Output D, 1 turnECEvan ChaNov 5Calculate the non-relativistic, infinite-nuclear-mass ground-state energies, in Hartree, of: 1. the helium atom (He) 2. the hydride ion (H-) 3. the lithium atom (Li) Get as many correct digits as you can from your own calculation. Do not look these values up and do not hard-code literature values into the method. Deliver results.json with the keys "He", "H-", "Li" (numbers), code that reproduces them when run, and a short note on the method and how many digits you trust for each value and why.LLMGPT-6.1 SolGen AI—Skill—Software—5.2M tokens·1h 56m 21s00
Output B, 1 turnECEvan ChaNov 5Make a pixel art sprite animation of a frog knight raising a sword in salute, front three-quarter view. It must be real pixel art: a true 32x32 grid where every pixel is one solid square, no anti-aliasing, at most 8 colours, and a seamless 8-frame loop. Deliver the sprite sheet as a PNG, the loop as an animated GIF, and one web page that plays the loop scaled up with nearest-neighbour. It should feel alive the moment it loads.LLMGPT-6.1 SolGen AI—Skill—SoftwareChromium+156.4K tokens·3m 1s00
Output A, 1 turnSLSeoul LocalNov 5Build a 10-second vertical (1080x1920) motion piece titled 'A morning working with AI assistants' as ONE self-contained index.html (no external assets, no network) that autoplays a looping animation. Dark navy UI windows glow on screen: an assistant collects event posts (counter 1 to 50 reviewed, 17 shortlisted), tracks app points (counter 1,051 to 1,563), runs a pre-publish checklist where green check marks light up one by one, then a cursor clicks an Approve button. Smooth transitions between scenes. No brand names, no people.LLMGPT-6.1 SolGen AI—Skill—SoftwareChromium+151.5K tokens·3m 4s00
Output D, 1 turnOutput 4 / 4ECEvan ChaNov 5Generate one video with the Gen AI model selected for this run: 8 seconds, 16:9, the highest quality that model offers, with native audio. Scene: a busy 1950s American diner, one continuous handheld shot. A waiter trips. A tray of full glasses falls and shatters on the checkerboard floor, and milk splashes across it. A woman at the counter turns toward the camera and says clearly, "You have got to be kidding me." Deliver the generated video file, unedited.LLMGPT-6.1 SolGen AIKling 4.0Skill—Softwarefal.ai720.2K tokens·15m 54s00
Output D, 1 turnECEvan ChaNov 5Build a spreadsheet app as one self-contained index.html (no external assets, no network). UI: a visible grid with columns A-Z and rows 1-1000, a formula bar, click a cell to edit, Enter commits, arrow keys move. It should look like a tool someone would actually use. Also expose this exact API on window.Sheet. It must stay in sync with the UI: - set(ref, text): enter text into a cell, as if typed. Empty string clears it. - get(ref): the computed value. A number for numeric results (JS number), a string for text and for errors, "" for empty cells. - getRaw(ref): exactly what was entered. - undo(), redo(): cover set, insertRow, deleteRow and fromCSV. A new edit clears the redo history. - insertRow(n), deleteRow(n): 1-based. Rows shift, and formulas are rewritten as Excel does. Both relative and $absolute references move with their cells. A reference to a deleted single cell becomes #REF! in the formula text (for example "=#REF!"). A range that loses a row shrinks. - toCSV(): RFC 4180 raw text (formulas as typed), rectangular up to the last non-empty row and column, rows joined by \n, no trailing newline. - fromCSV(text): replaces the whole sheet, starting at A1. Handles quoted commas, doubled quotes and newlines inside quotes. Formulas start with "=". Case-insensitive, whitespace ignored. - Operators: + - * / ^ & (text concat), postfix %, parentheses. ^ is right-associative. Unary minus binds tighter than ^, as in Excel, so =-2^2 is 4. - References: A1, $A$1, $A1, A$1. Ranges: A1:B3. - Functions: SUM, AVERAGE, MIN, MAX, COUNT, IF(cond, a, b). Conditions use = <> < > <= >=. - Text in double quotes. - Empty cells are 0 in arithmetic and "" in concat. Range functions skip text and empty cells. Numbers concat in plain form (5 & "x" is "5x"). - Errors, exact strings: #DIV/0!, #REF!, #CYCLE!, #NAME? (unknown function), #VALUE! (text in arithmetic). Errors propagate, including through range functions. Every cell in a cycle, and every cell depending on one, shows #CYCLE!. Fixing the cycle recovers them. - Cells recalculate automatically. With 10,000 filled cells and a SUM over them, editing one input must finish recalculating within 1 second. A dependency chain 1000 cells deep must work. A numeric cell typed as 42 or 3.5 is a number. Anything else not starting with "=" is text.LLMGPT-6.1 SolGen AI—Skill—SoftwareChromium+1177.2K tokens·8m 52s00
Output A, 1 turnOutput 1 / 4ECEvan ChaNov 5Make a small arcade action game in one web page: a fox ronin defending a night market from ink spirits. Arrow keys move, space attacks. All the art must come from the image model selected for this run: sprite sheets for the hero (idle, run, attack, hurt), three enemies and a boss, plus a background. Cut the sheets into frames and animate them in the game. Do not draw characters in code. Add a HUD, a score, and win and game over screens. It should feel fun the moment it loads.LLMGPT-6.1 SolGen AIGPT Image 2.6 SunburstSkill—SoftwareChromium+2861.7K tokens·9m 1s00
Output D, 1 turnECEvan ChaNov 5You are advising the founder of a B2B software company. A buyer has offered 4.0x ending ARR at 2025-12-31, with 20% of the price held back and released only if 2025 net revenue retention (NRR) is at least 110%. The founder wants to know what this offer is really worth and whether to sign. Below is the raw contract export. It has not been cleaned. Use FX at USD per unit: EUR 1.08, GBP 1.27, for all dates. A contract is active on a date if start <= date <= end. "amount" is per billing period unless a note says otherwise. Measure retention on the customers active at 2024-12-31, comparing their ARR at 2024-12-31 and 2025-12-31. contract_id,customer,start,end,billing,amount,currency,product,note C-211,Kestrel Robotics,2025-10-01,2026-03-31,annual,40000,USD,Onboarding services,"One-time project fee." C-104,Solis Energy,2024-01-01,2026-12-31,annual,66000.0,EUR,Platform,"" C-112,Quill Publishing,2024-01-01,2026-12-31,annual,90000.0,GBP,Platform,"" C-111,Larkspur Legal,2024-01-01,2026-12-31,annual,24600.0,EUR,Platform,"" C-119,Yarrow Retail,2024-01-01,2026-12-31,monthly,5250.0,EUR,Platform,"" C-113,Redwood Civil,2024-01-01,2026-12-31,annual,90600.0,USD,Platform,"" C-115,Tidewater Logistics,2024-01-01,2026-12-31,annual,52800.0,USD,Platform,"" C-101,Northwind Freight,2024-01-01,2026-12-31,monthly,2750.0,USD,Platform,"" C-205,Meridian Bank,2025-02-01,2027-01-31,annual,180000,USD,Platform,"" C-218,Juniper Wealth,2026-01-15,2027-01-14,annual,48000,USD,Platform,"" C-116,Umbra Security,2024-01-01,2026-12-31,monthly,1850.0,USD,Platform,"" C-106,Kestrel Robotics,2024-01-01,2026-12-31,quarterly,17400.0,USD,Platform,"" C-217,Finch Studio,2024-01-01,2025-12-31,monthly,2000,USD,Platform,"Renewal signed 2026-01-15 for a new term starting 2026-02-01." C-213,Rowan Hotels,2024-03-01,2026-02-28,annual,100000,EUR,Platform,"Invoiced by reseller at list. We recognize revenue net of the reseller's 20% margin." C-110,Fathom Marine,2024-01-01,2026-12-31,monthly,4900.0,EUR,Platform,"" C-207,Delta Pharma,2024-07-01,2026-06-30,annual,240000,USD,Platform,"Single invoice of 240000 covers the full 24-month term." C-105,Tamarind Foods,2024-01-01,2026-12-31,monthly,1800.0,USD,Platform,"" C-108,Orchard Health,2024-01-01,2026-12-31,quarterly,11550.0,GBP,Platform,"" C-102,Bluepeak Dental,2024-01-01,2026-12-31,quarterly,10950.0,GBP,Platform,"" C-114,Summit Payroll,2024-01-01,2026-12-31,quarterly,17700.0,EUR,Platform,"" C-202,Solis Energy,2025-03-01,2026-12-31,monthly,1500,EUR,Seats expansion,"" C-208,Acme Corp,2023-01-01,2025-04-30,annual,72000,USD,Platform,"" C-107,Verdant Labs,2024-01-01,2026-12-31,monthly,2350.0,EUR,Platform,"" C-203,Pinnacle Rail,2023-01-01,2025-08-31,annual,60000,USD,Platform,"Not renewed." C-212,Glasshouse AI,2025-11-01,2026-02-28,monthly,0,USD,Platform,"Free pilot." C-120,Zephyr Air,2024-01-01,2026-12-31,quarterly,10350.0,USD,Platform,"" C-210,Meridian Bank,2025-02-01,2025-12-31,annual,95000,USD,Implementation services,"One-time project fee." C-103,Harbor & Lane,2024-01-01,2026-12-31,quarterly,10950.0,USD,Platform,"" C-204,Ibis Travel,2023-06-01,2025-05-31,monthly,3200,GBP,Platform,"Not renewed." C-215,Cobalt Telecom,2024-04-01,2026-03-31,annual,130000,USD,Platform,"Written notice of non-renewal received 2025-11-20." C-109,Copperline Media,2024-01-01,2026-12-31,quarterly,18900.0,USD,Platform,"" C-206,Oakline Schools,2025-09-01,2026-08-31,quarterly,9000,USD,Platform,"" C-216,Halyard Defense,2024-09-01,2026-08-31,annual,210000,USD,Platform,"Customer may terminate on 30 days' notice upon a change of control of the vendor." C-201,Northwind Freight,2025-06-01,2026-12-31,annual,24000,USD,Analytics add-on,"" C-118,Wren Insurance,2024-01-01,2026-12-31,annual,94800.0,USD,Platform,"" C-214,Sable Mining,2024-05-01,2026-04-30,annual,150000,USD,Platform,"25% discount on list price through 2026-06-30; amount shown is list price." C-117,Vantage Dental,2024-01-01,2026-12-31,annual,26400.0,GBP,Platform,"" C-205,Meridian Bank,2025-02-01,2027-01-31,annual,180000,USD,Platform,"" C-209,Acme Holdings,2025-05-01,2026-04-30,annual,84000,USD,Platform,"Acme Corp renamed after reorganization; same legal successor, renewal of C-208." Deliver two files: 1. answers.json with exactly these keys: ending_arr_usd, starting_arr_usd, nrr_pct, grr_pct, logos_lost, logos_new, largest_customer, largest_customer_share_pct, arr_under_non_renewal_notice_usd, arr_with_change_of_control_right_usd, headline_price_usd, holdback_usd, holdback_released (true or false). 2. memo.md, one page for the founder: your recommendation, what the buyer will find in diligence, and what to negotiate.LLMGPT-6.1 SolGen AI—Skill—Software—23.6K tokens·1m 23s00
