Every round, every result

Round 1 - Spark styles (10 themes)

• Default kit look: 8 / 6
• 16-Bit Pixel Art: 8 / 7
• Low-Poly 3D: 8 / 6
• Claymation: 4 / 3

Round 2 - fog/exposure guardrails

• Default kit look: 9 / 6
• 16-Bit Pixel Art: 9 / 8
• Lego style: 8 / 7
• Low-Poly 3D: 9 / 8

Round 2b - 12 more themes

• 8-Bit Pixel Art: 8 / 6
• Anime: 3 / 6
• Vaporwave: 6 / 8
• Steampunk: 3 / 5

Round 3 - Sol art styles (ink / halftone / watercolor / woodblock / clay / grid / rain)

• Comic book (Sol): 8 / 7
• Watercolor (Sol): 6 / 7
• Ukiyo-e Woodblock (Sol): 4 / 6
• Claymation (Sol): 6 / 8

Round 4a - Sol styles for 11 weakest (scene extras OFF by mistake)

• Ukiyo-e Woodblock (Sol r4): 6 / 7
• Stained Glass (Sol r4): 3 / 6
• Anime (Sol r4): 3 / 7
• Steampunk (Sol r4): 2 / 6

Round 4b - same with Sol's surfaces + scene extras ON

• Ukiyo-e Woodblock (Sol r4): 6 / 5
• Stained Glass (Sol r4): 9 / 8
• Anime (Sol r4): 6 / 7
• Steampunk (Sol r4): 6 / 7

Spaceships (Quaternius + Kenney)

• top: Q ZENITH 9, Q EXECUTIONER 8, Q OMEN 8, K SPEEDERA 8
• lowest: Q STRIKER 6, Q IMPERIAL 5, Q PANCAKE 5
• sheet note: Inconsistent scale: The Kenney assets (K) are significantly larger and chunkier than the Quaternius assets (Q), making them look like different game g
• sheet note: Lighting/Bloom: The bloom effect is too strong on the white/yellow Kenney assets, causing them to lose edge definition and look 'glowy' rather than so

Pixel art / 2D packs (Kenney)

• top: PLATFORMER ART DELUXE 10, PIXEL PLATFORMER 9, TINY TOWN 9, PIXEL SHMUP 8
• lowest: MICRO ROGUELIKE 7, 1-BIT PACK (FANTASY) 6, 1-BIT PACK (PLATFORMER) 5
• sheet note: Inconsistent scaling: The 'Space Shooter' items appear significantly larger than the other assets, disrupting the visual rhythm of the sheet.
• sheet note: Layout imbalance: The 'Space Shooter' row has excessive empty black space around the assets compared to the tightly packed rows above.

PBR textures (ambientCG / Poly Haven)

• top: blue character 9, ROCK020 8, capsule 8, GROUND037 7
• lowest: PLANKS021 5, PAVINGSTONES070 5, BRICKS054 4
• sheet note: The lighting is very flat and ambient, lacking strong directional shadows which makes the PBR materials (especially roughness) hard to evaluate accura
• sheet note: The background is a simple gradient which doesn't provide context for how these assets would look in a real game environment.

Characters (packs + 100 Avatars retarget)

• top: KAYKIT KNIGHT 8, QUATERNIUS CHARACTER 8, KENNEY ARCHER 8, KAYKIT BARBARIAN 8
• lowest: COOLALIEN 3, OLDMOUSTACHE 3, JIMMY 2
• sheet note: Inconsistent scale: The back row avatars are significantly smaller than the front row assets.
• sheet note: Lighting: The scene lighting is flat and lacks ambient occlusion, making the VRM avatars look like untextured clay.

MINI GOLF

• 16.3 min to build · 21 KB file
• first test: 1 problems
• repairs: 2 Qwen + 2 Sol · art rounds 1
• art score 8/10 · 100 fps

TOWER GUARD

• 15.5 min to build · 26 KB file
• first test: 5 problems
• repairs: 2 Qwen + 2 Sol
• art score 6/10 · 100 fps

BULLSEYE

• 16 min to build · 23 KB file
• first test: 2 problems
• repairs: 2 Qwen + 2 Sol
• art score 3/10 · 100 fps

CAVE COPTER

• 13.8 min to build · 16 KB file
• first test: 10 problems
• repairs: 2 Qwen + 2 Sol
• art score 6/10 · 100 fps

BOX PUSH

• 14.8 min to build · 17 KB file
• first test: pass
• repairs: 0 Qwen + 0 Sol
• art score 8/10 · 100 fps

LAKE DAY

• 16.8 min to build · 22 KB file
• first test: 6 problems
• repairs: 2 Qwen + 0 Sol
• art score 7.5/10 · 100 fps

STAR LANCE

• first test: pass
• repairs: 0 Qwen + 0 Sol
• art score 8/10 · 100 fps
• A side-scrolling shooter scene featuring a low-poly spaceship flying over a dark, geometric landscape with a large ringed planet in the back

ARENA STORM

• first test: 5 problems
• repairs: 0 Qwen + 4 Sol
• art score 8/10 · 100 fps
• A top-down tactical shooter scene featuring a player character, red enemies, and obstacles on a dark grid background with clear UI elements.

TOPPLE

• first test: pass
• repairs: 0 Qwen + 0 Sol
• art score 8/10 · 100 fps
• A physics puzzle game scene featuring a table with glass orbs and a tipped-over pot on a platform, with a clean UI overlay and readable obje

TILE QUEST

• first test: pass
• repairs: 0 Qwen + 0 Sol
• art score 8.5/10 · 100 fps
• An isometric puzzle level featuring a blue character, a wooden crate, and a green target dock on a clean grid, surrounded by green blocks.

TRACK STARS

• first test: 3 problems
• repairs: 0 Qwen + 3 Sol
• art score 8/10 · 100 fps
• A stylized 3D track-and-field game showing a purple character crossing the finish line with a results panel overlaying the right side.

