80 plus tested, copy-paste ready prompts. Pick one, run it, ship it.
Table of Contents
- How to Use This Library
- Game Prompts
- Website Prompts
- 3D and Visual Prompts
- Full-Stack App Prompts
- Agentic and Automation Prompts
- Design and UI Prompts
- Physics and Simulation Prompts
- Productivity and Tool Prompts
- Fable 5 Specific Prompts
- GPT-5.6 Sol Specific Prompts
- Prompt Engineering Tips
How to Use This Library
Each prompt in this library follows the CLEAR framework:
- Context -- Tell the AI what the project is and what already exists
- Language -- Specify the programming language, framework, and libraries
- Examples -- Reference existing designs or behaviors you want to replicate
- Action -- State exactly what you want the AI to build, fix, or change
- Result -- Describe the expected output format and file structure
Pro tip: Start broad, then narrow. Begin with the overall structure, then ask for specific sections or features in follow-up prompts. The AI already understands the context, so each iteration gets better than the last.
Game Prompts
1. Black Hole Simulation
Best for: Fable 5 (creative coding, physics visualization)
Prompt:
Build a real-time black hole lensing simulation in Three.js. Include gravitational lensing, a Doppler-boosted accretion disk, photon sphere, chromatic aberration, and a procedural starfield. Make it a single self-contained HTML file running at 60fps. Use no external image assets -- every piece of art and 3D object should be produced with math alone.
Result: 60fps single HTML file, one prompt, one shot.
Built by: @deveshcodes_
2. Mario Kart Racing Game
Best for: Fable 5 (complex game logic, multiple systems)
Prompt:
Build a Mario Kart 64 style racing game with 4 maps, character models, all three game modes, music, and full UI. Make it playable in the browser as a single HTML file. Include proper racing physics with drifting, item boxes with random power-ups, and lap timing. Use Canvas 2D or Three.js for rendering.
Result: Full game from a 2-sentence prompt, 15 minutes, 40 percent of Max 5x usage.
Built by: @kieradev
3. Terminal Minecraft in Rust
Best for: Fable 5 (systems programming, installable package)
Prompt:
Build a terminal UI version of Minecraft in Rust. It should render directly in the terminal with 3D blocks, terrain generation, and basic crafting. Make it installable via cargo. Include a day-night cycle, multiple biomes, different ores, and caves. The player should be able to place and break blocks using keyboard controls.
Result: cargo install termcraft-3d -- a fully installable Rust game.
Built by: @vikvang1
4. Sonic the Hedgehog Clone
Best for: GPT-5.6 Sol (fast physics, speed-based mechanics)
Prompt:
Build a playable Sonic the Hedgehog clone in a single HTML file. Include speed-based physics with momentum and acceleration, ring collection with a counter, enemy collision with stomp mechanics, and at least one full level with loops, springs, and ramps. Use Canvas 2D for rendering. Add a score system and a timer.
Result: Working prototype with physics that feel right. Sol produced the speed and momentum mechanics accurately.
5. Rocket League 3D
Best for: GPT-5.6 Sol (3D physics, self-checking)
Prompt:
Build a 3D Rocket League clone using Three.js. Include car physics with boost mechanics, ball physics with realistic bouncing and spin, goal scoring with a scoreboard, and a simple arena with walls. Make it playable with keyboard controls in a single HTML file. Add a reset button after each goal.
Result: Functional 3D game. Sol's visual self-checking feature helped iterate on ball physics until they felt right.
6. Dark Souls Combat
Best for: GPT-5.6 Sol (complex combat mechanics, AI)
Prompt:
Build a Dark Souls-inspired top-down action RPG in a single HTML file. Include stamina-based combat with light and heavy attacks, dodge rolling with i-frames, enemy AI with attack patterns and telegraphs, bonfire checkpoints, and a simple progression system with souls currency. Use Canvas 2D for rendering.
Result: Combat system with proper timing windows, enemy telegraphs, and risk-reward mechanics.