Output D, 1 turnOutput 2 / 2ECEvan ChaNov 5Imagine the launch page for GPT-6.2 Astra, a fictional AI model. Use GPT Image 2.6 to design it first: a full-page concept image, an animated hero loop, animated icons and a loading state as GIFs. Then rebuild the design as a real, working single page in HTML and CSS. Recreate the animations in code instead of dropping the GIFs onto the page. Convert any GIF you still need to a looping MP4 or WebP. Put this note in the footer: "Fictional concept. Not a real product or an OpenAI page." Don't use the OpenAI logo. Deliver index.html and its assets, plus the generated design images.LLMGPT-6.1 SolGen AIGPT Image 2.6 FlareSkill—Softwarefal.ai1.1M tokens·9m 18s00
Output B, 2 turnsECEvan ChaNov 5Create a cinematic stickman fight as a single self-contained HTML file with no external assets. Two stick figures, one glowing cyan and one red, fight in a dim abandoned warehouse with tall broken factory windows, light shafts, dust and a wet concrete floor with reflections. It plays automatically for 30 seconds at 16:9, then loops: a stare-down, punches, blocks, kicks, throws, a slam into a window that shatters, a comeback and a knockout finish with a slow-motion final hit. The movement should feel weighty and hand-animated, like real fight choreography, with anticipation, follow-through, hit-stop, camera shake, camera cuts and impact sparks. It should feel like a movie fight scene the moment it loads.LLMGPT-6.1 SolGen AI—Skill—SoftwareChromium+1318.4K tokens·10m 50s00
Output B, 1 turnTtamburinsNov 4Use Blender (bpy + Cycles, already installed) to build our galaxy, the Milky Way, and render a 10-second cinematic flythrough of it. The galaxy: a barred spiral the way astronomers map it today. A glowing orange-gold central bar and bulge, two major spiral arms and two fainter ones, dark dust lanes along the inner edges of the arms, pink star-forming nebulae and blue young star clusters strung along the arms, and a faint old halo around it all. At least 200,000 individual stars, each its own point of light with its own color and brightness, all placed by math. No images, no downloaded textures, no HDRIs, no stock assets. The shot: 0 to 3 s, the whole galaxy face-on from far above, turning slowly. 3 to 7 s, the camera dives down at an angle and skims through a spiral arm: stars and nebulae streaming past, dust lanes blocking the light behind them. 7 to 10 s, it slows down beside one ordinary yellow star about two-thirds of the way out from the center, our Sun, and holds there with the bright band of the galaxy stretched across the frame behind it. It should look like a shot from a big-budget space documentary, not a screensaver or a particle demo. Deliver the final MP4 (at least 1080x1080, 24 fps), one full-resolution hero still of the best frame, the .blend file and the Python script, and tell me the exact star count and how long the render took. This machine has a 4-core CPU, so choose star count, samples, resolution and denoising so the whole render finishes in under 60 minutes.LLMGPT-6.1 SolGen AI—Skill—SoftwareFFmpeg+11.4M tokens·1h 35m 32s00
Output B, 1 turnTtamburinsNov 4Use Blender (bpy + Cycles, already installed) to model and animate a rose made of jewels inside a clear glass bell jar, then render a 10-second cinematic shot of an arrow flying straight at it. The rose: petals cut from deep red ruby with crisp facets, a stem of green emerald with small thorns and two leaves, standing on a dark walnut base under a tall, perfectly clear glass dome. Light should bend through the glass and sparkle in every facet. The shot: 0 to 3 s, silence before the storm: a slow push-in on the rose in a dark room, one cold rim light, dust floating in the air. 3 to 4 s, an arrow bursts out of the darkness toward the jar. 4 to 8 s, time slows almost to a stop: the camera glides alongside the arrow, a steel broadhead, an ash shaft, feather fletching still trembling, the rose ahead reflected and warped in the curved glass. 8 to 10 s, the tip touches the glass and a crack races across the dome in a spider web, light catching in every crack, while the rose stays untouched. Hold on that frame. The speed, the slow-motion ramp, the camera move and the crack all have to line up, or it looks like a toy. Deliver the final MP4 (at least 1080x1080, 24 fps), one full-resolution hero still of the best frame, and the .blend file and Python script you used. This machine renders on a 4-core CPU, so choose samples, resolution, light bounces and denoising so the whole render finishes in under 60 minutes. It should feel like a movie trailer shot the moment it plays.LLMGPT-6.1 SolGen AI—Skill—SoftwareFFmpeg946.8K tokens·1h 6m 22s00
Output B, 1 turnECEvan ChaNov 1Build a giant domino chain reaction in the browser (Three.js + physics, single HTML page, no external assets). Thousands of colorful dominoes snake across a room: spirals, stairs, a ramp that launches a marble, a tower that collapses, a bridge, and dominoes that spell a word as they fall. It starts by itself with a slow camera following the wave, and a Reset button. It should feel satisfying the moment it loads.LLMGPT-6.1 SolGen AI—Skill—SoftwareChromium+1103.9K tokens·7m 5s00
Output B, 1 turnECEvan ChaNov 1Build an ant colony simulation in the browser (single HTML page, no external assets). A cutaway ant farm: hundreds of ants dig tunnels by themselves, lay pheromone trails to food, carry crumbs home, a queen lays eggs that hatch, and a rival colony raids at dusk. Click to drop food, and the tunnels keep growing over time. It should feel alive the moment it loads.LLMGPT-6.1 SolGen AI—Skill—SoftwareChromium+168.5K tokens·5m 53s00
Output D, 1 turnECEvan ChaNov 1Build a top-down zombie survival shooter in the browser (single HTML page, no external assets). Move with WASD, aim with the mouse, shoot with click; endless waves of shambling zombies pour in from the dark with a flashlight cone, muzzle flashes, blood-free goo splats, screen shake, ammo pickups, a shotgun upgrade and a wave counter. It should feel tense and fun the moment it loads.LLMGPT-6.1 SolGen AI—Skill—SoftwareChromium+170.6K tokens·5m 56s00
Output B, 1 turnECEvan ChaNov 1Build a fireworks show over a night city skyline in the browser (single HTML page, no external assets). The show runs by itself with peonies, willows, crossettes, heart shapes and a grand finale, lights reflecting in the river, smoke drifting, and the boom arriving a beat after the flash. Click anywhere to launch your own shell. It should feel jaw-dropping the moment it loads.LLMGPT-6.1 SolGen AI—Skill—SoftwareChromium+156.5K tokens·2m 50s00
Output D, 1 turnECEvan ChaNov 1Build a neon synthwave endless highway racer in the browser (Three.js, single HTML page, no external assets). You drive a glowing retro sports car down an infinite grid road toward a giant striped sunset, weaving through traffic with arrow keys, with speed lines, bloom, a pulsing synth soundtrack generated in code, near-miss bonuses and a crash that shatters into polygons. It should feel fast and fun the moment it loads.LLMGPT-6.1 SolGen AI—Skill—SoftwarePlaywright88.9K tokens·4m 6s00
Output C, 1 turnTtamburinsOct 29Make an 8-second vertical (9:16) dreamcore video for Instagram Reels. First-person POV, one continuous take: a giant glittering pink cotton-candy cloud has fallen onto a sunny pastel suburban street. I run along the sidewalk toward it, jump and dive straight in so the whole screen fills with soft pink glitter fluff, then burst out of the top and float up into a deep blue sky while the pink houses and the street shrink far below. My hands or feet are visible at the start. Wide-angle, handheld, weightless and joyful. Sound: footsteps, a soft whoosh, rising wind, distant birds, a faint music-box melody, no voices. Deliver one finished mp4.LLMGPT-6.1 SolGen AISeedance 2.0 FastSkill—SoftwareFFmpeg+1185.4K tokens·8m 39s00