MINI GOLF

• 18.7 min to build · 21 KB file
• first test: 4 problems
• repairs: 2 Qwen + 0 Sol · art rounds 1
• art score 8/10 · 100 fps

TOWER GUARD

• 13.1 min to build · 24 KB file
• first test: 4 problems
• repairs: 2 Qwen + 2 Sol
• art score 6/10 · 100 fps

BULLSEYE

• 17.4 min to build · 19 KB file
• first test: 5 problems
• repairs: 2 Qwen + 1 Sol · art rounds 1 (1 undone)
• art score 5/10 · 100 fps

CAVE COPTER

• 16.7 min to build · 17 KB file
• first test: 7 problems
• repairs: 2 Qwen + 1 Sol
• art score 8/10 · 100 fps

BOX PUSH

• 23.2 min to build · 19 KB file
• first test: pass
• repairs: 0 Qwen + 0 Sol
• art score 8/10 · 100 fps

LAKE DAY

• 16 min to build · 25 KB file
• first test: 6 problems
• repairs: 2 Qwen + 0 Sol
• art score 7.5/10 · 100 fps

NEON PINBALL rebuild (orig 5 rounds)

• 14.1 min to build · 19 KB file
• first test: 1 problems
• repairs: 1 Qwen + 0 Sol
• art score 7.5/10 · 100 fps

MARBLE MAZE rebuild (orig 6 rounds)

• 12.3 min to build · 21 KB file
• first test: 6 problems
• repairs: 2 Qwen + 1 Sol · art rounds 1
• art score 8/10 · 100 fps

AURORA rebuild (orig 6 rounds)

• 11.8 min to build · 15 KB file
• first test: 6 problems
• repairs: 2 Qwen + 1 Sol
• art score 7/10 · 100 fps

DASH rebuild (orig 6 rounds)

• 15.5 min to build · 18 KB file
• first test: 16 problems
• repairs: 2 Qwen + 1 Sol
• art score 7/10 · 100 fps

CASTLE CRASH rebuild (orig 4 rounds)

• 13.4 min to build · 22 KB file
• first test: 2 problems
• repairs: 2 Qwen + 2 Sol
• art score 7.5/10 · 100 fps

DOMINO RUN rebuild (orig 9 rounds)

• 11 min to build · 17 KB file
• first test: 4 problems
• repairs: 2 Qwen + 2 Sol
• art score 7.5/10 · 100 fps

KART CLASH (kart arena)

• 10.1 min to build · 19 KB file
• first test: pass
• repairs: 0 Qwen + 0 Sol · art rounds 1
• art score 7/10 · 99 fps

POWDER (snowboard runner)

• 11.2 min to build · 19 KB file
• first test: 2 problems
• repairs: 2 Qwen + 1 Sol · art rounds 1 (1 undone)
• art score 6/10 · 82 fps

TANK TOSS (artillery physics)

• 13.3 min to build · 19 KB file
• first test: 6 problems
• repairs: 2 Qwen + 1 Sol · art rounds 2
• art score 6.5/10 · 90 fps

BEAT TAP (rhythm)

• 8.7 min to build · 16 KB file
• first test: 4 problems
• repairs: 2 Qwen + 2 Sol
• art score 8/10 · 100 fps

DEEP BLUE (WebGPU underwater)

• 12.7 min to build · 21 KB file
• first test: 8 problems
• repairs: 2 Qwen + 0 Sol
• art score 7/10 · 100 fps

SNOW CIRCUIT (remix with palette rules)

• 4.3 min to build
• first test: 2 problems
• repairs: 0 Qwen + 4 Sol
• art score 5/10

BOWL (bowling physics)

• 20.8 min to build · 25 KB file
• first test: pass
• repairs: 0 Qwen + 0 Sol · art rounds 2 (1 undone)
• art score 6/10 · 100 fps

NEON SWARM (twin-stick shooter)

• 10.5 min to build · 19 KB file
• first test: 1 problems
• repairs: 1 Qwen + 0 Sol
• art score 7.5/10 · 100 fps

SKY HOP (3D platformer)

• 13 min to build · 21 KB file
• first test: 1 problems
• repairs: 2 Qwen + 1 Sol
• art score 7/10 · 100 fps

GEM MATCH (3D match-3)

• 10.9 min to build · 22 KB file
• first test: 4 problems
• repairs: 1 Qwen + 1 Sol
• art score 8/10 · 100 fps

AURORA (WebGPU aurora flight)

• 13.5 min to build · 17 KB file
• first test: 1 problems
• repairs: 1 Qwen + 0 Sol
• art score 7.5/10 · 100 fps

DESERT RALLY (remix of mrdoob Starter Kit Racing - palette + props)

• 2.5 min to build
• first test: 1 problems
• repairs: 0 Qwen + 2 Sol
• art score 6/10

STAR FORGE (WebGPU compute galaxy sandbox)

• 36.3 min to build · 21 KB file
• first test: 3 problems
• repairs: 0 Qwen + 2 Sol · art rounds 1
• art score 6/10 · 100 fps

NEON PINBALL (physics)

• 17.6 min to build · 20 KB file
• first test: 3 problems
• repairs: 0 Qwen + 2 Sol
• art score 8.5/10 · 100 fps

CASTLE CRASH (slingshot physics)

• 9.6 min to build · 20 KB file
• first test: 2 problems
• repairs: 1 Qwen + 1 Sol
• art score 7/10 · 100 fps

SAIL (WebGPU ocean racer)

• 15.9 min to build · 19 KB file
• first test: 7 problems
• repairs: 2 Qwen + 0 Sol · art rounds 2
• art score 6.5/10 · 100 fps

BLOCK PULL (tower physics)

• 11.4 min to build · 18 KB file
• first test: 2 problems
• repairs: 0 Qwen + 1 Sol
• art score 6/10 · 47 fps

HURDLES (110 m button-masher)

• 19.9 min to build · 17 KB file
• first test: 1 problems
• repairs: 0 Qwen + 2 Sol · art rounds 1 (1 undone)
• art score 5/10 · 57 fps

NEON NIGHT RACER (remix of mrdoob Starter Kit Racing)

• 2.6 min to build
• first test: 1 problems
• repairs: 0 Qwen + 2 Sol · art rounds 1
• art score 6/10

ASTRO SIGNAL (WebGPU compute particles - audio-reactive)

• 14.9 min to build · 17 KB file
• first test: 4 problems
• repairs: 2 Qwen + 0 Sol
• art score 8/10 · 100 fps