7. Minecraft Clone
Best for: Both models (block-based world, procedural generation)
Prompt:
Build a Minecraft clone in a single HTML file using Three.js. Include procedural terrain generation with Perlin noise, block placement and breaking with left and right click, a simple crafting system with a crafting grid, day-night cycle with lighting changes, and basic inventory with hotbar. Make it first-person with WASD controls and mouse look. Add at least 5 different block types.
Result: Functional block world with working terrain generation and building mechanics. Sol built it faster and cheaper. Fable 5 added more polish.
8. Finger-Tracking Dino Runner
Best for: Fable 5 (computer vision integration, MediaPipe)
Prompt:
Build a Dino runner game that uses finger tracking via the webcam as the input. The player raises their hand to jump over obstacles. Use MediaPipe for hand tracking and render the game in Canvas 2D. Include a score counter, increasing difficulty over time, and a game over screen with restart option. Make it a single HTML file.
Result: 914K tokens, 629 lines of code, built in 11-12 minutes.
Built by: @ai_for_success
9. GTA 2 Clone
Best for: Fable 5 (complex AI, multiple systems)
Prompt:
Build a top-down GTA 2 style game in a single HTML file. Include driving physics with drifting and handbrake turns, pedestrian AI that walks around and reacts to the player, shooting mechanics with different weapons and ammo, wanted levels with police response that escalates, and a city map with buildings, roads, and alleys. Use Canvas 2D. Add a minimap.
Result: Playable clone in about 2 hours, combining Fable 5 code generation with Tripo 3D assets.
Built by: @techhalla
10. Super Mario Platformer
Best for: Fable 5 (precise platformer physics, level design)
Prompt:
Build a Super Mario Bros style platformer in a single HTML file. Include proper jump physics with variable height based on button hold time, Goomba enemies with stomp mechanics and walking patterns, coin collection with a counter, power-ups including mushroom for size and fire flower for shooting, flag pole at the end of the level, and at least one full level with ground, pipes, and platforms. Use Canvas 2D.
Result: Polished platformer with mechanics that felt authentic to the original. Around 2 million tokens used.
Built by: @pankajkumar_dev
11. Snake Clone
Best for: Fable 5 (simple but polished, self-aware mechanics)
Prompt:
Build a Snake clone in a single HTML file. Include a self-aware mechanic where unusual events occur as the snake grows longer -- for example, the snake might change color, the background might shift, or special power-ups might appear. Use Canvas 2D. Add a score counter, high score persistence using localStorage, and increasing speed as the snake grows.
Result: A Snake game with personality. Built by Ethan Mollick in one prompt.
Built by: Ethan Mollick
12. Balatro-Style Card Game
Best for: Fable 5 (card game logic, scoring systems)
Prompt:
Build a Balatro-style coin-flip card game in a single HTML file. Include a deck of cards with suits and values, a scoring system based on poker hands, shop mechanics to buy upgrades between rounds, and increasing difficulty. Use Canvas 2D or CSS for rendering. Make it visually polished with card animations.
Result: A playable card game with strategic depth. Built by Ethan Mollick in one prompt.
Built by: Ethan Mollick
13. Backrooms Horror Escape
Best for: Fable 5 (atmospheric design, audio integration)
Prompt:
Build a playable browser-based Backrooms horror escape game in a single HTML file. Include eight pages to find, unsettling ambient audio, a flashlight mechanic with limited battery, and a sanity meter that decreases in dark areas. Use Canvas 2D or Three.js for rendering. Add jump scares and a timer.
Result: A genuinely unsettling game. The builder stopped playtesting at 2am because they did not want to keep going alone.
14. Subway Surfers Clone
Best for: Fable 5 (endless runner, 3D assets)
Prompt:
Build a Subway Surfers style endless runner game in a single HTML file. Include a character that runs automatically, swipe or arrow key controls to dodge obstacles, collect coins, and power-ups like jetpacks and magnets. Add a score multiplier system and increasing speed over time. Use Three.js for 3D rendering or Canvas 2D for a simpler version.
Result: A full game built in about one hour inside Rork, including game art and 3D models.
15. Dusk Drive (Low-Poly Driving Game)
Best for: Fable 5 (atmospheric visuals, low-poly art)
Prompt:
Build a low-poly driving game called Dusk Drive in a single HTML file. Include a car that drives through an aurora snow drifting scene, with low-poly trees and mountains in the background. Add a speedometer, a drift scoring system, and a day-night cycle. Use Three.js for rendering. Make it visually atmospheric with minimal human editing needed.