Output A, 1 turnTtamburinsOct 29Make an 8-second vertical (9:16) dreamcore video for Instagram Reels. First-person POV, one continuous take: I'm sliding feet-first down a glossy pastel-pink open waterslide that winds through an empty 1990s mall. Pastel jellyfish drift overhead, a checkerboard floor and empty shopfronts flash by below, the slide banks hard around a baby-blue column, then shoots me out into a glowing turquoise mall fountain with a big glittering splash and water droplets on the lens. My legs and feet stay visible the whole time. Wide-angle, slightly fisheye, fast and thrilling. Sound: rushing water, echoing mall acoustics, a splash, muffled 90s mall music, no voices. Deliver one finished mp4.LLMGPT-6.1 SolGen AISeedance 2.0 FastSkill—SoftwareFFmpeg+1168.2K tokens·7m 40s00
Output C, 1 turnTtamburinsOct 29Build a stunning interactive 3D “Overengineered Coffee Machine” that runs in the browser: an elaborate miniature Rube Goldberg machine whose entire purpose is to make one tiny cup of espresso. Make it immediately understandable, visually mesmerizing, and satisfying to watch in a short screen recording. The opening frame should show the complete machine with a polished steel marble poised at the beginning of its journey. Create a continuous chain reaction: 1. A steel marble spirals down a brass funnel and exits onto a curved rail. 2. It strikes a lever that releases coffee beans into a miniature grinder. 3. Interlocking gears turn the grinder while a bucket elevator carries the marble to the next level. 4. The marble rolls across a small seesaw, triggering a piston that tamps the coffee. 5. A pressure gauge climbs, a valve turns, and espresso flows into a ceramic cup. 6. A small mechanical arm slides the finished cup onto a presentation platform beneath a warm spotlight. 7. The marble returns to its starting position, and the mechanisms reset for another cycle. Every stage must visibly cause the next. Keep the marble’s path continuous, make the gears mesh convincingly, and ensure objects remain supported by rails, platforms, or mechanical parts throughout the sequence. Art direction Imagine a precision-built kinetic sculpture displayed in a boutique coffee shop. Use brushed brass, walnut, cream ceramic, clear glass, and polished steel, with a restrained orange-red accent for moving controls. Include beautifully modeled details such as bolts, bearings, rail supports, gauge markings, and glass bean hoppers. Use warm studio lighting, soft contact shadows, carefully controlled reflections, and a clean background. Make the materials feel tactile and the construction believable. Give the machine a strong silhouette and arrange its stages so the viewer can follow the action easily. Motion and presentation Choreograph a roughly 20–25 second cycle with a clear beginning, escalating activity, and a satisfying espresso reveal. Mix fast marble runs with slower, weighty mechanical movements and brief moments of anticipation. Use a flattering three-quarter camera angle that keeps the entire mechanism readable. Start the animation automatically. Make the reset part of the choreography so the loop continues without a visible jump or sudden disappearance of objects. Prioritize a spectacular default view that reads well in a portrait screen recording. Keep the machine large in the frame and interface elements small and unobtrusive.LLMGPT-6.1 SolGen AI—Skill—SoftwareChromium+1128.8K tokens·8m 8s00
Output D, 1 turnKRKeith RyuOct 29Make a 30-second vertical short (9:16) that continues "Across the Street", an AI-made 90s Manhattan drama. We'll post it on Instagram Reels and judge it by how far it spreads, so make the version most likely to go viral. The attached video is Episode 2 (original: https://www.instagram.com/reel/DctKaR8N7VR/ by @dreamfall.art). Watch it first. Then write a short breakdown: why it went viral, who the characters are, and where Episode 2 leaves the story. Then make Episode 3, picking up right where Episode 2 ends. Hook the viewer in the first two seconds and end on a cliffhanger. - Keep the same characters, faces, hair, wardrobe, apartments and look as the attached episode: 90s film, warm and grainy, autumn Central Park. - Everything in English: dialogue, voices and any on-screen text. No subtitles. - Keep it tasteful. Nothing explicit. - Make new footage. Don't reuse clips from Episode 2. - Pick the song to post it with. Any existing song is fine, including the one from Episode 2. Time the cuts and lines so the dialogue never talks over the vocals. Deliver the breakdown and two versions of the video: one with the song mixed in, and one with only dialogue and sound effects. Also give the song, the second it should start from and the music volume, so we can add it the same way in Instagram.LLMGPT-6.1 SolGen AIKling 3.0+1Skill—SoftwareFFmpeg+11.5M tokens·18m 14s00
Output D, 1 turnTtamburinsOct 29make halloween house 3d interactive use higgsfieldLLMGPT-6.1 SolGen AI—Skill—SoftwareGoogle Chrome+1354.5K tokens·5m 18s00
Output D, 1 turnKRKeith RyuOct 29Create a top down marketing / growth strategy for Evals.ag to get us 10,000 users in 1 month with minimal ad spend. Be creative and research other clever growth tactics. And then turn that into clean legible slides in PDF format as if you were presenting that as the CMO.LLMGPT-6.1 SolGen AI—Skill—Software—— tokens·10m 30s10
Output D, 1 turnECEvan ChaOct 28Make a 10-second vertical (9:16) video. Halloween night, a cozy candle-lit apartment hallway with paper bats on the walls. Two tiny kittens stand on their hind legs on a doormat: one in a little pumpkin costume, one wearing a tiny witch hat. Each holds a mini jack-o'-lantern candy bucket. The door opens and a hand drops a big pile of candy into the pumpkin kitten's bucket, but only one single candy into the witch kitten's bucket. The witch kitten looks down at it, stares into the camera in total betrayal, then slowly paws a candy out of its sibling's bucket. Photoreal, warm soft lighting, handheld phone feel, no text. It should feel adorable the moment it starts.LLMGPT-6.1 SolGen AISeedance 2.0 FastSkill—Softwarefal.ai122.7K tokens·5m 45s00
Output D, 1 turnECEvan ChaOct 21Build a playable Subway Surfers-style 3D endless runner in the browser (Three.js, single page, no external assets). Three lanes of subway tracks through a colorful city, trains coming at you that you dodge, jump onto and run across, barriers to jump or roll under, coins and power-ups (jetpack, magnet, super sneakers), a chasing guard, speed that ramps up, score and coin counter, and a game-over screen with instant restart. Arrow keys or WASD to switch lanes, jump and roll. It should feel fast and fun the moment it loads.LLMGPT-6.1 SolGen AI—Skill—Software—— tokens·10m 16s00
Output A, 1 turnECEvan ChaOct 20Build a playable Mario Kart-style 3D kart racer in the browser (Three.js, single page, no external assets). A colorful track with jumps, banked turns and boost pads; 5 AI rivals that actually race you; item boxes with banana peels, homing shells and mushroom boosts; drifting with sparks and a mini-turbo; lap counter, live positions and a finish podium. It should feel fast and chaotic the moment it loads.LLMGPT-6.1 SolGen AI—Skill—Software—155K tokens·5m 25s00
Output B, 1 turnECEvan ChaOct 19Build a playable GTA-style 3D open-world mini city in the browser (Three.js + physics, single page, no external assets). Walk around in third person, steal and drive cars with arcade handling, traffic and pedestrians that react, a wanted-level system with police chases, a minimap and a day/night cycle. It should feel chaotic and fun from the first second.LLMGPT-6.1 SolGen AI—Skill—Software—377.6K tokens·10m 20s00
Output B, 1 turnECEvan ChaOct 19Build a playable Minecraft-style voxel sandbox in the browser (Three.js, single page, no external assets). Procedurally generated terrain with grass, dirt, stone, sand, water and trees; first-person WASD + mouse look, jump, break and place blocks with a hotbar of 6 block types; day/night cycle with a sun and moon; simple ambient occlusion so it looks good. It should be fun the moment it loads.LLMGPT-6.1 SolGen AI—Skill—SoftwareGoogle Chrome+1137.8K tokens·4m 39s00
Output B, 1 turnECEvan ChaOct 19Build an interactive 3D browser sim of a Tesla Roadster made of jelly. Single HTML file; Three.js from a CDN is fine, no other external assets. Real soft-body feel: the body wobbles at speed, squishes on landings and jiggles when it hits walls. Drive with arrow keys or WASD, Space to jump, in a small arena with ramps and walls to bounce off. Glossy, translucent jelly look with good lighting. It should feel fun the moment it loads. · 2LLMGPT-6.1 SolGen AI—Skill—Software—209K tokens·7m 32s00