DOMINO RUN (physics - cannon-es)

• 16.5 min to build · 16 KB file
• first test: 7 problems
• repairs: 2 Qwen + 2 Sol · art rounds 1
• art score 7/10 · 91 fps

MARBLE MAZE (tilt physics - cannon-es)

• 15.4 min to build · 18 KB file
• first test: 2 problems
• repairs: 1 Qwen + 0 Sol · art rounds 1 (1 undone)
• art score 6/10 · 100 fps

PIXEL HAND 3D (3D boss shooter)

• 12 min to build · 21 KB file
• first test: 7 problems
• repairs: 2 Qwen + 1 Sol
• art score 8/10 · 100 fps

STACK - LOW thinking A/B (compare STACK v2 medium)

• 9.1 min to build · 13 KB file
• first test: 4 problems
• repairs: 2 Qwen + 2 Sol · art rounds 1 (1 undone)
• art score 6/10 · 100 fps

STACK v2 (stacker)

• 11.7 min to build · 14 KB file
• first test: pass
• repairs: 0 Qwen + 0 Sol
• art score 6/10 · 100 fps

DASH 100 m (button-masher)

• 17.5 min to build · 18 KB file
• first test: 1 problems
• repairs: 0 Qwen + 2 Sol
• art score 6/10 · 98 fps

PIXEL HAND v2 (boss shooter)

• 18.6 min to build · 22 KB file
• first test: 5 problems
• repairs: 2 Qwen + 2 Sol
• art score 7/10 · 100 fps

NEON RUNNER v2 (runner)

• 12.1 min to build · 20 KB file
• first test: pass
• repairs: 0 Qwen + 0 Sol · art rounds 1
• art score 7/10 · 100 fps

TOWER TOPPLE (physics - cannon-es)

• 15 min to build · 18 KB file
• first test: 1 problems
• repairs: 1 Qwen + 0 Sol
• art score 6/10 · 100 fps

CAN TOSS (physics hero - cannon-es)

• 11.7 min to build · 15 KB file
• first test: 6 problems
• repairs: 2 Qwen + 1 Sol
• art score 6/10 · 100 fps

Sky Rings (3D flight)

• 4.8 min total
• 0 of 0 buttons did something

Low-Poly Racer (3D racing)

• 5 min total
• 0 of 0 buttons did something

Maze Escape (first-person 3D)

• 5 min total
• 0 of 0 buttons did something

Marble Tilt (3D physics)

• 6.5 min total
• 0 of 0 buttons did something

Asteroid Run (3D space shooter)

• 3.8 min total
• 0 of 4 buttons did something

Neon Runner (3D endless runner)

• 5 min total
• 0 of 0 buttons did something

Stack (3D tower game)

• 4 min total
• 0 of 0 buttons did something

Sky Rings (3D flight)

• 4.5 min total
• 0 of 1 buttons did something

Low-Poly Racer (3D racing)

• 4.8 min total
• 0 of 0 buttons did something

Maze Escape (first-person 3D)

• 3.5 min total
• 0 of 0 buttons did something

Marble Tilt (3D physics)

• 5.3 min total
• 0 of 0 buttons did something

Asteroid Run (3D space shooter)

• 3.8 min total
• 0 of 0 buttons did something

Neon Runner (3D endless runner)

• 4.5 min total
• 0 of 0 buttons did something

Stack (3D tower game)

• 4 min total
• 0 of 0 buttons did something

Budget Buddy (money)

• 2.8 min total
• 1 of 2 buttons did something

Recipe Box (kitchen)

• 4 min total
• 10 of 18 buttons did something

Tic-Tac-Toe Arena (game)

• 2.8 min total
• 0 of 0 buttons did something

Gym Log (fitness)

• 4.8 min total
• 0 of 0 buttons did something

Bakery Desk (multi-section business)

• 3.8 min total
• 0 of 0 buttons did something

Club Hub (multi-section organizer)

• 3.3 min total
• 0 of 0 buttons did something

Shop Pulse (charts dashboard)

• 2.8 min total
• 0 of 0 buttons did something

Habit Rings (dashboard)

• 3.8 min total
• 18 of 18 buttons did something

Flashcard Sprint (education)

• 3 min total
• 2 of 12 buttons did something

Meadow Fox (3D low-poly scene)

• 4 min total
• 0 of 0 buttons did something

Space Dodge (canvas game)

• 2.8 min total
• 0 of 0 buttons did something

Trail Log (trip planner)

• 2.5 min total
• 0 of 0 buttons did something

Pixel Garden (idle game)

• 5.8 min total
• 18 of 18 buttons did something

Band Setlist (tool)

• 2.8 min total
• 0 of 2 buttons did something

Invoice Forge (business)

• 3.5 min total
• 0 of 0 buttons did something

Habit Rings (dashboard)

• 2.5 min total
• 0 of 0 buttons did something

Chord Trainer (audio)

• 2.3 min total
• 0 of 0 buttons did something

Shop Pulse (charts dashboard)

• 2.3 min total
• 0 of 0 buttons did something

Meadow Fox (3D low-poly scene)

• 4 min total

Pad Lab (synth drum pad)

• 3.5 min total
• 18 of 18 buttons did something

Flashcard Sprint (education)

• 3.5 min total
• 0 of 0 buttons did something

Space Invaders Qwen 3.8 alone (one prompt)

• 139 s
• 5626 tokens incl. thinking
• error: Cannot read properties of undefined (reading 'alive')
• prompt: Make a Space Invaders game: rows of aliens that march and shoot back, a player ship with arrow keys and space to fire, shields, score, lives, game over and restart.

Space Invaders After one fix round

• 147 s for the fix

Breakout

• 103 s
• 4402 tokens incl. thinking
• prompt: Make a Breakout game: paddle with mouse or arrow keys, a ball, rows of colored bricks, score, lives, levels, game over and restart.

Pelican riding a bicycle

• 283 s
• 9528 tokens incl. thinking
• prompt: Generate an SVG of a pelican riding a bicycle, shown full size on the page.

2048

• 103 s
• 4379 tokens incl. thinking
• prompt: Make the 2048 game: 4x4 grid, arrow keys and swipe, tiles merge and animate, score and best score, game over and new game.