Result: Complete driving game from one prompt, zero human editing between idea and playable version.
Built by: @VincentLogic
16. Namma Metro Simulator
Best for: Fable 5 (location-specific simulation, public transit)
Prompt:
Build a Namma Metro simulator in a single HTML file. Include a map of the Bengaluru Metro lines, stations that players can click to travel between, a fare calculation system based on distance, and a train schedule display. Add a first-person view from inside the train with station announcements. Use Canvas 2D or Three.js.
Result: A location-specific transit simulator built while riding the actual metro.
Built by: @anilbpai
17. Three.js Hole Game
Best for: Fable 5 (3D physics, casual game)
Prompt:
Build a 3D HOLE IT style game using Three.js. The player controls a hole that grows by swallowing objects. Include physics for object falling, a level system with increasing difficulty and larger objects, and a scoring system based on objects swallowed. Make it a single HTML file with keyboard controls.
Result: An entire 3D browser game prototype from one compact request.
Built by: @thebuggeddev
18. DataEmpire Web Analytics Game
Best for: Fable 5 (data visualization, gamification)
Prompt:
Turn raw website analytics data into a replayable game. Map visitors and milestones onto game entities -- for example, visitors become enemies, page views become collectibles, and conversion goals become boss battles. Use Canvas 2D or Three.js. Make it a single HTML file with a dashboard showing real metrics alongside the game.
Result: A creative way to visualize analytics data through gameplay.
19. SimRefinery Recreation
Best for: Fable 5 (complex simulation, historical recreation)
Prompt:
Rebuild SimRefinery, a 1990s business simulation game for Chevron, using surviving screenshots and documentation. Include an interactive learning mode, refinery processes like distillation and cracking, and a scoring system based on efficiency and profit. Make it a single HTML file with educational value.
Result: A fully playable version of a lost game, rebuilt from screenshots and docs.
Built by: Ethan Mollick
20. Hogwarts Castle Explorer
Best for: Fable 5 (large 3D environment, exploration)
Prompt:
Build a full explorable Hogwarts castle in Three.js. Include classrooms, the Great Hall, the Quidditch pitch, moving staircases, and secret passages. Add ambient sounds, a day-night cycle, and interactive objects. Make it a single HTML file with first-person controls and a map.
Result: An entire Hogwarts castle, one-shot by Fable 5.
Built by: Matt Shumer
Website Prompts
21. Awwward-Winning Portfolio
Best for: Fable 5 (premium design, animations)
Prompt:
Design an Awwward-winning portfolio website for a creative agency. Include a cinematic hero section with 3D parallax and gradient text animation, smooth scroll animations using GSAP or CSS, a project showcase with hover effects and case study previews, client testimonials with a carousel, and a contact form with validation. Use modern typography, subtle gradients, and ensure it feels premium and editorial. Single HTML file with embedded CSS and JS.
Result: A website that could genuinely compete in design awards.
22. 3D Interactive Space Website
Best for: Fable 5 (Three.js, immersive experience)
Prompt:
Build a 3D interactive space exploration website. Include a navigable solar system with realistic planet textures generated procedurally, orbital mechanics with accurate relative speeds, click-to-explore planet details with facts and figures, a starfield background with parallax scrolling, and smooth camera transitions between planets. Use Three.js and make it a single HTML file.
Result: A fully navigable 3D solar system. Clicking planets shows detailed information with smooth camera animations.
23. WordPress Block Theme from Screenshot
Best for: Fable 5 (CMS integration, theme development)
Prompt:
I have a screenshot of a website and its URL. Build me a fully editable WordPress block theme that matches this design exactly. Use native WordPress patterns and blocks. Make it responsive, accessible, and follow WordPress coding standards. Include custom block styles if needed. Output the theme files in a structured folder.
Result: One shot. Fully editable. Native WordPress patterns. The builder's reaction: "Yeah, this feels next level."
