Shoebill bird

@Psalteric Perfect design

The feature that kills the blank-chat-box problem is already designed. Fable 5 just showed what a few prompts can build. No design team. No weeks of iteration. Just Fable 5 and a handful of prompts — and out came the ChatGPT 6.5 — AI Assistant Suggestions Experience. Here's what Fable 5 designed 🔹 Expanded, smarter suggestions the moment the chat opens — no more staring at a blank box. 🔹 Options shaped around user interests, past chats, and personal context — not generic filler. 🔹 Built to help people start conversations faster and surface what actually matters to them. The wild part? It went from idea → prompts → finished concept in one sitting. That's the new speed of building: prompt it, ship it, sell it. Concept fully designed by Fable 5

@Psalteric That result is such a big deal

MAN SOLD TWO CINEMATIC 3D GLOBE MOCKUPS AND WALKED AWAY $3,000 RICHER — BUT ONLY AFTER HE PUT THEM SIDE BY SIDE FIRST. Before handing them to the client, Man lined up both AI/ML API screens and studied them like a buyer would. On the left, Sonnet 5 — $0.10, 87% cheaper: a dark Earth wrapped in glowing orange network trails. Lean and stealthy. On the right, Fable 5 — $0.77: a vivid blue Earth threaded with luminous cyan lines. Bright and premium. That side-by-side check was the move. Once Man saw them together, he knew exactly how to pitch each one — and the client bought both on the spot.

@Psalteric In one prompt one world game is insane

CLAUDE FABLE 5 JUST BUILT NO MAN'S SKY FROM A SINGLE PROMPT. Nobody wrote a line of code. No game engine was touched. There wasn't even a team. Just one sentence. Just one model. And out came "ECHO HORIZON" — a fully playable 3D space exploration game running in the browser. What's inside it? → A galaxy you can actually fly through, with star systems like Synach Prime and Riash Prime → Four hand-feeling, procedurally-named planets, each with its own biome and gravity: • a volcanic hellscape of lava and jagged cliffs (1.13g) • a green world alive with forests, oceans, and wildlife (0.63g) • an endless water planet (1.03g) • a frozen realm of snow, pines, mountains, and glittering planetary rings (0.77g) → Real spaceship flight, complete with a speed + altitude HUD → A pulse drive that hurls you between planets → Smooth atmospheric entry, landing, and takeoff → Step off the ship and explore on foot — sprint, jump, wander → Surface structures scattered across alien terrain → Day/night lighting that shifts as worlds turn Now hold that next to reality: Hello Games needed four years and fifteen people to ship No Man's Sky. Claude knocked this out somewhere between ordering lunch and finishing it. Curious what the prompt looked like? Roughly this: "Build a complete 3D space exploration game in the browser. Multiple star systems, procedurally generated planets — each with a unique biome (volcanic, lush, oceanic, glacial), its own gravity, and a generated name. Give me a flyable spaceship with a speed/altitude HUD, a pulse drive for interplanetary travel, atmospheric entry, planet landing, and on-foot exploration. Throw in terrain generation, trees, rocks, structures, day/night lighting, and planetary rings. Make the whole thing playable with mouse + keyboard." A $200B industry. Disrupted. By a text box.

@Psalteric O yep, wait for that article because also interested into design

@Psalteric I guess Anthropic will boost Fable 5 and we will win

Meet claude-fable-5-lite. 😄 The model got nerfed — weakened, taste stripped out. On its own it designs safe and dull: grey text, one blue button. But wire it up to Opus 4.8 and it's actually good: bold type, a real accent color, whitespace it commits to. Verdict: the "Fable 5 magic" was ~70% prompt, not weights. 🎨

@Psalteric 7 minute to get a full UI and UX screens app. That's wild

Fable 5 is officially back and fully usable again 🎉 — and last night, all it took was a movie. That's it. That's the whole "work session." A man put on a film, dropped a few small prompts into Fable 5, and the video below is the result: a complete, market-ready app that looks and feels like a team of designers spent months on it. Watch it build itself. The UX is chef's kiss. The UI is clean enough to launch tomorrow. And it's the man's own — ready to sell, ready to earn. Fable 5 didn't come back to help you prototype. It came back to help you ship. 💸

I've watched this clip four times and I still can't stop laughing. The guy is racing the Monaco Grand Prix from a kitchen stool. Six seconds later — he's leading the pack at 300 km/h. No, really. He's sitting on a wobbly stool in the corner of his kitchen, dish rack behind him, gripping a frying pan like it's a steering wheel and leaning into a corner that only exists in his head. AI turns him into a Formula 1 driver mid-apex. Same lean. Same white-knuckle grip. The frying pan becomes a carbon-fiber wheel, the kitchen becomes a cockpit screaming down the straight. Then he plants both feet, throws his weight sideways, and yanks the "wheel" hard. AI drops him into a rally car sliding through a dust storm in the desert. Same posture, down to the tension in his shoulders. The dish rack? Now it's a wall of gravel flying past the window. Third shot — he slouches back, one hand loose on the pan, the other on an invisible gearshift, chewing gum like he's got all night. AI turns him into a trucker rolling down an empty highway at 3 a.m., neon reflecting off the windshield. What's actually in his other hand is a TV remote. And here's the thing to understand: AI no longer needs a car, a track, or a stunt budget. It needs a pose. Badly acted, in socks, with a frying pan, a broom, or a laundry basket. Pose transfer is so precise that a kitchen stool becomes a race seat and a spatula becomes a gearshift. Framework: → RECORD — act the pose with whatever's within reach → REIMAGINE — ChatGPT rebuilds the world around it → ANIMATE — Kling puts it in motion Between a stunt driver on a contract and a guy with a frying pan, there's only one difference left: the guy doesn't have an agent. This week I'm breaking down a few more finds like this.