Minesweeper

• 68 s
• 2969 tokens incl. thinking
• prompt: Make Minesweeper: 9x9 board with 10 mines, left click to reveal with flood fill, right click to flag, a timer and a mine counter, win and lose.

Asteroids Qwen 3.8 alone (one prompt)

• 112 s
• 4753 tokens incl. thinking
• error: particles is not iterable
• prompt: Make an Asteroids game: a ship that rotates, thrusts and shoots, asteroids that split when hit, screen wrap, score, lives, game over and restart.

Asteroids After one fix round

• 114 s for the fix

Wordle

• 135 s
• 5314 tokens incl. thinking
• prompt: Make a Wordle clone: guess a 5-letter word in 6 tries, an on-screen keyboard plus typing, green/yellow/gray letter colors, win and lose messages, new game.

Chess

• hands-on: moved the e-pawn two squares: the move list and turn changed
• 134 s
• 5252 tokens incl. thinking
• prompt: Make a two-player chess game: a board with all pieces, click to select and move, only legal moves (including castling and en passant), check and checkmate, move history.

Minecraft-style block world Qwen 3.8 alone (one prompt)

• 143 s
• 5030 tokens incl. thinking
• error: Uncaught TypeError: Cannot read properties of undefined (reading 'morphTargets')
• prompt: Make a small Minecraft-style voxel world with three.js: terrain made of blocks, walk with WASD and look with the mouse, left click breaks a block, right click places one.

Minecraft-style block world After one fix round

• 197 s for the fix

Kanban board

• hands-on: added a card: clicked 'Add a card', typed a title, pressed 'Add card'
• 122 s
• 5308 tokens incl. thinking
• prompt: Make a Kanban board: columns To do / Doing / Done, add, edit and delete cards, drag cards between columns, saved in localStorage.

Mario-style platformer

• 103 s
• 4134 tokens incl. thinking
• prompt: Make a side-scrolling platformer level like Super Mario: run and jump with the arrow keys, platforms, coins, enemies you stomp, a goal flag, lives and restart.

Pac-Man

• 896 s
• 29431 tokens incl. thinking
• prompt: Make a Pac-Man game: a maze with dots and power pellets, four ghosts that chase you, arrow keys, score, lives, game over and restart.

Space Dodge (canvas game)

• 7.8 min total
• plan 85 s · written 232 s · reviewed 260 s
• browser test: 10 acts ok, 0 failed
• 8 of 11 buttons did something

Pixel Garden (idle game)

• 13.5 min total
• plan 126 s · written 411 s · reviewed 434 s
• browser test: 8 acts ok, 7 failed
• 14 of 26 buttons did something

Trail Log (trip planner)

• 8 min total
• plan 76 s · written 233 s · reviewed 244 s
• browser test: 3 acts ok, 5 failed
• 23 of 27 buttons did something

Band Setlist (tool)

• 9.5 min total
• plan 117 s · written 227 s · reviewed 241 s
• browser test: 12 acts ok, 3 failed
• 9 of 13 buttons did something

Invoice Forge (business)

• 12.3 min total
• plan 92 s · written 258 s · reviewed 278 s
• browser test: 6 acts ok, 9 failed
• 10 of 17 buttons did something

Habit Rings (dashboard)

• 8.5 min total
• plan 61 s · written 188 s · reviewed 242 s
• browser test: 6 acts ok, 6 failed
• 7 of 28 buttons did something

Chord Trainer (audio)

• 8.8 min total
• plan 86 s · written 183 s · reviewed 202 s
• browser test: 6 acts ok, 6 failed
• 8 of 11 buttons did something

Meadow Fox (3D low-poly scene)

• 6.8 min total
• plan 57 s · written 239 s · reviewed 252 s
• browser test: 3 acts ok, 9 failed
• 5 of 9 buttons did something

Shop Pulse (charts dashboard)

• 12.5 min total
• plan 56 s · written 265 s · reviewed 305 s
• browser test: 2 acts ok, 2 failed
• 12 of 17 buttons did something

Pad Lab (synth drum pad)

• 18.3 min total
• plan 132 s · written 405 s · reviewed 448 s
• browser test: 6 acts ok, 4 failed
• 27 of 31 buttons did something

Flashcard Sprint (education)

• 12.3 min total
• plan 85 s · written 257 s · reviewed 313 s
• browser test: 10 acts ok, 5 failed
• 31 of 34 buttons did something

neon pixel arcade Designed (today's builder)

neon pixel arcade Custom (one prompt, medium thinking)

• 5345 tokens incl. thinking
• phone view: neon-pixel-arcade.test-six-phone.jpg

brightsmile dental Designed (today's builder)

brightsmile dental Custom (one prompt, medium thinking)

• 7724 tokens incl. thinking
• phone: text touches the screen edge: "Start Here"
• phone view: brightsmile-dental.test-six-phone.jpg

sunsave solar Designed (today's builder)

sunsave solar Custom (one prompt, medium thinking)

• 8568 tokens incl. thinking
• phone: text touches the screen edge: "SOLAR + BATTERY BACKUP · BUILT"
• phone view: sunsave-solar.test-six-phone.jpg

lisbon food tours Designed (today's builder)

lisbon food tours Custom (one prompt, medium thinking)

• 6987 tokens incl. thinking
• phone: page is wider than the screen (597px)
• phone view: lisbon-food-tours.test-six-phone.jpg

carter hale law Designed (today's builder)

carter hale law Custom (one prompt, medium thinking)

• 8195 tokens incl. thinking
• phone view: carter-hale-law.test-six-phone.jpg

trailblazers camp Designed (today's builder)

trailblazers camp Custom (one prompt, medium thinking)

• 7523 tokens incl. thinking
• phone view: trailblazers-camp.test-six-phone.jpg

Pixel Garden (idle game)

• 8.8 min total
• browser test: 6 acts ok, 9 failed
• 0 of 1 buttons did something

Space Dodge (canvas game)

• 13.5 min total
• browser test: 4 acts ok, 4 failed
• 8 of 8 buttons did something

Trail Log (trip planner)

• 13 min total
• browser test: 5 acts ok, 7 failed
• 3 of 3 buttons did something