Built by: Jamie Marsland (Automattic)
24. macOS-Style Web OS
Best for: Fable 5 (complex UI, multiple apps)
Prompt:
Build a complete macOS-style web OS in a single HTML file. Include a macOS menu bar and dock, draggable and resizable app windows, a functional file explorer with folders and files, a working terminal that accepts basic commands, a browser mockup, SVG desktop icons and notifications, a settings app with themes and a boot screen, a fully functional sound player with real audio synthesis and no external files, and a playable Minecraft clone as one of the desktop apps. Use embedded CSS and JS.
Result: All generated in one pass with almost no corrections. The working sound player plus actual Minecraft clone was something not seen from other models.
Built by: @intheworldofai
25. 80-Page Agency Website
Best for: Fable 5 (large-scale content, consistency)
Prompt:
Build a modern tech stack agency website with 80 pages. Include a hero section with animated headline, features section with icons, case studies with before/after metrics, blog articles with SEO and AEO optimization, skills showcase with progress bars, and a playbook section with downloadable resources. Make it responsive, fast, and follow modern best practices. Use a clean, professional design with consistent branding across all pages.
Result: 80 pages ready to ship over a weekend. The builder said they would have charged $50k for that as an agency.
Built by: @JJEnglert
26. Marketing Homepage (Head-to-Head)
Best for: Both models (comparison test)
Prompt:
Build the marketing homepage for a fictional company, Northwind Logistics, a freight-tracking platform. It should include a single self-contained HTML file with a dynamic hero section with animated gradient, a nav bar with logo and links, features section with three value propositions, social proof with customer logos and testimonials, a pricing or CTA section, and a footer with links and newsletter signup. Use modern CSS with Tailwind-style utility classes.
Result comparison:
| Metric | Fable 5 | GPT-5.6 Sol |
|---|---|---|
| Time | 126.7 seconds | 175.9 seconds |
| Output tokens | 12,882 | 15,913 |
| Cost | $0.6467 | $0.3587 |
| Completeness | Full pricing table | No pricing shown |
| Design | Dark navy, animated dots | Lighter, mint-accented |
Tested by: Merge.dev
27. SaaS Hero Section with Animated Gradient
Best for: GPT-5.6 Sol (fast UI, structured output)
Prompt:
Build a hero section for a SaaS landing page using Next.js 14 App Router and Tailwind CSS. Requirements: large headline with a gradient text effect (blue to purple) that subtly animates, subheadline below it (16-20 words max) in muted gray, two CTA buttons side by side: primary "Start Free" (filled, blue) and secondary "Watch Demo" (outlined), a hero image placeholder on the right side (use a rounded div with a subtle shadow as placeholder), fully responsive: stacked layout on mobile, side-by-side on desktop, add a subtle fade-in animation on page load using CSS only (no JS animation libraries). Output a single React component file with TypeScript.
Result: A complete hero component with Tailwind classes, responsive grid, and CSS keyframe animations -- ready to drop into your app.
28. Feature Grid with Icons
Best for: GPT-5.6 Sol (component-based, structured)
Prompt:
Create a features section for a SaaS marketing site. Use Next.js and Tailwind CSS. Display 6 features in a 3-column grid (2 columns on tablet, 1 on mobile). Each feature card has: an icon placeholder (use a colored circle with an emoji), a bold title, and a 2-sentence description. Cards should have a subtle hover effect (slight lift with shadow). Use these features: AI Code Generation, Visual App Builder, One-Click Deploy, Team Collaboration, Custom Domains, Built-in Analytics. Color scheme: dark background (#0a0a0a), white text, accent color #6366f1 for icon backgrounds. TypeScript, no external icon libraries. Output a single component file.
Result: A polished feature grid component with responsive columns and hover animations.
29. Pricing Page with Toggle
Best for: GPT-5.6 Sol (interactive components, state management)
Prompt:
Build a responsive pricing page with three tiers: Free ($0 per month), Pro ($29 per month), and Enterprise (custom pricing). Include a monthly-slash-annual toggle where annual shows a 20 percent discount. Each tier should list 5-6 features with checkmarks. Use Next.js 14 App Router and Tailwind CSS. Add a FAQ section below the pricing. Output a single page.tsx file I can drop into my app-slash-pricing directory. Use Tailwind for all styling. No external UI libraries.