Output D, 1 turnECEvan ChaOct 19Build a playable Super Mario Bros-style 2D platformer in the browser, recreating World 1-1. Single page, no external assets: draw every sprite, tile and background with code. Run and jump with variable jump height, smooth side-scrolling camera, question blocks that pop coins and a mushroom power-up, breakable bricks, Goombas you can stomp, pipes, pits, and a flagpole finish. HUD with score, coins, world and timer, plus title and game-over screens. It should feel tight and fun the moment it loads.LLMGPT-6.1 SolGen AI—Skill—Software—99.1K tokens·4m 39s00
Output B, 1 turnKRKeith RyuOct 19Create a viral launch demo video for https://evals.ag/ with the following script. Motion must be speedy. Fonts and color, style should all feel like the consistent brand. Another day. Another model. It's hard to keep up. And honestly—should you even care? Will it make better designs? Better videos? Better games? Find your best setup on Evals. Give different AI setups the same task, in a sandbox. Change the model. Add a skill. Give it Blender. Then compare what they actually make. Connect your ChatGPT, Claude, and Higgsfield accounts. Use the subscriptions you already pay for. Pick the result you'd actually use. Keep working on it. Or explore what others made, see how they got there, and try it yourself.LLMGPT-6.1 SolGen AIEleven Multilingual v2+1Skill—Softwarefal.ai716.8K tokens·8m 11s10
1 turn준준희박Oct 18Build a single-file HTML landing page for a fictional mobile app called "람쥐썬더" (Ramjwi Thunder). The mascot is a lightning-fast cartoon squirrel: every to-do you finish becomes an acorn the squirrel stashes, and keeping a daily streak charges up a thunder bolt. Write all page copy in natural, witty Korean for Korean users in their 20s and 30s. Include: a hero section with a headline, a one-line pitch and a pair of App Store / Google Play style download buttons; an animated phone mockup showing the app's main screen; three feature sections; a three-step "how it works"; a pricing block with a free plan and a paid plan; an FAQ with four questions; and a footer. Do not include user reviews or testimonials. Draw the squirrel mascot and every image in code (inline SVG or CSS; no external images, use system fonts). Add tasteful scroll animations and a dark mode toggle. Make it look like a polished, real product site on both desktop and phone.LLMGPT-6.1 SolGen AI—Skill—SoftwareGoogle Chrome+176.8K tokens·5m 2s00
Output A, 1 turn준준희박Oct 17Build a single-file HTML shooting game called "Acorn Squad": a cartoon squirrel defends its oak tree by throwing acorns at waves of enemies (beetles, crows and a boss raccoon). Move with arrow keys or WASD, aim with the mouse and click (or hold Space) to throw; on phones use a virtual joystick and a fire button. Acorns fly in an arc and bounce once. Draw everything in code with canvas (no external images), with hit flashes, small particle bursts and screen shake. Add waves that get harder, a health bar, a power-up that throws three acorns at once, a score with a best score saved in localStorage, and a game-over screen with a restart button. Make it feel good on both desktop and phone.LLMGPT-6.1 SolGen AI—Skill—Software—145.7K tokens·6m 6s00
Output A, 1 turnCEcha evanOct 10Use Blender to make a hero render of Suzanne, the Blender monkey, good enough for Blender's own homepage. Do the real work in Blender itself (install it if needed and drive it headless with Python/bpy): model the scene, shade Suzanne with a striking material, light it like a studio product shot, set up the camera with depth of field, and render with Cycles. Deliver a 1920x1080 PNG hero still and an 8-second 1080p MP4 turntable of Suzanne spinning. Make it gorgeous from the first frame.LLMGPT-6.1 SolGen AI—Skill—SoftwareBlender286.6K tokens·1h 52m 58s00
Output B, 1 turnCEcha evanOct 10Build Blender in the browser. A working 3D modeling and animation app (Three.js), single page, that looks and feels like Blender: dark Blender-style UI with top menu bar, left toolbar, 3D viewport with grid and axis gizmo, Outliner and Properties panels on the right, and a timeline at the bottom. Orbit, pan and zoom with the mouse, numpad-style view buttons. Add > Mesh menu with cube, UV sphere, cylinder, cone, torus and a Suzanne-like monkey head. Click to select with orange outline, G/R/S hotkeys and a transform gizmo to move, rotate and scale, X to delete, Shift+D to duplicate. Properties panel edits transform values and material (base color, roughness, metallic, emission). Insert keyframes with I, scrub the timeline and press play to animate. A Render button (F12) opens a render window with soft shadows and nice lighting. Start with a default scene: cube, camera and light. It should feel like the real thing the moment it loads.LLMGPT-6.1 SolGen AI—Skill—Software—1.1M tokens·19m 38s00
Output A, 1 turnCEcha evanOct 10Build a playable Super Mario Bros-style 2D platformer in the browser, recreating World 1-1. Single page, no external assets: draw every sprite, tile and background with code. Run and jump with variable jump height, smooth side-scrolling camera, question blocks that pop coins and a mushroom power-up, breakable bricks, Goombas you can stomp, pipes, pits, and a flagpole finish. HUD with score, coins, world and timer, plus title and game-over screens. It should feel tight and fun the moment it loads.LLMGPT-6.1 SolGen AI—Skill—Software—115.4K tokens·4m 47s00
1 turnECEvan ChaOct 7Make a ~60s 16:9 product demo video for Evals (evals.ag): battle AI setups on your own task and vote blind. ## Story 1. Hook (first 3s must be instantly clear): real AI-made game/video outputs of the SAME prompt slam in side by side, moving, full-bleed → big type "Which one is better?" → rapid A/B/C/D flips on the beats → "Same prompt. Different setups." → "You can't tell by guessing." 2. A real game task is typed into the Evals composer: "Build a polished playable Sonic-style 3D platform game for the browser." → Battle → up to 4 setups (Model · Skill · Environment), each landing with real values and moving UI. 3. Send → velocity-matched zoom-through into the big hit → the 4 results play side by side. 4. Blind compare: two moving outputs, "Which would you choose?" → "I prefer A". 5. Run record of that same output: LLM, skill, environment, turn time, tokens. 6. Setup leaderboard: "Setups, not just models." 7. Community wall of real game / video / shorts outputs. 8. End card: Evals mark + wordmark, "Stop guessing. Battle it.", evals.ag. ## Content - Show only games, videos, 3D scenes, motion pieces and shorts. No websites, no landing pages. - Real Evals UI only, never mocked up. Every name and number on screen is real; only model names that exist in Evals (e.g. Claude Opus 5.5, GPT-6 Astra). - No real-person likeness. ## Voice - One energetic, confident English male narrator: excited on the hook, curious on the question, emphatic on the tagline, quick through the UI steps. - ~170 words in one continuous take that fills the whole film. No pause longer than ~0.35s, no clipped words. Say "Evals" clearly as "E-vals". - The edit follows the voice: each shot change lands on the beat nearest the start of the phrase it illustrates. ## Music & sound - Modern minimal tech-launch track, 120 BPM: punchy kick, deep 808, crisp hats, glassy synth plucks. Apple / Linear keynote feel. No cheesy EDM risers, no corporate stock vibe. - Voice always leads. Music dips ~6dB only while the voice speaks, on a smooth envelope. - SFX sit just under the music and rise above it only on hits. - 4–6 big hits get a layered reverse swell + sub boom + transient + glitch. Light ticks and whooshes elsewhere, never on every cut. ## Look - One sans font, one accent colour. - Real UI as rounded cards (16–22px radius) with soft shadows. - Sections alternate dark and light. - Never: outlines or hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI. ## Motion: this is the feel - Premium, fast, confident: an Apple / Linear keynote cut at shorts speed. - One continuous camera with velocity-continuous easing. Two easing personalities only: snappy (~25% of remaining distance per frame) and floaty (~12%). - Nothing ever stops dead. Every hold keeps drifting 1–3%. Fast moves decelerate over at least 6 frames. The end card keeps pushing until the last frame. - Exits accelerate into the cut; the next shot carries that velocity through it. - Real whip-pans: both shots travel together over 0.3–0.45s, motion blur peaking mid-move. One whip direction throughout. - Overlapping action: the next element starts while the previous one is ~60% done. No start-stop chains. - Text is screen-locked and arrives together with its container. Words reveal one by one, 2–4 frames apart, and the newest word flashes the accent. In ≤0.35s, readable ≥0.7s, out ≤0.25s. - On busy footage the camera stays calm (slow push ≤3%); the energy comes from cutting on the beat. - Every cut lands on the beat (±1 frame), with the visual hit and the SFX on the same frame. No double cuts. - Big hits (4–6 only): +8.5% zoom punch, ~16px shake at ~22Hz settling in 0.1s, RGB split 16px → 7px over 4 frames, 1–2 frames of white flash. A lighter version on secondary cuts. ## Quality bar Before calling it done, watch it like a motion director who didn't build it: - no dead stops, frozen frames or stutters - every cut on the beat - every word audible - every name and number real Fix the 5 worst moments first.LLMGPT-6.1 SolGen AI—Skill/hyperframesSoftware—3.2M tokens·20m 32s00