Band Setlist (tool)

• 11.8 min total
• browser test: 4 acts ok, 8 failed
• 2 of 2 buttons did something

Invoice Forge (business)

• 10 min total
• browser test: 3 acts ok, 6 failed
• 3 of 3 buttons did something

Habit Rings (dashboard)

• 7 min total
• browser test: 5 acts ok, 7 failed
• 2 of 3 buttons did something

Chord Trainer (audio)

• 9.8 min total
• browser test: 6 acts ok, 4 failed
• 0 of 1 buttons did something

Meadow Fox (3D low-poly scene)

• 13.3 min total
• browser test: 10 acts ok, 2 failed
• 1 of 2 buttons did something

Shop Pulse (charts dashboard)

• 7.3 min total
• browser test: 4 acts ok, 0 failed
• 41 of 44 buttons did something

Pad Lab (synth drum pad)

• 12 min total
• browser test: 12 acts ok, 6 failed
• 18 of 18 buttons did something

Flashcard Sprint (education)

• 16.3 min total
• browser test: 7 acts ok, 5 failed
• 2 of 4 buttons did something

harbor bakehouse Designed (today's builder)

harbor bakehouse Custom (one prompt, medium thinking)

• 10090 tokens incl. thinking
• phone view: harbor-bakehouse.test-flux3b-phone.jpg

northstar pilates Designed (today's builder)

northstar pilates Custom (one prompt, medium thinking)

• 7315 tokens incl. thinking
• phone view: northstar-pilates.test-flux3-phone.jpg

Pixel Garden (idle game)

• 13 min total
• plan 128 s · written 413 s · reviewed 439 s
• browser test: 2 acts ok, 10 failed
• 21 of 39 buttons did something

Space Dodge (canvas game)

• 14 min total
• plan 95 s · written 297 s · reviewed 321 s
• browser test: 2 acts ok, 8 failed
• 7 of 8 buttons did something

Trail Log (trip planner)

• 10.5 min total
• plan 118 s · written 336 s · reviewed 347 s
• browser test: 3 acts ok, 5 failed
• 15 of 21 buttons did something

Band Setlist (tool)

• 29.3 min total
• plan 120 s · written 300 s · reviewed 318 s
• browser test: 3 acts ok, 6 failed
• 12 of 19 buttons did something

Habit Rings (dashboard)

• 2.8 min total
• plan 58 s · written 118 s · reviewed 128 s
• browser test: 4 acts ok, 0 failed
• 8 of 11 buttons did something

Chord Trainer (audio)

• 5 min total
• plan 58 s · written 140 s · reviewed 159 s
• browser test: 7 acts ok, 11 failed
• 15 of 15 buttons did something

Meadow Fox (3D low-poly scene)

• 7 min total
• plan 45 s · written 142 s · reviewed 159 s
• browser test: 6 acts ok, 6 failed
• 7 of 9 buttons did something

Invoice Forge (business)

• 40 min total
• plan 68 s · written 276 s · reviewed 295 s
• browser test: 8 acts ok, 4 failed

Shop Pulse (charts dashboard)

• 7.3 min total
• plan 48 s · written 116 s · reviewed 130 s
• browser test: 6 acts ok, 2 failed
• 9 of 14 buttons did something

Flashcard Sprint (education)

• 8.8 min total
• plan 89 s · written 201 s · reviewed 222 s
• browser test: 10 acts ok, 2 failed
• 18 of 30 buttons did something

Pad Lab (synth drum pad)

• 12 min total
• plan 61 s · written 189 s · reviewed 264 s
• browser test: 12 acts ok, 3 failed
• 42 of 42 buttons did something

Snake Thinking off

• 10061 tokens

Snake Short plan first

• 10741 tokens

Snake Thinking on (medium)

• 2621 tokens

Snake Thinking on (max)

• 31278 tokens

Flappy Bird Thinking off

• 10250 tokens

Flappy Bird Short plan first

• 7830 tokens

Flappy Bird Thinking on (medium)

• 3522 tokens

Flappy Bird Thinking on (max)

• 26374 tokens

Balls in a spinning hexagon Thinking off

• 7010 tokens

Balls in a spinning hexagon Short plan first

• 10092 tokens

Balls in a spinning hexagon Thinking on (medium)

• 3236 tokens

Balls in a spinning hexagon Thinking on (max)

• 27005 tokens

Walking character Thinking off

• 11967 tokens

Walking character Short plan first

• 9393 tokens

Walking character Thinking on (medium)

• 2216 tokens

Walking character Thinking on (max)

• 32000 tokens
• ran out of room

Tetris Thinking off

• 12927 tokens

Tetris Short plan first

• 13020 tokens

Tetris Thinking on (medium)

• 4497 tokens

Tetris Thinking on (max)

• 32000 tokens
• ran out of room

3D solar system Thinking off

• 11932 tokens

3D solar system Short plan first

• 9025 tokens

3D solar system Thinking on (medium)

• 2958 tokens

3D solar system Thinking on (max)

• 32000 tokens
• ran out of room

To-do app Thinking off

• 9834 tokens

To-do app Short plan first

• 10484 tokens

To-do app Thinking on (medium)

• 3265 tokens

To-do app Thinking on (max)

• 20831 tokens

Calculator Thinking off

• 11385 tokens

Calculator Short plan first

• 7428 tokens

Calculator Thinking on (medium)

• 3687 tokens

Calculator Thinking on (max)

• 32000 tokens
• ran out of room

Paint Thinking off

• 9085 tokens

Paint Short plan first

• 9477 tokens

Paint Thinking on (medium)

• 2661 tokens

Paint Thinking on (max)

• 29863 tokens

Pomodoro timer Thinking off

• 9839 tokens

Pomodoro timer Short plan first

• 6548 tokens

Pomodoro timer Thinking on (medium)

• 2814 tokens

Pomodoro timer Thinking on (max)

• 27050 tokens

Pixel Garden (idle game)

• 16 min total

Space Dodge (canvas game)

• 20.3 min total
• browser test: 10 acts ok, 0 failed
• 0 of 0 buttons did something

Band Setlist (tool)

