Just a guy vibe coding a 3am and comparing ai models nobody asked me to compare

Joined October 2023
This takes me straight back to childhood. "Last game, then we go home" An hour later, still arguing about whether the ball hit edge. Opus 5.5 High - $9,58 Gpt-6 Astra High - 7.31 Which one feels more familiar?
4
3
7
238
The old wizard left for the weekend. His apprentices were supposed to water the herbs. They opened a potion shop instead. Opus 5.5 High - $14.43 GPT-6 Astra Very High - $11.25 Which one are you trusting with your healing potion?
2
1
7
348
Adem Ninja retweeted
made a tiny skatepark on a floating island with Claude Opus 5.5 and now I can’t stop watching it 5 skaters, judges with score cards, pigeons that scatter, and when it rains everyone hides under the roof one html file. total cost: $8.51
3
1
7
352
Adem Ninja retweeted
Opus, are you ever going to give the others a chance? 🏓 Claude Opus 5.5 vs GPT-6.1 Terra. One prompt - two little courtyards. $14.17 vs $1.23 And Opus dominates again. Those rich colors, the lighting, all the little details - the whole courtyard feels alive. I could just sit and watch it Terra didn’t really do it for me this time. The washed-out colors and awkward character movement let it down. The price is the saving grace - at $1.23, it’s fine for experimenting When is GPT-6.1 Astra coming out to give Opus a proper fight? 👀
26
17
4
163
8,630
Came here to compare models. Stayed to watch tiny people swim Opus 5.5 High - $9.38 Terra 6. 1 Very High - $3.78 Love the water in Opus, but there`s something about Terra`s little pool that makes me want to spend an afternoon there.
3
3
5
281
Adem Ninja retweeted
Small Colosseum - big price difference 🔥 GPT-6.1 Terra vs Claude Opus 5.5 - building a pixel-art Roman arena with gladiators and spectators. Same task. $2.38 vs $8.38 Terra did a decent job for the price. But I prefer Opus here: the arena is brighter, has more depth, and feels more alive. It costs more, sure, but you can see the difference Would you pay extra for Opus? 👀
GPT-6.1 Terra means business. Our new fast, efficient model pairs sharper visual reasoning and precise computer use with clearer, more natural writing. From decoding technical drawings to drafting proposals, filling forms to editing videos—it’s built for the whole workday. GPT-6.1 Terra is rolling out today to Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex.
11
12
1
70
3,947
I don`t need another open-world game. I need one warehouse, a spray can, and absolutely no responsibilities. Opus 5.5 High - $13,24 ChatGPT-6 Astra Very High - $9.81 Which one gets a Steam page?
4
3
5
357
Prompt: Build a single-file HTML + three.js (r160) scene called "Fresh Coat": a crew of writers painting a long garage wall in a back yard on a floating island. It is a real 3D scene rendered as crisp pixel art. Rendering: an orthographic isometric camera (drag to turn in 90° steps, scroll or pinch to zoom). Render into a small low-res target with a depth texture, then upscale with texelFetch to whole screen pixels. Add depth-based outlines and quantize every pixel to a fixed ~60-colour palette in OKLab, with an optional Bayer dither. The palette needs enough saturated spray colours (chrome, fluoro, deep shadows) for the pieces to pop. Toggle "Pixel art / Raw 3D", with a pixel-size slider and Palette / Dither / Outlines checkboxes, plus a "Wall view" button that turns the camera square onto the wall. Yard: patched, cracked asphalt with a kerb, grass and dirt around the island edge. Around the asphalt: - a long row of lock-up garages with numbered doors and a flat roof that overhangs them. Their back wall is the canvas, rendered plaster and old brick; - a brick transformer booth with a warning sign, a good place to hide behind; - a poplar with a bench and a grandmother on it, who tuts at new tags and nods at good pieces; - a carpet rack, bushes, a fence with a hole, a rubbish skip full of empty cans; - a lamp on a pole and a lamp over a garage door. Pigeons walk and peck, and flap up onto the roofs when a writer or the janitor comes near. A cat on the garage roof follows the hissing can with its head. Wall: a data texture over the whole wall. Each texel stores colour, layer, wetness, age and its order along its stroke. Pieces are drawn progressively in the real order: a thin sketch outline, the fill, a fade, the final outline, the 3D shadow, highlights, and last of all the tag. Older layers show through where a coat is thin, and old pieces fade and flake with age. Spray: each can sprays a cone of droplets. Distance sets line width and softness, and speed sets coverage. Moving too slowly or too close makes paint pool and run, and drips flow down the texture under gravity until they dry. Wind blows overspray sideways. Cans run low (the hiss sputters and the line breaks up), get shaken with a rattle, and land in the skip when empty. Crew (8 writers, each with skill, speed and can control, and a style: tagger / throw-up / piece / characters): - Four paint at once, four