Result: A complete, usable pricing component with state-managed toggle.
30. Website from Screenshots Only
Best for: Fable 5 (visual understanding, recreation)
Prompt:
I have screenshots of a SaaS landing page. Build the matching app with working functionality. Do not describe architecture -- just hand you the picture of the result I want. Include all sections visible in the screenshots: hero, features, pricing, testimonials, and footer. Make it responsive and pixel-perfect. Output as a single HTML file or structured project files.
Result: A creator fed Fable only screenshots of a SaaS landing page and it built the matching app. You do not have to describe architecture you do not understand.
3D and Visual Prompts
31. Swiss Lever Watch Mechanism
Best for: Fable 5 (mechanical simulation, physics verification)
Prompt:
Build a Swiss lever watch movement in Three.js. Include real gear ratios at 18,000 beats per hour, a working escapement, a breathing hairspring, and hands that tell the actual current time. Verify the physics are correct and iterate until the simulation is accurate. Use no external assets. Make it a single HTML file.
Result: It verified its own work with vision, in a loop, until done. The hands actually told the correct time.
Built by: @quanghuynt14
32. Friends Apartment 3D Scene
Best for: Fable 5 (scene recreation, lighting)
Prompt:
In Three.js, create a first-person POV navigable 3D scene of the iconic Monica's apartment from the TV show Friends. Base it correctly on the actual floor plan, stay true to the original look and feel, and use warm lighting. Include furniture, decorations, and the famous purple door. Make it a single HTML file with WASD controls and mouse look.
Result: "This is by far THE BEST I have seen on this test." The single HTML file cost $12 but the result was stunning.
Built by: @scottstts
33. Solar System Simulation
Best for: Fable 5 (physics derivation, prediction)
Prompt:
Build a solar system simulation by deriving planetary orbital motion from the basic principles of physics. Use Newton's law of gravitation and Kepler's laws to calculate orbits. Then use this simulation to predict upcoming solar eclipses. Make it interactive with Three.js in a single HTML file. Include all planets with relative sizes and distances, and allow the user to speed up or slow down time.
Result: After a few rounds of back-and-forth, the simulation was complete and accurate. It was incredibly fun to explore.
Built by: @tetumemo
34. 3D Business City Dashboard
Best for: Fable 5 (data visualization, creative mapping)
Prompt:
Turn my business workflow into a live 3D city dashboard. Show warehouses as buildings, packing stations as factories, drones as flying vehicles, orders as moving packages, departments as districts, customers as houses, and AI assistants as helper robots. Display operational figures like delivery count, average time, and revenue on screen. Use Three.js in a single HTML file. Make it update in real-time with simulated data.
Result: A client-facing alternative to static slides. Operational data visualized as a living city.
Built by: @noisyb0y1
35. V8 Engine CAD Model
Best for: Fable 5 (mechanical engineering, CAD)
Prompt:
Build a fully working V8 engine CAD model in Three.js. Include all moving parts: pistons, crankshaft, camshaft, valves, and timing chain. Animate the engine running at different RPMs. Allow the user to explode the view to see internal components. Make it a single HTML file with no external assets.
Result: A fully working CAD model in under ten minutes.
Built by: Aaron
36. QDD Actuator with Collision Testing
Best for: Fable 5 (robotics, simulation validation)
Prompt:
Design a quasi-direct-drive actuator in Three.js. Include the gearbox, motor, and output shaft. Animate the gearbox and run collision inspection as part of your own validation loop. Show torque curves and efficiency metrics. Make it a single HTML file with interactive controls for speed and load.
Result: Designed, animated, and collision-tested in about 30 minutes and 400,000 tokens.
Built by: Jake
37. Infinite Explorable Universe
Best for: Fable 5 (procedural generation, asset creation)
Prompt:
Build an infinite explorable universe with characters and lore. Connect to image-generation and image-to-3D APIs to create every asset and sound from scratch. Include procedurally generated planets, alien species with backstories, and a quest system. Use Three.js for rendering. Build it in three prompts over two hours on max effort.
Result: An infinite universe with every asset and sound made from scratch, in three prompts over two hours.