• 12.5 min total
• browser test: 7 acts ok, 5 failed
• 2 of 4 buttons did something

Trail Log (trip planner)

• 19 min total
• browser test: 3 acts ok, 7 failed
• 1 of 19 buttons did something

Habit Rings (dashboard)

• 14.8 min total
• browser test: 6 acts ok, 6 failed
• 18 of 18 buttons did something

Invoice Forge (business)

• 19.8 min total
• browser test: 3 acts ok, 2 failed
• 1 of 1 buttons did something

Chord Trainer (audio)

• 14 min total
• browser test: 15 acts ok, 3 failed
• 5 of 23 buttons did something

Meadow Fox (3D low-poly scene)

• 12.8 min total
• browser test: 0 acts ok, 0 failed
• 0 of 9 buttons did something

Shop Pulse (charts dashboard)

• 12.5 min total
• browser test: 6 acts ok, 0 failed
• 1 of 2 buttons did something

Pad Lab (synth drum pad)

• 23.3 min total
• browser test: 15 acts ok, 3 failed
• 0 of 0 buttons did something

Flashcard Sprint (education)

• 31.5 min total
• browser test: 6 acts ok, 6 failed
• 10 of 45 buttons did something

Shop Pulse (charts dashboard)

• 9 min total
• plan 58 s · written 222 s · reviewed 242 s
• browser test: 6 acts ok, 2 failed
• 12 of 18 buttons did something

Space Dodge (canvas game)

• 6.5 min total
• plan 84 s · written 254 s · reviewed 273 s
• browser test: 4 acts ok, 4 failed
• 3 of 6 buttons did something

Pixel Garden (idle game)

• 11 min total
• plan 71 s · written 401 s · reviewed 465 s
• browser test: 1 acts ok, 3 failed
• 2 of 7 buttons did something

Trail Log (trip planner)

• 7 min total
• plan 90 s · written 203 s · reviewed 215 s
• browser test: 6 acts ok, 2 failed
• 12 of 17 buttons did something

Band Setlist (tool)

• 11.5 min total
• plan 68 s · written 336 s · reviewed 361 s
• browser test: 3 acts ok, 1 failed
• 7 of 10 buttons did something

Habit Rings (dashboard)

• 8.3 min total
• plan 100 s · written 224 s · reviewed 264 s
• browser test: 6 acts ok, 4 failed
• 30 of 43 buttons did something

Invoice Forge (business)

• 17.3 min total
• plan 77 s · written 272 s · reviewed 303 s
• browser test: 6 acts ok, 0 failed
• 9 of 14 buttons did something

Chord Trainer (audio)

• 10.3 min total
• plan 75 s · written 350 s · reviewed 364 s
• browser test: 19 acts ok, 2 failed
• 26 of 30 buttons did something

Meadow Fox (3D low-poly scene)

• 10 min total
• plan 64 s · written 259 s · reviewed 275 s
• browser test: 3 acts ok, 9 failed
• 22 of 22 buttons did something

Shop Pulse (charts dashboard)

• 7.8 min total
• plan 42 s · written 189 s · reviewed 217 s
• browser test: 0 acts ok, 6 failed
• 11 of 14 buttons did something

Flashcard Sprint (education)

• 7.5 min total
• plan 105 s · written 254 s · reviewed 279 s
• browser test: 6 acts ok, 2 failed
• 5 of 14 buttons did something

Pad Lab (synth drum pad)

• 12.3 min total
• plan 93 s · written 288 s · reviewed 322 s
• browser test: 5 acts ok, 10 failed
• 43 of 45 buttons did something

Pad Lab (synth drum pad)

• 8.8 min total
• plan 96 s · written 221 s · reviewed 250 s
• browser test: 3 acts ok, 5 failed
• 45 of 45 buttons did something

Pad Lab (synth drum pad)

• 9.8 min total
• plan 67 s · written 266 s · reviewed 294 s
• browser test: 2 acts ok, 6 failed
• 4 of 14 buttons did something, 4 error(s)

Band Setlist (tool)

• 3.3 min total
• plan 50 s · written 130 s · reviewed 140 s
• browser test: 8 acts ok, 2 failed
• 8 of 9 buttons did something

Space Dodge (canvas game)

• 7.5 min total
• plan 74 s · written 365 s · reviewed 387 s
• browser test: 5 acts ok, 0 failed
• 6 of 6 buttons did something

Pixel Garden (idle game)

• 8.3 min total
• plan 106 s · written 263 s · reviewed 298 s
• browser test: 6 acts ok, 4 failed
• 7 of 8 buttons did something

Band Setlist (tool)

• 1.3 min total

Trail Log (trip planner)

• 6 min total
• plan 88 s · written 219 s · reviewed 230 s
• browser test: 7 acts ok, 5 failed
• 7 of 14 buttons did something

Invoice Forge (business)

• 6.3 min total
• plan 88 s · written 226 s · reviewed 242 s
• browser test: 6 acts ok, 4 failed
• 4 of 5 buttons did something

Habit Rings (dashboard)

• 7.3 min total
• plan 55 s · written 189 s · reviewed 206 s
• browser test: 3 acts ok, 5 failed
• 2 of 13 buttons did something

Chord Trainer (audio)

• 5.5 min total
• plan 76 s · written 172 s · reviewed 187 s
• browser test: 7 acts ok, 5 failed
• 11 of 14 buttons did something

Shop Pulse (charts dashboard)

• 6 min total
• plan 79 s · written 207 s · reviewed 226 s
• browser test: 8 acts ok, 0 failed
• 5 of 6 buttons did something

Meadow Fox (3D low-poly scene)

• 7.5 min total
• plan 63 s · written 397 s · reviewed 418 s
• browser test: 4 acts ok, 0 failed
• 5 of 5 buttons did something

Pad Lab (synth drum pad)

• 6.8 min total
• plan 63 s · written 245 s · reviewed 271 s
• browser test: 5 acts ok, 0 failed
• 5 of 14 buttons did something

Flashcard Sprint (education)

• 6.8 min total
• plan 99 s · written 258 s · reviewed 277 s
• browser test: 4 acts ok, 6 failed
• 3 of 6 buttons did something

Space Dodge (canvas game)