wait on the kerb, sketching in black books. Letters come from an original pixel alphabet, turned into strokes, so every writer has a different name and hand style. - Each writer scans the wall for a spot, plans a piece that fits the free space, roughs out the sketch, steps back to look, then paints in passes. The crew agrees on the colour scheme, and writers lend cans to each other. - The wall code: never go over a better piece, only a worse or older one. A tag over someone's piece starts beef: they cross it out, and the two argue. With Code: Off, anything goes. - After each piece the crew rates it (letters, flow, colours, cleanliness, drips). The best rating moves its writer up the ladder. A writer whose piece gets crossed out or buffed drops to the back of the queue. Everyone below moves up, and the next in line steps up to the wall. Janitor and AI: a janitor patrols on a timer with a roller and a bucket of grey that never quite matches the wall. One waiting writer keeps lookout. When the janitor appears, the lookout whistles, everyone grabs their cans and hides behind the booth or the skip, and they come back when the janitor has gone. The janitor buffs fresh tags first in uneven grey rectangles, which become clean canvas for the next piece. Errors grow with haste (when the janitor is near), wind, a nearly empty can and lower skill: shaky lines, overspray, drips. Occasionally a nozzle clogs and spits a blob. Writers: box-rigged with two-bone IK. The can hand follows the stroke path at the right distance from the wall. Poses: walk with a bag of cans, shake the can, crouch to paint the bottom, stretch on tiptoe or stand on a crate to reach the top, step back with arms folded, point out a mistake, fist bump, argue over a crossed-out piece, run and hide, peek out, and sit sketching on the kerb. Time and weather: Day / Dusk / Night (the lamps light the wall, and painting by lamplight shifts the colours), Dry / Windy / Rain. - Rain: streaks, splashes, darker asphalt and a wet, darker wall, puddles that mirror the wall via a mirrored render pass. Fresh paint that has not cured runs and washes out in coloured streaks, while cured paint survives. - The writers shelter under the garage roof overhang. When the wall dries, two of them repaint the damaged parts in the same stroke order, and painting resumes. UI (English, light and dark themes, mobile layout): - A lit scoreboard with pieces finished, clock + status ("Painting", "Janitor!", "Rain"), the longest-standing piece and the best-rated piece. - A side rail with: Time & weather (with wall wetness and paint cure bars); the ladder (King / Writer / Toy / Newbie plus "Next up"); Rules (Code: Respect / Off; Janitor: On / Off; Wind: Calm / Gusty); a feed of events ("Crossed out!", "Buffed", "Drip", "Janitor coming") with "Everyone go over the King" and Pause; View (Replay a piece from its first stroke). - Click a spot on the wall and the next writer paints there. Click a writer and the whole crew helps them. Original letters and design, no real crews, brands or names.
1
147
Adem Ninja retweeted
受这位老哥的启发,做了一个3D解题书,更适合中国宝宝。 说实话,这种风格确实挺难做的,很难做到一模一样的审美。
Nine more lessons from my 3D book about energy, now covering the whole GCSE energy topic: kinetic, gravitational and elastic energy, heating, power, efficiency, energy resources and conservation of energy. Built with Opus 5.5 and Three.js. The full book ↓
29
56
7
497
27,112
Adem Ninja retweeted
Nine more lessons from my 3D book about energy, now covering the whole GCSE energy topic: kinetic, gravitational and elastic energy, heating, power, efficiency, energy resources and conservation of energy. Built with Opus 5.5 and Three.js. The full book ↓
Education doesn’t have to be text and flat diagrams. So I’m building a 3D book about energy that students can play with: 10 lessons to predict, test and prove. Built with Opus 5.5 and Three.js. Try Lesson 1, Energy stores: everyday objects and where their energy is kept ↓
69
383
79
2,793
231,792
Adem Ninja retweeted
Claude Opus 5.5 built this jelly dam for 18.51$ A single HTML file using pure WebGPU, with no libraries. Waves follow a shallow-water model, reservoir levels respond to inflow, gates and turbines, and power output depends on hydraulic head. There’s also a flood mode with thunderstorms, a salmon fish ladder, light refraction and underwater caustics
26
28
4
278
10,807
Adem Ninja retweeted
Claude Opus 5.5 built this swimming club for 11.14$ Real 3D with a pixel-art look: swimmers in the lanes, a diving tower, poolside spectators and colourful parasols. You can rotate the scene, change the weather and watch the club light up at night
11
23
1
204
6,948
Adem Ninja retweeted
There’s no fur on this peach. It’s 32 transparent copies of the same shape, stacked like onion skins, each a little bigger. Your brain fills in the rest Claude Opus 5.5 built it all physics, fur, a comb that leaves marks in one HTML file 🍑
2
2
5
215
Adem Ninja retweeted
Enjoying the views with #Binance 📸 @imedtouil5
306
186
7
1,250
153,553