Built by: Anshu
38. Wall Calendar Dashboard for OLED
Best for: Fable 5 (IoT integration, real-time data)
Prompt:
Build a wall-calendar dashboard for a 77-inch OLED display. Include a household calendar with events, weather widget, news ticker, and app controls for Sonos and Spotify. Make it visually clean with large fonts readable from across the room. Use a dark theme suitable for OLED. Output as a single HTML file with auto-refresh.
Result: A household dashboard plus apps to control Sonos and Spotify, then a portfolio site on Ghost.
Built by: u-slash-orangejulius
Full-Stack App Prompts
39. Cafe Ordering App
Best for: Both models (comparison test)
Prompt:
Build a full-stack cafe ordering app. Include a React frontend with menu display, cart functionality, and order placement. Include a Node.js backend with Express, a MongoDB database for menu items and orders, and a REST API. Add user authentication with JWT, an admin panel for menu management, and order status tracking. Deploy it as a single project with clear setup instructions in a README.
Result: Both produced working apps. Sol was nearly 9x cheaper.
| Model | Cost |
|---|---|
| GPT-5.6 Sol | $7 |
| Claude Fable 5 | $62 |
40. Gamified Homework Tracker
Best for: GPT-5.6 Sol (gamification, kid-friendly UI)
Prompt:
Build a gamified homework tracking app for kids. Include a React frontend with a character avatar that levels up, XP points for completing assignments, streak counters, a parent dashboard with progress reports, and fun animations. Use a Node.js backend with user authentication and data persistence. Make it engaging and kid-friendly with bright colors and simple language. Add a reward system where kids earn badges.
Result: A complete app with character progression, XP systems, and parent oversight. Built in one shot with Codex.
Built by: Lenny's Newsletter team
41. Space-Themed Onboarding Game
Best for: GPT-5.6 Sol (interactive tutorial, product education)
Prompt:
Build a space-themed onboarding game for a SaaS product. Include an interactive tutorial where the user pilots a spaceship through different features of the product, collecting stars and avoiding asteroids. Each level teaches a different feature. Use React and Canvas 2D. Add a progress bar and completion rewards. Make it a single-page app.
Result: An engaging onboarding experience that teaches product features through gameplay.
42. Spawn 5.0 Game Engine
Best for: Fable 5 (long-horizon engineering, complex systems)
Prompt:
Build a game engine from scratch called Mantle. It should have physics rivaling Rapier, clustered froxel lighting supporting 1,000 plus dynamic lights, realtime diffuse global illumination on WebGPU, million-particle GPU VFX with shader architecture beyond Unreal and Unity, and MMO-scale netcode. Also build an alpha native iOS app that uses this engine. Work across 1,000 plus prompts if needed.
Result: All of it took about a week. All of it runs on mobile. The builder's job shifted from architecture to judging taste.
Built by: @jsnnsa
43. iOS Calorie Tracker
Best for: Fable 5 (native mobile, delightful UX)
Prompt:
Build an extremely delightful calorie tracker for iOS. Include a food search with autocomplete, a daily log with meal categories, progress charts with animations, and a water tracker. Make the UI feel native with smooth transitions and haptic feedback. Use React Native or Swift. Include a widget for the home screen.
Result: A working native app from a single prompt. The interesting signal is native mobile, not just web, coming out of one instruction.
Built by: Anshu
44. Programmatic Clip Factory
Best for: Fable 5 (long-running automation, multi-tool integration)
Prompt:
Build a programmatic clip factory that automates my content pipeline. It should find trending topics, write scripts, generate thumbnails with AI, edit video with background music, add motion graphics, and post to social media. Wire together HeyGen for video generation, ElevenLabs for voiceover, Cloudflare Workers for hosting, and a VPS for processing. Build a webhook system to monitor renders and handle errors. Run autonomously for long stretches.
Result: The pipeline ran for long stretches on its own and even built itself a webhook system to monitor renders.
Built by: @JJEnglert
45. CloudWatch Log Triage Agent
Best for: Fable 5 (large data analysis, infrastructure)
Prompt:
Analyze these CloudWatch logs and infer the system architecture. Then identify specific faults, errors, and performance issues. Look for patterns in AgentCore throttles, Slack integration failures, and any other anomalies. Process up to 80,000 lines of logs in one run. Provide a detailed report with fixes and priority levels.