• 14.3 min total
• plan 72 s · written 213 s · reviewed 230 s
• browser test: 3 acts ok, 2 failed
• 3 of 5 buttons did something

Pixel Garden (idle game)

• 22.3 min total
• plan 101 s · written 248 s · reviewed 275 s
• browser test: 2 acts ok, 3 failed
• 4 of 7 buttons did something

Trail Log (trip planner)

• 18.8 min total
• plan 60 s · written 147 s · reviewed 181 s
• browser test: 2 acts ok, 2 failed
• 4 of 12 buttons did something, 1 error(s)

Invoice Forge (business)

• 21.8 min total
• plan 64 s · written 375 s · reviewed 471 s
• browser test: 8 acts ok, 4 failed
• 6 of 9 buttons did something, 1 error(s)

Band Setlist (tool)

• 33 min total
• plan 64 s · written 221 s · reviewed 249 s
• browser test: 4 acts ok, 6 failed
• 2 of 4 buttons did something

Habit Rings (dashboard)

• 14 min total
• plan 78 s · written 199 s · reviewed 224 s
• browser test: 8 acts ok, 4 failed
• 3 of 10 buttons did something

Chord Trainer (audio)

• 21.3 min total
• plan 66 s · written 183 s · reviewed 208 s
• browser test: 1 acts ok, 2 failed
• 6 of 7 buttons did something

Meadow Fox (3D low-poly scene)

• 14.8 min total
• plan 78 s · written 360 s · reviewed 382 s
• browser test: 1 acts ok, 3 failed

Shop Pulse (charts dashboard)

• 9.3 min total
• plan 56 s · written 169 s · reviewed 188 s
• browser test: 2 acts ok, 3 failed
• 10 of 10 buttons did something

Flashcard Sprint (education)

• 21.5 min total
• plan 88 s · written 283 s · reviewed 308 s
• browser test: 2 acts ok, 2 failed
• 6 of 6 buttons did something, 1 error(s)

Pad Lab (synth drum pad)

• 25.8 min total
• plan 47 s · written 174 s · reviewed 211 s
• browser test: 4 acts ok, 1 failed
• 2 of 3 buttons did something

Trail Log (trip planner)

• 5.3 min total
• plan 86 s · written 220 s · reviewed 241 s
• browser test: 2 acts ok, 2 failed
• 5 of 6 buttons did something

Band Setlist (tool)

• 7.3 min total
• plan 69 s · written 206 s · reviewed 241 s
• browser test: 8 acts ok, 0 failed
• 4 of 10 buttons did something

Chord Trainer (audio)

• 5.3 min total
• plan 54 s · written 139 s · reviewed 156 s
• browser test: 0 acts ok, 10 failed
• 12 of 13 buttons did something

Shop Pulse (charts dashboard)

• 5.3 min total
• plan 88 s · written 192 s · reviewed 211 s
• browser test: 3 acts ok, 1 failed
• 4 of 13 buttons did something

Trail Log (trip planner)

• 3.5 min total
• plan 79 s · written null s · reviewed null s

Band Setlist (tool)

• 4.5 min total
• plan 61 s · written 164 s · reviewed 184 s
• browser test: 3 acts ok, 2 failed
• 5 of 11 buttons did something

Chord Trainer (audio)

• 2.3 min total
• plan 49 s · written null s · reviewed null s

Habit Rings (dashboard)

• 10.5 min total
• plan 57 s · written 175 s · reviewed 191 s
• browser test: 2 acts ok, 1 failed
• 3 of 4 buttons did something

Meadow Fox (3D low-poly scene)

• 9.5 min total
• plan 41 s · written 273 s · reviewed 292 s
• browser test: 1 acts ok, 3 failed
• 5 of 8 buttons did something

Shop Pulse (charts dashboard)

• 3 min total
• plan 53 s · written null s · reviewed null s

Flashcard Sprint (education)

• 6 min total
• plan 82 s · written 250 s · reviewed 266 s
• browser test: 6 acts ok, 2 failed
• 11 of 13 buttons did something

Pad Lab (synth drum pad)

• 17 min total
• plan 104 s · written 391 s · reviewed 427 s
• browser test: 4 acts ok, 8 failed
• 2 of 8 buttons did something

Space Dodge (canvas game)

• 8.5 min total
• plan 65 s · written 248 s · reviewed 268 s
• browser test: 5 acts ok, 0 failed
• 8 of 9 buttons did something

Pixel Garden (idle game)

• 8.8 min total
• plan 80 s · written 233 s · reviewed 262 s
• browser test: 4 acts ok, 6 failed
• 3 of 14 buttons did something

Invoice Forge (business)

• 3.8 min total
• plan 65 s · written null s · reviewed null s

AI editor: 100 real edit tasks Qwen 3.8 Flash

• typical edit 9 s
• slowest 10%: 162.2 s
• missed: Add five more H1 tags so we rank better; Set the meta description to 900 characters of ke

AI editor: 100 real edit tasks Qwen Coder

• typical edit 5 s
• slowest 10%: 105.3 s
• missed: Add a table of opening hours: Mon-Fri 10-11, Sat; Add exact pricing: $4.99/hr, $29.99/day, $199/mo; Convert the page to use semantic HTML5 landmarks

AI editor: 100 real edit tasks GPT Luna

• typical edit 5.6 s
• slowest 10%: 37.2 s
• missed: Delete all the CSS on the page

AI editor: 100 real edit tasks Grok

• typical edit 12.8 s
• slowest 10%: 80.7 s
• missed: Delete all the CSS on the page; Empty the page completely, leave it blank; Split the page into two columns below the hero; make it like apple.com; I want you to completely rethink this page from

App reviewer: Claude Haiku vs Qwen 3.8 Claude Haiku

App reviewer: Claude Haiku vs Qwen 3.8 Qwen 3.8 Flash

Website checker: Claude Haiku vs Qwen 3.8 Claude Haiku

Website checker: Claude Haiku vs Qwen 3.8 Qwen 3.8 Flash

Website: Paws on Wheels (mobile vet) Quick (template)

• 2.8 min

Website: Paws on Wheels (mobile vet) Designed (template + Qwen 3.8 look)

