bus info a…Show allShow less

Late Bus Rescue

Tagline: Get a 2-line backup route from your transit screenshot instantly.

Flow

  1. User uploads transit app screenshot and confirms start/end.
  2. App reads the missed bus info and current time from the image.
  3. App returns a 2-line fastest backup route with time and mode.

Models

  • Muse Spark 1.3 for text and image understanding and structured answers to extract route, stop, delay from screenshot and generate the 2-line summary.
  • Muse Wonder 1.0 for long multi-step planning to compare alternatives and pick fastest backup.

Limits

  • Requires clear screenshot with readable route/time; blurry images fail.
  • No live map data, so estimates assume current traffic is normal.
  • Only outputs 2 lines, no detailed turn-by-turn.

Cost per run: ~$0.008

[no final answer within 512 new tokens]

Base better

Base 9 · Tuned 1

Delivers complete brief in correct format with valid models; the other provides no answer at all.

Prompt 39

PDF of my son Jayden's report card attached, need an app that summarizes grades and missing work for a stressed mom

**Name:** Jayden Report Calm **Tagline:** A calm, clear summary of grades and missing work. **Flow:** 1. Upload Jayden's report card PDF. 2. Get a simple table of grades, missing work, and teacher…Show allShow less

Name: Jayden Report Calm

Tagline: A calm, clear summary of grades and missing work.

Flow:

  1. Upload Jayden's report card PDF.
  2. Get a simple table of grades, missing work, and teacher notes.
  3. Tap to make a to-do checklist and encouraging message.

Models:

  • Muse Spark 1.3 for text and image understanding and structured answers

Limits:

  • May misread blurry scans or handwritten comments.
  • Cannot check live grades in school portals, only the uploaded PDF.
  • Not academic or parenting advice, just a summary.

Cost per run: ~$0.004

Name: Jayden Report Summarizer
Tagline: Calm, clear grade snapshot for busy parents.

Flow

  1. Upload Jayden's report card PDF or photo.
  2. App extracts courses, grades, and
Name: Jayden Report Helper Tagline: Calm, clear grades and next steps for Jayden. Flow: 1. Upload Jayden's report card PDF. 2. Get a simple summary of grades and missing work. 3. Get a short checklis…Show allShow less

Name: Jayden Report Helper
Tagline: Calm, clear grades and next steps for Jayden.
Flow:

  1. Upload Jayden's report card PDF.
  2. Get a simple summary of grades and missing work.
  3. Get a short checklist to share with teacher.
    Models:
  • Muse Spark 1.3 for text and image understanding and structured answers to read the PDF and summarize grades and missing assignments.
    Limits:
  • May misread blurry scans or handwritten notes.
  • Not official school advice, check with teacher.
    Cost per run: $0.004
Tuned better

Base 3 · Tuned 9

Complete brief follows required shape with models limits and cost; the other is cut off mid-flow.

Prompt 40

Pic of the empty formula shelf at Target attached, I'm exhausted, need an app that finds what to do next in 2 calm steps

**Name:** Calm Formula Finder **Tagline:** Empty shelf to next step, calmly. **Flow:** 1. Snap the empty shelf, share zip code. 2. See 2 calm options: where else to check or what to ask. 3. Tap for…Show allShow less

Name: Calm Formula Finder

Tagline: Empty shelf to next step, calmly.

Flow:

  1. Snap the empty shelf, share zip code.
  2. See 2 calm options: where else to check or what to ask.
  3. Tap for map, call script, or rest note.

Models:

  • Muse Spark 1.3 for image understanding and structured answers
  • SAM 3.1 for finding and counting remaining cans on shelf
  • Muse Wonder 1.0 for planning next stores and steps

Limits:

  • No live Target inventory, info may be outdated.
  • No medical advice on formula swaps, confirm with doctor.

Cost per run: $0.018

**Name:** Shelf Calm Finder **Tagline:** Calm next steps when the shelf is empty. **Flow** 1. User uploads photo of empty formula shelf and taps “Find next step”. 2. App confirms what’s missing and…Show allShow less

Name: Shelf Calm Finder

Tagline: Calm next steps when the shelf is empty.

Flow

  1. User uploads photo of empty formula shelf and taps “Find next step”.
  2. App confirms what’s missing and suggests 2 calm actions.
  3. User taps an action to get a ready-to-send message or store check.

Models

  • Muse Spark 1.3 for text and image understanding and structured answers — reads shelf labels, context, and returns a calm 2-step plan.
  • SAM 3.1 for finding, outlining and counting objects in photos and video — detects empty slots and product types in the Target shelf image.
  • Muse Wonder 1.0 for long multi-step planning — sequences the 2 calm steps with fallback options.

Limits

  • Works only on clear shelf photos; blurry or dark images reduce accuracy.
  • Does not place orders or verify real-time store inventory.
  • Advice is general, not medical nutrition guidance.

Cost per run: ~$0.018

Name: Calm Shelf Helper Tagline: Turn an empty shelf photo into 2 calm next steps. Flow: 1. Upload your Target formula shelf pic. 2. Confirm your baby's age and formula type. 3. Get 2 calm steps to d…Show allShow less

Name: Calm Shelf Helper
Tagline: Turn an empty shelf photo into 2 calm next steps.
Flow:

  1. Upload your Target formula shelf pic.
  2. Confirm your baby's age and formula type.
  3. Get 2 calm steps to do next.
    Models:
  • SAM 3.1 for finding, outlining empty shelf space in the photo
  • Muse Spark 1.3 for text and image understanding and structured answers to give the 2-step plan
    Limits:
  • Can't check live Target stock or delivery times
  • Not medical advice, check with pediatrician for formula changes
  • May misread blurry or crowded shelf photos
    Cost per run: ~$0.008
Base better

Base 9 · Tuned 7

Gives three model plan with sequencing and calmer two action flow; the other asks extra input and skips planning model.