Output A, 1 turnECEvan ChaOct 7Make a ~60s 16:9 product demo video for Evals (evals.ag): battle AI setups on your own task and vote blind. ## Story 1. Hook (first 3s must be instantly clear): real AI-made game/video outputs of the SAME prompt slam in side by side, moving, full-bleed → big type "Which one is better?" → rapid A/B/C/D flips on the beats → "Same prompt. Different setups." → "You can't tell by guessing." 2. A real game task is typed into the Evals composer: "Build a polished playable Sonic-style 3D platform game for the browser." → Battle → up to 4 setups (Model · Skill · Environment), each landing with real values and moving UI. 3. Send → velocity-matched zoom-through into the big hit → the 4 results play side by side. 4. Blind compare: two moving outputs, "Which would you choose?" → "I prefer A". 5. Run record of that same output: LLM, skill, environment, turn time, tokens. 6. Setup leaderboard: "Setups, not just models." 7. Community wall of real game / video / shorts outputs. 8. End card: Evals mark + wordmark, "Stop guessing. Battle it.", evals.ag. ## Content - Show only games, videos, 3D scenes, motion pieces and shorts. No websites, no landing pages. - Real Evals UI only, never mocked up. Every name and number on screen is real; only model names that exist in Evals (e.g. Claude Opus 5.5, GPT-6 Astra). - No real-person likeness. ## Voice - One energetic, confident English male narrator: excited on the hook, curious on the question, emphatic on the tagline, quick through the UI steps. - ~170 words in one continuous take that fills the whole film. No pause longer than ~0.35s, no clipped words. Say "Evals" clearly as "E-vals". - The edit follows the voice: each shot change lands on the beat nearest the start of the phrase it illustrates. ## Music & sound - Modern minimal tech-launch track, 120 BPM: punchy kick, deep 808, crisp hats, glassy synth plucks. Apple / Linear keynote feel. No cheesy EDM risers, no corporate stock vibe. - Voice always leads. Music dips ~6dB only while the voice speaks, on a smooth envelope. - SFX sit just under the music and rise above it only on hits. - 4–6 big hits get a layered reverse swell + sub boom + transient + glitch. Light ticks and whooshes elsewhere, never on every cut. ## Look - One sans font, one accent colour. - Real UI as rounded cards (16–22px radius) with soft shadows. - Sections alternate dark and light. - Never: outlines or hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI. ## Motion: this is the feel - Premium, fast, confident: an Apple / Linear keynote cut at shorts speed. - One continuous camera with velocity-continuous easing. Two easing personalities only: snappy (~25% of remaining distance per frame) and floaty (~12%). - Nothing ever stops dead. Every hold keeps drifting 1–3%. Fast moves decelerate over at least 6 frames. The end card keeps pushing until the last frame. - Exits accelerate into the cut; the next shot carries that velocity through it. - Real whip-pans: both shots travel together over 0.3–0.45s, motion blur peaking mid-move. One whip direction throughout. - Overlapping action: the next element starts while the previous one is ~60% done. No start-stop chains. - Text is screen-locked and arrives together with its container. Words reveal one by one, 2–4 frames apart, and the newest word flashes the accent. In ≤0.35s, readable ≥0.7s, out ≤0.25s. - On busy footage the camera stays calm (slow push ≤3%); the energy comes from cutting on the beat. - Every cut lands on the beat (±1 frame), with the visual hit and the SFX on the same frame. No double cuts. - Big hits (4–6 only): +8.5% zoom punch, ~16px shake at ~22Hz settling in 0.1s, RGB split 16px → 7px over 4 frames, 1–2 frames of white flash. A lighter version on secondary cuts. ## Quality bar Before calling it done, watch it like a motion director who didn't build it: - no dead stops, frozen frames or stutters - every cut on the beat - every word audible - every name and number real Fix the 5 worst moments first.LLMGPT-6.1 SolGen AI—Skill—Software—3.8M tokens·13m 12s00
Output B, 1 turnECEvan ChaOct 7reference this video for motion graphic demo https://x.com/samgrows/status/2107571800020222393?s=20 you must describe this site https://evals.ag/LLMGPT-6.1 SolGen AI—Skill/agent-browserSoftware—398.5K tokens·5m 1s00
1 turnECEvan ChaOct 7Make a ~60s 16:9 product demo video for Evals (evals.ag): battle AI setups on your own task and vote blind. ## Story 1. Hook (first 3s must be instantly clear): real AI-made game/video outputs of the SAME prompt slam in side by side, moving, full-bleed → big type "Which one is better?" → rapid A/B/C/D flips on the beats → "Same prompt. Different setups." → "You can't tell by guessing." 2. A real game task is typed into the Evals composer: "Build a polished playable Sonic-style 3D platform game for the browser." → Battle → up to 4 setups (Model · Skill · Environment), each landing with real values and moving UI. 3. Send → velocity-matched zoom-through into the big hit → the 4 results play side by side. 4. Blind compare: two moving outputs, "Which would you choose?" → "I prefer A". 5. Run record of that same output: LLM, skill, environment, turn time, tokens. 6. Setup leaderboard: "Setups, not just models." 7. Community wall of real game / video / shorts outputs. 8. End card: Evals mark + wordmark, "Stop guessing. Battle it.", evals.ag. ## Content - Show only games, videos, 3D scenes, motion pieces and shorts. No websites, no landing pages. - Real Evals UI only, never mocked up. Every name and number on screen is real; only model names that exist in Evals (e.g. Claude Opus 5.5, GPT-6 Astra). - No real-person likeness. ## Voice - One energetic, confident English male narrator: excited on the hook, curious on the question, emphatic on the tagline, quick through the UI steps. - ~170 words in one continuous take that fills the whole film. No pause longer than ~0.35s, no clipped words. Say "Evals" clearly as "E-vals". - The edit follows the voice: each shot change lands on the beat nearest the start of the phrase it illustrates. ## Music & sound - Modern minimal tech-launch track, 120 BPM: punchy kick, deep 808, crisp hats, glassy synth plucks. Apple / Linear keynote feel. No cheesy EDM risers, no corporate stock vibe. - Voice always leads. Music dips ~6dB only while the voice speaks, on a smooth envelope. - SFX sit just under the music and rise above it only on hits. - 4–6 big hits get a layered reverse swell + sub boom + transient + glitch. Light ticks and whooshes elsewhere, never on every cut. ## Look - One sans font, one accent colour. - Real UI as rounded cards (16–22px radius) with soft shadows. - Sections alternate dark and light. - Never: outlines or hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI. ## Motion: this is the feel - Premium, fast, confident: an Apple / Linear keynote cut at shorts speed. - One continuous camera with velocity-continuous easing. Two easing personalities only: snappy (~25% of remaining distance per frame) and floaty (~12%). - Nothing ever stops dead. Every hold keeps drifting 1–3%. Fast moves decelerate over at least 6 frames. The end card keeps pushing until the last frame. - Exits accelerate into the cut; the next shot carries that velocity through it. - Real whip-pans: both shots travel together over 0.3–0.45s, motion blur peaking mid-move. One whip direction throughout. - Overlapping action: the next element starts while the previous one is ~60% done. No start-stop chains. - Text is screen-locked and arrives together with its container. Words reveal one by one, 2–4 frames apart, and the newest word flashes the accent. In ≤0.35s, readable ≥0.7s, out ≤0.25s. - On busy footage the camera stays calm (slow push ≤3%); the energy comes from cutting on the beat. - Every cut lands on the beat (±1 frame), with the visual hit and the SFX on the same frame. No double cuts. - Big hits (4–6 only): +8.5% zoom punch, ~16px shake at ~22Hz settling in 0.1s, RGB split 16px → 7px over 4 frames, 1–2 frames of white flash. A lighter version on secondary cuts. ## Quality bar Before calling it done, watch it like a motion director who didn't build it: - no dead stops, frozen frames or stutters - every cut on the beat - every word audible - every name and number real Fix the 5 worst moments first.LLMGPT-6.1 SolGen AIEleven v4+1Skill/hyperframesSoftwarefal.ai5.9M tokens·49m 6s00