Tell me where the camera was hidden in this shot. You can't — because there was no camera. No crew. No stunt team. No CGI. Just one prompt typed into an AI. Watch closely: A lone samurai stands at the edge of a cliff, the wind dragging through his hakama. Mist rolls below. Mount Fuji burns red behind him. Then he moves. → The blade leaves the saya in a single breath — too fast to follow. → The camera doesn't cut away. It stays locked to the steel, riding the arc of the strike. → Petals freeze mid-air. Rain hangs in place. → Time slows to almost nothing. The world goes silent. The camera circles his face — calm, focused, unshaken. He exhales. His eyes lift to meet yours. Then — The blade snaps back into its sheath. Click. And everything rushes back to full speed at once. One continuous shot. No cuts. No transitions. No edits. This is the level of cinematic control AI video models have already reached — full choreography, real weight, real motion, real emotion, all from words on a screen. And this is the worst it will ever be from here. 📌 Save this post. Follow for AI prompts and workflows that look impossible — right up until you watch them happen.

𝗢𝗻𝗲 𝗽𝗿𝗼𝗺𝗽𝘁. 𝗡𝗼 𝗰𝗮𝗺𝗲𝗿𝗮. 𝗡𝗼 𝗱𝗿𝗼𝗻𝗲. 𝗡𝗼 𝗰𝗿𝗲𝘄. 𝗔 𝗳𝘂𝗹𝗹 𝗰𝗶𝗻𝗲𝗺𝗮𝘁𝗶𝗰 𝗳𝗹𝗶𝗴𝗵𝘁 𝘁𝗵𝗿𝗼𝘂𝗴𝗵 𝘁𝗵𝗲 𝗰𝗶𝘁𝘆 — 𝘀𝗲𝗲𝗻 𝘁𝗵𝗿𝗼𝘂𝗴𝗵 𝘁𝗵𝗲 𝗲𝘆𝗲𝘀 𝗼𝗳 𝗮 𝗽𝗶𝗴𝗲𝗼𝗻. 𝗦𝗲𝗲𝗱𝗮𝗻𝗰𝗲 𝟮.𝟬 𝗷𝘂𝘀𝘁 𝘀𝗵𝗼𝘄𝗲𝗱 𝘄𝗵𝗮𝘁 𝗶𝘁 𝗰𝗮𝗻 𝗱𝗼 𝘄𝗵𝗲𝗻 𝘆𝗼𝘂 𝗴𝗶𝘃𝗲 𝗶𝘁 𝘁𝗵𝗲 𝗿𝗶𝗴𝗵𝘁 𝗱𝗲𝘁𝗮𝗶𝗹𝗲𝗱 𝗽𝗿𝗼𝗺𝗽𝘁. The scene starts with a tight close-up on the pigeon's eye and head as it pushes off a rooftop ledge and drops into the morning city. Then in one continuous shot the camera pulls back and glides right alongside the bird — you feel the rush of wind through its feathers, the flicker of every wingbeat, the glass towers rushing past, and the streets far below. It weaves between buildings, skimming past windows and balconies. Suddenly a flock scatters in front of it — ultra slow motion begins. The pigeon banks hard and threads the gap, the camera following from below and behind, showing every detail: spread feathers, the tension in the wings, the city reflected in its eye. Then time returns to normal speed. It flares its wings, lands cleanly on a lamppost, ruffles its feathers, turns its head to the camera and tilts it with that confident, knowing look. No cuts. No transitions. Just one smooth cinematic shot with the real urban skyline, the sound of wind and wingbeats, speed, and that final confident glance. The most interesting part: We're at the point where one person can create dynamic, high-quality aerial action scenes that previously required drones, professional rigs, and a full crew. This changes the game for creators who want to make viral, eye-catching content. Save this post so you don't miss the full prompt and more content in this style 🤝 Follow for more powerful AI video prompts, cinematic examples, and next-level workflows ❤️‍🩹

@Psalteric I always know that OpenClaw is a powerful tool, now i figured out how it works

He cloned himself into AI agents on OpenClaw — now those bots design his product cards and banners, run his marketplace sales, and do it all while he sleeps. Sounds fake, but it's completely real. Watch till the very end to see the exact setup so you can copy it

Not gonna lie — I totally get why Fable 5 got banned for international users, and honestly the reasoning makes sense. But I'm beyond hyped it's finally BACK. I'm dropping a video with real examples of what Claude Fable 5 can actually do.

@Psalteric GLM become a great competitor for Opus and GPT

Open source is quietly winning the AI race. Same prompt — a "football player scores a goal" animation in one HTML file — across four models. Opus 4.8 - $0.41 GPT 5.5 - $0.17 GLM 5.2 - $0.05 Kimi K2.7 - $0.03 Same quality, 10x cheaper. Another win for open source.

@Psalteric Great article

Your AI writes flawless code but builds bad products. Why? It knows what you did, never why. Vercel's fix: turn design judgment into a system agents can read → SKILL.md → references/ → exemplars/. Stop feeding agents adjectives.