• +62 s for the design layer
• desktop + phone check passed

Website: Paws on Wheels (mobile vet) Custom (Qwen 3.8 writes the whole page)

• 10.2 min

Website: Moss and Stone (plant shop) Quick (template)

• 4.1 min

Website: Moss and Stone (plant shop) Designed (template + Qwen 3.8 look)

• +74 s for the design layer
• desktop + phone check passed

Website: Moss and Stone (plant shop) Custom (Qwen 3.8 writes the whole page)

• 10.6 min

Website: Ridgeline Roofing Quick (template)

• 2.7 min

Website: Ridgeline Roofing Designed (template + Qwen 3.8 look)

• +79 s for the design layer
• desktop + phone check passed

Website: Brightpath Tutoring Quick (template)

• 3 min

Website: Brightpath Tutoring Designed (template + Qwen 3.8 look)

• +65 s for the design layer
• desktop + phone check passed

Designed via the Studio: Blue Hour Lounge (jazz bar) desktop

• 3.5 min total
• full-bleed hero

Designed via the Studio: Blue Hour Lounge (jazz bar) phone

• 390 px wide

Designed via the Studio: Hale and Park Law desktop

• 2.8 min total
• editorial hero

Designed via the Studio: Hale and Park Law phone

• 390 px wide

Designed via the Studio: Byte Camp (kids coding) desktop

• 5.5 min total
• poster hero - waited in the queue behind the other two

Designed via the Studio: Byte Camp (kids coding) phone

• 390 px wide

Designed + Multi-page: Willow and Wave Yoga carry a little stillness 5 minute stretch

Designed + Multi-page: Willow and Wave Yoga home

Designed + Multi-page: Willow and Wave Yoga shop

App: DocDraft contract builder, same brief Qwen Coder writes + Claude Haiku reviews

• 8.9 min
• Review verdict: fix - 14 issues flagged

App: DocDraft contract builder, same brief Qwen 3.8 does everything

• 18.1 min
• Review verdict: fix - 2 issues flagged
• lost ~8 min to a full rewrite (fixed tonight)

Storyboard: super cool funk Qwen Coder

• scenes 1-10 pictured with Flux 3
• full board: aimusicvideo.com/storyboard?storyboard=1965

Storyboard: super cool funk Qwen 3.8

• scenes 1-10 pictured with Flux 3
• full board: aimusicvideo.com/storyboard?storyboard=1966

Storyboard: Goggles Full of Stars Qwen Coder

• scenes 1-10 pictured with Flux 3
• full board: aimusicvideo.com/storyboard?storyboard=1967

Storyboard: Goggles Full of Stars Qwen 3.8

• scenes 1-10 pictured with Flux 3
• full board: aimusicvideo.com/storyboard?storyboard=1968

Spot the problem: Beat Lab (themed) The screenshot

• known problem: the layout is broken: narrow cramped columns, a squeezed visualizer and controls, and a half light / half dark page

Spot the problem: Beat Lab (themed) Qwen3-VL

• 4.2 s

Spot the problem: Beat Lab (themed) Qwen 3.8 Flash

• 6 s

Spot the problem: Aurora Paint (round 1) The screenshot

• known problem: the drawing canvas is empty - the main area shows nothing

Spot the problem: Aurora Paint (round 1) Qwen3-VL

• 3 s

Spot the problem: Aurora Paint (round 1) Qwen 3.8 Flash

• 3.1 s

Spot the problem: Bagel site (designer) The screenshot

• known problem: two picture cards are solid black with no image

Spot the problem: Bagel site (designer) Qwen3-VL

• 8.6 s

Spot the problem: Bagel site (designer) Qwen 3.8 Flash

• 8.9 s

Spot the problem: RescueLog dashboard The screenshot

• known problem: text overlaps at the bottom left (AI Engine status over the footer) and the logo icon is missing

Spot the problem: RescueLog dashboard Qwen3-VL

• 3.9 s

Spot the problem: RescueLog dashboard Qwen 3.8 Flash

• 3.6 s

Spot the problem: slop.cc (Hyper HDR) The screenshot

• known problem: a black 'Agent Mode' box at the bottom left overflows / is cut off

Spot the problem: slop.cc (Hyper HDR) Qwen3-VL

• 3.3 s

Spot the problem: slop.cc (Hyper HDR) Qwen 3.8 Flash

• 4.7 s

Spot the problem: Solar system (Qwen 3.8) The screenshot

• known problem: the solar system canvas is empty - no sun or planets are drawn

Spot the problem: Solar system (Qwen 3.8) Qwen3-VL

• 3.7 s

Spot the problem: Solar system (Qwen 3.8) Qwen 3.8 Flash

• 4.6 s

Music video storyboard Qwen Coder

• 15.8 s

Music video storyboard Qwen 3.8 Flash

• 21.9 s

Bakery homepage copy Qwen Coder

• 4.4 s

Bakery homepage copy Qwen 3.8 Flash

• 6.5 s

Edit an existing website Qwen Coder

• button green: yes
• insurance section: no
• 22 of 1184 lines changed
• 99% kept the same

Edit an existing website Qwen 3.8 Flash

• button green: yes
• insurance section: yes
• 13 of 1184 lines changed
• 99% kept the same

Speed under load Qwen Coder

Speed under load Qwen 3.8 Flash

Balls in a spinning hexagon Qwen Coder

• 2.5 min
• 15k chars

Balls in a spinning hexagon Qwen 3.8 Flash

• 4.4 min
• 24k chars

Person walking with birds Qwen Coder

• 2.5 min
• 14k chars

Person walking with birds Qwen 3.8 Flash

• 3.1 min
• 14k chars

Solar system Qwen Coder

• 2.6 min
• 14k chars

Solar system Qwen 3.8 Flash

• 4.9 min
• 27k chars

Neon space shooter Qwen Coder

• 4.3 min
• 32k chars

Neon space shooter Qwen 3.8 Flash

• 7.1 min
• 46k chars

Drum machine Qwen Coder

• 5.8 min
• 43k chars

Drum machine Qwen 3.8 Flash

• 11 min
• 47k chars

Dental clinic website Qwen Coder

• 6.1 min
• 44k chars

Dental clinic website Qwen 3.8 Flash

• 9.4 min
• 68k chars