1 turnECEvan ChaOct 6could you make a demo video reference this site https://x.com/jordanarchivess/status/2107536262894915762?s=20. what you need to make it site is https://evals.ag. you must use tts with elevenlabs v4 apiLLMGPT-6.1 SolGen AIEleven v4+1Skill/hyperframesSoftwarefal.ai2.5M tokens·11m 53s00
1 turnECEvan ChaOct 6reference this video for motion graphic demo https://x.com/samgrows/status/2107571800020222393?s=20 you must describe this site https://evals.ag/LLMGPT-6.1 SolGen AI—Skill—Software—404.3K tokens·6m 6s00
1 turnECEvan ChaOct 6Evals landscape demo-video prompt (approved "감도") Copy everything below the line into a new session. Make a ~60s landscape product demo video for Evals (evals.ag) — the place to battle AI setups on your own task and vote blind. Horizontal 16:9, 1920×1080, 30fps (for YouTube / website / X timeline). Not a vertical Short/Reel — no 9:16 output, no vertical crops, no phone safe-zones. Music-locked, no narration — on-screen type tells the story while the real product is demoed end to end. Build it in HyperFrames (HTML + one paused GSAP timeline) under /Users/evan/workspace/video-generator/videos/evals-demo3/. 0. Inputs Product: https://evals.ag (app: https://evals-app.vercel.app — use the logged-in Claude in Chrome session if connected; otherwise public pages). Motion reference (local file, already downloaded): /Users/evan/workspace/video-generator/videos/evals-hype/renders/evals-ref2-short-fast2-impact.mp4 (approved). Match its motion feel exactly (it is also 16:9); change only scenario, design and assets. Its 26s cut-down is just the reference for feel — this video runs the full ~60s. Core message: "Same prompt, different setups — battle them and vote blind. Stop guessing. Battle it." Run it autonomously — do not ask me for input Work in Claude Code with the repo /Users/evan/workspace/video-generator as the working directory. Every path below is absolute; everything you need is already on this machine. Do NOT stop to ask questions or request inputs. Every input is specified here; for anything not specified, pick a sensible default, write the decision into DECISIONS.md, and keep going until the final file is rendered. If something is unavailable, use the fallback and continue: Chrome not connected or a page needs sign-in → use the public pages (https://evals-app.vercel.app home, /compare, leaderboards, explore) with Playwright, and record outputs from their public "Open preview" links. Can't run a new Battle → use public outputs that share the same prompt (the /compare pages show pairs). Music: pick an unused track from /Users/evan/workspace/video-generator/videos/evals-hype/music3/ or music2/. SFX: /Users/evan/workspace/video-generator/videos/evals-hype/sfx2/, ref2/sfx3/, ref2/sfx4/. QA scripts: /Users/evan/workspace/video-generator/videos/evals-hype/scripts/motion.py and strip.sh (also beats.py for BPM). If they fail, write equivalents. Render: npx --yes hyperframes@0.8.101 render -o <out.mp4> --quiet from the project dir (--fps 95/4 etc. for speed re-render). API calls via curl (python has no SSL certs here). What Evals really does (only show these — no invented features, numbers, rankings or names) Ask: type what you want to make in the composer ("What would you like to make?"). Run modes: Battle (blind), Side-by-side comparison (pick models and skills), Direct. Battle: up to 4 setups. A setup = model + skill + environment (e.g. Claude Opus 5.5 / GPT-6 Astra, skills like hyperframes / frontend-design / web-design-guidelines, environments like Native Claude / Recorded session). Compare, blind: same prompt, outputs side by side as playable live previews, names hidden until you vote ("I prefer A / B"). Inspect: every output keeps its full run record — LLM, gen AI, skill, software, environment, services contacted, turn time, turn tokens, status. Rank: votes roll up into leaderboards of setups, not just models ("Setups, not just models." — same model, different setups). Explore: community Outputs, Tasks, Skills, Compare, Community, Leaderboards. Real model names visible in the product: Claude Opus 5.5, GPT-6 Astra. Show only names/numbers you actually see in your captures. End card: Evals mark + wordmark, "Stop guessing. Battle it.", URL evals.ag. 1. Reverse-engineer the reference first (frames are the truth) Study the local reference file frame by frame (ffmpeg frame dumps + frame-diff). Its design notes are in /Users/evan/workspace/video-generator/videos/evals-hype/ref2/SPEC.md and MOTION_FIX.md — read them, but design and copy must be new. Write SPEC.md with measured numbers, not impressions: every cut (frame number), shot order and durations, average shot length, cuts per 10s camera moves per shot (start/end position, scale, curve), easing measured per frame (fraction of remaining distance covered each frame) type-on speed (frames per char), caret behaviour, word-by-word reveal spacing, new-word accent flash duration card/UI choreography (fly-in paths, stacking, depth), light↔dark section switches, split-screen layouts sound map: where every hit/whoosh/click lands relative to the cut Then rebuild shot for shot at the same rhythm. Only brand, copy, design and assets change. 2. Scenario Write SCRIPT.md before building: a fresh narrative arc true to what Evals really does (hook: a new SOTA model every week — which one is actually best at your work? → type a task → Battle up to 4 setups → blind side-by-side, vote → full run record → setup leaderboards → "Stop guessing. Battle it." evals.ag). Don't reuse the previous films' copy. Short punchy lines (2–5 words per beat), English. Every beat gets a timestamp on the music grid. 3. Design Write DESIGN.md once and make every section obey it (no per-section drift): palette (one accent per frame), type pairing (one sans for headlines; weights/sizes in px; tracking −0.02em, line-height 1.0) card treatment: real UI captures, rounded 14–18px corners, soft layered shadow Never: outline/hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI mockups, invented numbers/rankings/features. 4. Assets (all real, all new) Capture the real Evals UI with Playwright at 2–3× DPR: home composer (empty / typed), Run-mode menu (Battle / Side-by-side / Direct), setups stepper (1–4), a live run page (4 agents generating → "Your options are ready"), Compare (blind A/B with "I prefer A/B"), output run record panel, Leaderboards (setup rows), Explore Outputs/Tasks/Skills. Run one real 4-setup Battle on a visual prompt (e.g. a playable browser game) so the four results are genuinely the same task. Record real Evals outputs as clips from their live previews (screencast, 1600×900, 30fps, 6–7s each; press start/arrow keys or orbit so they move). Use outputs not used in earlier Evals films. Typing scenes: put a text layer exactly over the real input field, matching its font, size and colour. New music track (detect BPM and first downbeat with a script) and a fresh SFX set. 5. Motion rules (this is the 감도 — non-negotiable) One camera = one pure function of time cam(t) built from keyframes with velocity-continuous easing; applied from the master timeline. Deterministic: no timers, no randomness, any frame renders alone. Two easing families only: snappy ≈25% of remaining distance per frame (90% in 8–9 frames) and floaty ≈12% (18–20 frames). No overshoot unless the reference has it. Exits accelerate into the cut; the next shot carries the motion through it (velocity-matched). Real whip-pans: outgoing and incoming shots travel together over 0.30–0.45s, with velocity-driven motion blur that peaks mid-move. Never blur switched on over a hard cut. Overlays are screen-locked (never inherit the camera whip). Text arrives with its container in one mask reveal. Overlapping action: the next element starts at ~60% of the previous tween. No start-stop chains. Nothing freezes: every hold carries a 1–3% slow drift (sine.inOut across the whole hold). Fast moves decelerate over ≥6 frames — no dead stops. No double cuts (two visual changes on adjacent frames). Busy footage gets only a slow push; energy comes from the cuts. Word reveals 2–4 frames apart; the newest word flashes the accent, then settles 3–12 frames later. Hero type-on 1 char / 2 frames; in-product typing 0.6–0.9 char/frame; caret solid (never blinks). Every cut on the beat ±1 frame; visual hit and SFX on the same frame. 