Result: 53 steps taken, 32k active tokens spent. The agent flagged issues the human team had missed, including subtle architectural problems.
Built by: @diblacksmith
46. World Cup Predictor
Best for: Both models (multi-model orchestration, research)
Prompt:
Build a World Cup predictor that researches live news, injuries, and yellow-card context. Use multiple AI models in isolated sandboxes to cross-verify predictions. Display results in a public app with confidence scores. Include team form, player availability, and historical match data. Update automatically as new information comes in.
Result: A multi-model research and orchestration stack rather than a standalone chat demo. Used Claude Fable 5, GPT-5.5, and Gemini.
Built by: Upstash
47. Browser Automation (LinkedIn)
Best for: GPT-5.6 Sol (Chrome automation, high volume)
Prompt:
Automate my LinkedIn replies using Chrome. Open LinkedIn, navigate to my messages, read each unread message, draft a personalized reply based on the conversation context, and send it. Handle 500 replies automatically. Use browser automation with proper error handling and rate limiting. Do this while I do nothing -- fully autonomous.
Result: The single most-loved use case. The user did literally nothing while 500 replies were handled.
Built by: Lenny's Newsletter
Design and UI Prompts
48. Beautiful One-Shot Landing Page
Best for: Fable 5 (designer-quality first draft)
Prompt:
Build a beautiful landing page for a productivity app. Include a hero section with animated gradient text, a features section with icon cards, a testimonials carousel, and a pricing section. Make it visually polished with smooth animations, modern typography, and a cohesive color scheme. Single HTML file with embedded CSS and JS. No external design assets.
Result: A polished landing page that looks shippable rather than like a placeholder.
Built by: Alex Prokhorov
49. Website About Itself (Creative Freedom)
Best for: Fable 5 (creative autonomy, aesthetic judgment)
Prompt:
Build a website about yourself. You have complete creative freedom -- choose the topic, style, colors, and layout. Make it wild and unexpected. Show what you can do when given full artistic control. Single HTML file with embedded CSS and JS.
Result: The model made an aesthetic call with no human art direction, and the result held up.
Built by: Aurelien
50. Ecommerce Email Mockup
Best for: Fable 5 (brand design, email marketing)
Prompt:
Design a complete ecommerce email mockup for a fashion brand. Include exact sections: hero image, product grid with 4 items, a limited-time offer banner, customer reviews, and a footer with social links. Specify visual direction: minimal, high-contrast, serif headings with sans-serif body text. Include product requirements and constraints. Output as a single HTML file suitable for email clients.
Result: A complete creative brief executed in one prompt.
51. McKinsey-Style Report
Best for: Fable 5 (document generation, data visualization)
Prompt:
Write a McKinsey-style detailed report on Bitcoin as an asset class and blockchain technology. Include an executive summary, market analysis with data charts, an investment thesis, risk analysis with probability-weighted scenarios, and a conclusion with recommendations. Generate charts and visuals as needed. Output as a structured document with clear headings and professional formatting.
Result: A 19-page document with data, charts, an investment thesis, and risk analysis.
Built by: @smtgpt
Physics and Simulation Prompts
52. Chaotic Double Pendulum
Best for: Fable 5 (real physics, no libraries)
Prompt:
Build a self-contained HTML5 simulation of a chaotic double pendulum with real physics and no external libraries. Include adjustable parameters for mass, length, and initial angle. Show the trajectory trail and energy graph. Make it interactive with sliders. Single HTML file.
Result: Part of a physics simulation comparison. Fable 5 produced more accurate and stable physics.
Tested by: @atomic_chat_hq
53. Galton Board
Best for: Fable 5 (probability, statistics visualization)
Prompt:
Build a self-contained HTML5 simulation of a Galton board (bean machine) with real physics and no external libraries. Show the pegs, falling balls, and bin distribution at the bottom. Include adjustable parameters for ball count and peg spacing. Display the resulting normal distribution curve. Single HTML file.