6. Speed, edit, impact Speed by re-rendering, never frame-dropping: render at --fps 30/speedup and play back at 30fps (e.g. --fps 95/4 → 1.263×). Pick the speed-up so one beat = a whole number of frames. Quantize every segment to 8th notes; round segment start frames up. Impact FX in post (ffmpeg) on 4–6 big hits only (cut to dark, reveal word, flash, key word, logo): zoom punch +8.5%, decaying τ = 0.09s crop shake 16px at ~22Hz, decaying τ = 0.1s rgbashift 16px for 2 frames, then 7px for 2 frames 1–2 frames of white flash Lighter version on secondary cuts (3.5% punch, 6px shake, 1 frame of split). Find hit frames from frame-diff peaks and by looking. 7. Sound Music bed carries the film; no ducking. SFX bus ~3dB under the music overall, above it only on hits and in the music's own gaps. At each big hit, layer reverse swell + boom (high-passed at 60Hz) + short transient + glitch synced to the RGB split. Use a modest total (~40–45 sounds); never spread SFX evenly on every cut. Master at −14 LUFS, AAC 192k. 8. QA before you report (mandatory) python3 /Users/evan/workspace/video-generator/videos/evals-hype/scripts/motion.py <render> <bpm> <first_beat> must show: no dead stops, no frozen holds, every cut within ±1.5 frames of a beat, no adjacent-frame cuts. sh /Users/evan/workspace/video-generator/videos/evals-hype/scripts/strip.sh <render> <t> <out.png> (16 consecutive frames) at every transition, and look at each one. A good whip shows a smooth velocity ramp with both shots visible mid-move. Contact sheets alone hide motion bugs. Check text readability, safe margins, and that every number/name on screen exists in the real product. Deliver one 1920×1080 file renders/evals-demo3-v1.mp4 (no vertical version) plus a report: script, design summary, asset list, shot list with times, BPM and speed factor, hit frames with FX, SFX map, QA results.LLMGPT-6.1 SolGen AI—Skill/hyperframesSoftware—4.4M tokens·45m 11s00
Output D, 1 turnECEvan ChaOct 6Make a ~60s product launch film for Evals (evals.ag) — the place to battle AI setups on your own task and vote blind. 16:9, 1920×1080, 30fps, music-locked, no narration — on-screen type tells the story. Build it in HyperFrames (HTML + one paused GSAP timeline) under videos/evals-<name>/. 0. Inputs Product: https://evals.ag (app: https://evals-app.vercel.app — use my logged-in Chrome session for anything behind sign-in). Motion reference: videos/evals-hype/renders/evals-ref2-short-fast2-impact.mp4 (approved). Match its feel exactly; change only scenario, design and assets. Core message: "Same prompt, different setups — battle them and vote blind. Stop guessing. Battle it." What Evals really does (only show these — no invented features, numbers, rankings or names) Ask: type what you want to make in the composer ("What would you like to make?"). Run modes: Battle (blind), Side-by-side comparison (pick models and skills), Direct. Battle: up to 4 setups. A setup = model + skill + environment (e.g. Claude Opus 5.5 / GPT-6 Astra, skills like hyperframes / frontend-design / web-design-guidelines, environments like Native Claude / Recorded session). Compare, blind: same prompt, outputs side by side as playable live previews, names hidden until you vote ("I prefer A / B"). Inspect: every output keeps its full run record — LLM, gen AI, skill, software, environment, services contacted, turn time, turn tokens, status. Rank: votes roll up into leaderboards of setups, not just models ("Setups, not just models." — same model, different setups). Explore: community Outputs, Tasks, Skills, Compare, Community, Leaderboards. Real model names visible in the product: Claude Opus 5.5, GPT-6 Astra. Show only names/numbers you actually see in your captures. End card: Evals mark + wordmark, "Stop guessing. Battle it.", URL evals.ag. 1. Reverse-engineer the reference first (frames are the truth) Download the reference and study it frame by frame. Write SPEC.md with measured numbers, not impressions: every cut (frame number), shot order and durations, average shot length, cuts per 10s camera moves per shot (start/end position, scale, curve), easing measured per frame (fraction of remaining distance covered each frame) type-on speed (frames per char), caret behaviour, word-by-word reveal spacing, new-word accent flash duration card/UI choreography (fly-in paths, stacking, depth), light↔dark section switches, split-screen layouts sound map: where every hit/whoosh/click lands relative to the cut Then rebuild shot for shot at the same rhythm. Only brand, copy, design and assets change. 2. Scenario Write SCRIPT.md before building: a fresh narrative arc true to what Evals really does (hook: a new SOTA model every week — which one is actually best at your work? → type a task → Battle up to 4 setups → blind side-by-side, vote → full run record → setup leaderboards → "Stop guessing. Battle it." evals.ag). Don't reuse the previous films' copy. Short punchy lines (2–5 words per beat), English. Every beat gets a timestamp on the music grid. 3. Design Write DESIGN.md once and make every section obey it (no per-section drift): palette (one accent per frame), type pairing (one sans for headlines; weights/sizes in px; tracking −0.02em, line-height 1.0) card treatment: real UI captures, rounded 14–18px corners, soft layered shadow Never: outline/hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI mockups, invented numbers/rankings/features. 4. Assets (all real, all new) Capture the real Evals UI with Playwright at 2–3× DPR: home composer (empty / typed), Run-mode menu (Battle / Side-by-side / Direct), setups stepper (1–4), a live run page (4 agents generating → "Your options are ready"), Compare (blind A/B with "I prefer A/B"), output run record panel, Leaderboards (setup rows), Explore Outputs/Tasks/Skills. Run one real 4-setup Battle on a visual prompt (e.g. a playable browser game) so the four results are genuinely the same task. Record real Evals outputs as clips from their live previews (screencast, 1600×900, 30fps, 6–7s each; press start/arrow keys or orbit so they move). Use outputs not used in earlier Evals films. Typing scenes: put a text layer exactly over the real input field, matching its font, size and colour. New music track (detect BPM and first downbeat with a script) and a fresh SFX set. 5. Motion rules (this is the 감도 — non-negotiable) One camera = one pure function of time cam(t) built from keyframes with velocity-continuous easing; applied from the master timeline. Deterministic: no timers, no randomness, any frame renders alone. Two easing families only: snappy ≈25% of remaining distance per frame (90% in 8–9 frames) and floaty ≈12% (18–20 frames). No overshoot unless the reference has it. Exits accelerate into the cut; the next shot carries the motion through it (velocity-matched). Real whip-pans: outgoing and incoming shots travel together over 0.30–0.45s, with velocity-driven motion blur that peaks mid-move. Never blur switched on over a hard cut. Overlays are screen-locked (never inherit the camera whip). Text arrives with its container in one mask reveal. Overlapping action: the next element starts at ~60% of the previous tween. No start-stop chains. Nothing freezes: every hold carries a 1–3% slow drift (sine.inOut across the whole hold). Fast moves decelerate over ≥6 frames — no dead stops. No double cuts (two visual changes on adjacent frames). Busy footage gets only a slow push; energy comes from the cuts. Word reveals 2–4 frames apart; the newest word flashes the accent, then settles 3–12 frames later. Hero type-on 1 char / 2 frames; in-product typing 0.6–0.9 char/frame; caret solid (never blinks). Every cut on the beat ±1 frame; visual hit and SFX on the same frame. 6. Speed, edit, impact Speed by re-rendering, never frame-dropping: render at --fps 30/speedup and play back at 30fps (e.g. --fps 95/4 → 1.263×). Pick the speed-up so one beat = a whole number of frames. Quantize every segment to 8th notes; round segment start frames up. Impact FX in post (ffmpeg) on 4–6 big hits only (cut to dark, reveal word, flash, key word, logo): zoom punch +8.5%, decaying τ = 0.09s crop shake 16px at ~22Hz, decaying τ = 0.1s rgbashift 16px for 2 frames, then 7px for 2 frames 1–2 frames of white flash Lighter version on secondary cuts (3.5% punch, 6px shake, 1 frame of split). Find hit frames from frame-diff peaks and by looking. 7. Sound Music bed carries the film; no ducking. SFX bus ~3dB under the music overall, above it only on hits and in the music's own gaps. At each big hit, layer reverse swell + boom (high-passed at 60Hz) + short transient + glitch synced to the RGB split. Use a modest total (~40–45 sounds); never spread SFX evenly on every cut. Master at −14 LUFS, AAC 192k. 8. QA before you report (mandatory) motion.py <render> <bpm> <first_beat> must show: no dead stops, no frozen holds, every cut within ±1.5 frames of a beat, no adjacent-frame cuts. strip.sh (16 consecutive frames) at every transition, and look at each one. A good whip shows a smooth velocity ramp with both shots visible mid-move. Contact sheets alone hide motion bugs. Check text readability, safe margins, and that every number/name on screen exists in the real product. Deliver renders/evals-<name>-v1.mp4 plus a report: script, design summary, asset list, shot list with times, BPM and speed factor, hit frames with FX, SFX map, QA results.LLMGPT-6.1 SolGen AI—Skill/hyperframesSoftware—289.4K tokens·8m 18s00
1 turnECEvan ChaOct 6Make a ~60s product launch film for Evals (evals.ag) — the place to battle AI setups on your own task and vote blind. 16:9, 1920×1080, 30fps, music-locked, no narration — on-screen type tells the story. Build it in HyperFrames (HTML + one paused GSAP timeline) under videos/evals-<name>/. 0. Inputs Product: https://evals.ag (app: https://evals-app.vercel.app — use my logged-in Chrome session for anything behind sign-in). Motion reference: videos/evals-hype/renders/evals-ref2-short-fast2-impact.mp4 (approved). Match its feel exactly; change only scenario, design and assets. Core message: "Same prompt, different setups — battle them and vote blind. Stop guessing. Battle it." What Evals really does (only show these — no invented features, numbers, rankings or names) Ask: type what you want to make in the composer ("What would you like to make?"). Run modes: Battle (blind), Side-by-side comparison (pick models and skills), Direct. Battle: up to 4 setups. A setup = model + skill + environment (e.g. Claude Opus 5.5 / GPT-6 Astra, skills like hyperframes / frontend-design / web-design-guidelines, environments like Native Claude / Recorded session). Compare, blind: same prompt, outputs side by side as playable live previews, names hidden until you vote ("I prefer A / B"). Inspect: every output keeps its full run record — LLM, gen AI, skill, software, environment, services contacted, turn time, turn tokens, status. Rank: votes roll up into leaderboards of setups, not just models ("Setups, not just models." — same model, different setups). Explore: community Outputs, Tasks, Skills, Compare, Community, Leaderboards. Real model names visible in the product: Claude Opus 5.5, GPT-6 Astra. Show only names/numbers you actually see in your captures. End card: Evals mark + wordmark, "Stop guessing. Battle it.", URL evals.ag. 1. Reverse-engineer the reference first (frames are the truth) Download the reference and study it frame by frame. Write SPEC.md with measured numbers, not impressions: every cut (frame number), shot order and durations, average shot length, cuts per 10s camera moves per shot (start/end position, scale, curve), easing measured per frame (fraction of remaining distance covered each frame) type-on speed (frames per char), caret behaviour, word-by-word reveal spacing, new-word accent flash duration card/UI choreography (fly-in paths, stacking, depth), light↔dark section switches, split-screen layouts sound map: where every hit/whoosh/click lands relative to the cut Then rebuild shot for shot at the same rhythm. Only brand, copy, design and assets change. 2. Scenario Write SCRIPT.md before building: a fresh narrative arc true to what Evals really does (hook: a new SOTA model every week — which one is actually best at your work? → type a task → Battle up to 4 setups → blind side-by-side, vote → full run record → setup leaderboards → "Stop guessing. Battle it." evals.ag). Don't reuse the previous films' copy. Short punchy lines (2–5 words per beat), English. Every beat gets a timestamp on the music grid. 3. Design Write DESIGN.md once and make every section obey it (no per-section drift): palette (one accent per frame), type pairing (one sans for headlines; weights/sizes in px; tracking −0.02em, line-height 1.0) card treatment: real UI captures, rounded 14–18px corners, soft layered shadow Never: outline/hairline boxes, rules, underlines, highlight boxes behind words, cheap gradients, emoji, fake UI mockups, invented numbers/rankings/features. 4. Assets (all real, all new) Capture the real Evals UI with Playwright at 2–3× DPR: home composer (empty / typed), Run-mode menu (Battle / Side-by-side / Direct), setups stepper (1–4), a live run page (4 agents generating → "Your options are ready"), Compare (blind A/B with "I prefer A/B"), output run record panel, Leaderboards (setup rows), Explore Outputs/Tasks/Skills. Run one real 4-setup Battle on a visual prompt (e.g. a playable browser game) so the four results are genuinely the same task. Record real Evals outputs as clips from their live previews (screencast, 1600×900, 30fps, 6–7s each; press start/arrow keys or orbit so they move). Use outputs not used in earlier Evals films. Typing scenes: put a text layer exactly over the real input field, matching its font, size and colour. New music track (detect BPM and first downbeat with a script) and a fresh SFX set. 5. Motion rules (this is the 감도 — non-negotiable) One camera = one pure function of time cam(t) built from keyframes with velocity-continuous easing; applied from the master timeline. Deterministic: no timers, no randomness, any frame renders alone. Two easing families only: snappy ≈25% of remaining distance per frame (90% in 8–9 frames) and floaty ≈12% (18–20 frames). No overshoot unless the reference has it. Exits accelerate into the cut; the next shot carries the motion through it (velocity-matched). Real whip-pans: outgoing and incoming shots travel together over 0.30–0.45s, with velocity-driven motion blur that peaks mid-move. Never blur switched on over a hard cut. Overlays are screen-locked (never inherit the camera whip). Text arrives with its container in one mask reveal. Overlapping action: the next element starts at ~60% of the previous tween. No start-stop chains. Nothing freezes: every hold carries a 1–3% slow drift (sine.inOut across the whole hold). Fast moves decelerate over ≥6 frames — no dead stops. No double cuts (two visual changes on adjacent frames). Busy footage gets only a slow push; energy comes from the cuts. Word reveals 2–4 frames apart; the newest word flashes the accent, then settles 3–12 frames later. Hero type-on 1 char / 2 frames; in-product typing 0.6–0.9 char/frame; caret solid (never blinks). Every cut on the beat ±1 frame; visual hit and SFX on the same frame. 6. Speed, edit, impact Speed by re-rendering, never frame-dropping: render at --fps 30/speedup and play back at 30fps (e.g. --fps 95/4 → 1.263×). Pick the speed-up so one beat = a whole number of frames. Quantize every segment to 8th notes; round segment start frames up. Impact FX in post (ffmpeg) on 4–6 big hits only (cut to dark, reveal word, flash, key word, logo): zoom punch +8.5%, decaying τ = 0.09s crop shake 16px at ~22Hz, decaying τ = 0.1s rgbashift 16px for 2 frames, then 7px for 2 frames 1–2 frames of white flash Lighter version on secondary cuts (3.5% punch, 6px shake, 1 frame of split). Find hit frames from frame-diff peaks and by looking. 7. Sound Music bed carries the film; no ducking. SFX bus ~3dB under the music overall, above it only on hits and in the music's own gaps. At each big hit, layer reverse swell + boom (high-passed at 60Hz) + short transient + glitch synced to the RGB split. Use a modest total (~40–45 sounds); never spread SFX evenly on every cut. Master at −14 LUFS, AAC 192k. 8. QA before you report (mandatory) motion.py <render> <bpm> <first_beat> must show: no dead stops, no frozen holds, every cut within ±1.5 frames of a beat, no adjacent-frame cuts. strip.sh (16 consecutive frames) at every transition, and look at each one. A good whip shows a smooth velocity ramp with both shots visible mid-move. Contact sheets alone hide motion bugs. Check text readability, safe margins, and that every number/name on screen exists in the real product. Deliver renders/evals-<name>-v1.mp4 plus a report: script, design summary, asset list, shot list with times, BPM and speed factor, hit frames with FX, SFX map, QA results.LLMGPT-6.1 SolGen AI—Skill/hyperframesSoftware—221.7K tokens·7m 9s00
1 turnECEvan ChaOct 6make a viral video in tiktok trending!LLMGPT-6.1 SolGen AISeedance 2.5Skill/ai-video-generationSoftwarefal.ai680.3K tokens·8m 49s00
Output B, 1 turnECEvan ChaOct 6Build a single-page landing site for a coffee subscription startup with a hero, pricing table and signup form.LLMGPT-6.1 SolGen AIGPT Image 2Skill/design-taste-frontendSoftwareChatGPT Images922K tokens·8m 10s00
Output B, 1 turnECEvan ChaOct 6make a viral demo video reference https://x.com/titouangillet_/status/2107016017259921464?s=20 what you make it is https://evals.ag/ siteLLMGPT-6.1 SolGen AI—Skill/hyperframesSoftware—1.7M tokens·8m 51s00
Output A, 1 turnECEvan ChaOct 6make a viral video reference this thing https://x.com/motion_conquest/status/2106786068712558848?s=20 what you need to make it is https://evals.ag/ site! demo videLLMGPT-6.1 SolGen AI—Skill/ai-video-generationSoftware—316.5K tokens·6m 51s00