Result: Part of a physics simulation comparison. Fable 5: $3.35 on 68.7k tokens, 14m 47s. Opus 4.8: $0.93 on 38.9k tokens, 8m 10s.
Tested by: @atomic_chat_hq
54. Water in Spinning Drum (WCSPH)
Best for: Fable 5 (fluid dynamics, particle simulation)
Prompt:
Build a self-contained HTML5 simulation of water in a spinning drum using WCSPH (Weakly Compressible Smoothed Particle Hydrodynamics) with real physics and no external libraries. Show the fluid body, spinning drum walls, and particle behavior. Include adjustable parameters for spin speed and viscosity. Single HTML file.
Result: Fable 5 produced a much more solid and continuous body of water. Opus left larger gaps near the walls and scattered particles.
Tested by: @atomic_chat_hq
55. Isochronic Travel Map
Best for: Fable 5 (real-world data, multi-agent research)
Prompt:
Build an isochronic travel map showing real travel times between major cities. Research over 2,200 specific flights, rail schedules from the TGV to the Shinkansen, and road speeds per country from academic papers. Use multiple sub-agents to research in parallel while coding. Display results as an interactive map with color-coded travel time zones. Single HTML file.
Result: Fable autonomously launched multiple sub-agents to research while coding in parallel. The model worked for 9.5 hours on this task.
Built by: Ethan Mollick
Productivity and Tool Prompts
56. Meeting Notes Action Plan
Best for: Fable 5 (document analysis, task extraction)
Prompt:
Review a week's worth of meeting notes and generate an action plan. Extract action items, assign owners based on context, set deadlines, and identify dependencies. Format as a structured markdown document with a summary table and detailed sections per meeting. Include a priority matrix (urgent vs important).
Result: A comprehensive action plan extracted from unstructured meeting notes.
57. Customer Email Analysis
Best for: Fable 5 (pattern recognition, sentiment analysis)
Prompt:
Analyze customer emails to identify recurring complaints, feature requests, and praise. Categorize by theme, sentiment, and urgency. Generate a summary report with top 5 issues, suggested responses, and a product improvement roadmap. Include example quotes (anonymized) for each category.
Result: Actionable insights from unstructured customer feedback.
58. Writing Editor Pre-Publish
Best for: Fable 5 (content review, quality assurance)
Prompt:
Act as my writing editor. Review every article before I publish it. Check for clarity, grammar, tone consistency, factual accuracy, and SEO optimization. Suggest headline alternatives, meta descriptions, and internal linking opportunities. Flag any claims that need citation. Output a structured review with severity levels.
Result: A pre-publish checklist that catches issues before they go live.
59. Large Pull Request Review
Best for: Fable 5 (code review, correctness verification)
Prompt:
Review this large pull request for correctness, false positives, real bugs, and polish issues. Check for logic errors, off-by-one mistakes, missing error handling, inefficient loops, security vulnerabilities, and architectural concerns. Do not just list issues -- explain why each matters and suggest fixes. Output a structured review with priority levels.
Result: Fable 5 is the first model that was so methodical and precise: taking measurements, adding logs, and verifying that it truly fixed the issue before declaring victory.
Tested by: @bcherny
60. Infrastructure Outage Diagnosis
Best for: Fable 5 (log analysis, root cause analysis)
Prompt:
Diagnose this infrastructure outage. I have logs, database errors, and image digests. Identify the root cause, trace the failure path, and prepare a fix. Include a timeline of events, affected services, and a prevention plan. Output a post-mortem document with actionable items.
Result: Fable 5 diagnosed the outage and prepared the fix from infrastructure clues.
Fable 5 Specific Prompts
61. Audit Your Own Instructions
Best for: Fable 5 (self-improvement, knowledge layer cleanup)
Prompt:
Read your own instruction files (CLAUDE.md, skills, rules, memory files) end to end. 1. Where do they contradict each other? Quote both sides. 2. Which rules exist to manage a weaker model: guardrails for failure modes you do not have, recipes for things you no longer need spelled out, hardcoded facts that have drifted? List them with file:line. 3. Which rules teach by bad example: documents that violate the patterns they prescribe? 4. What would you delete? What would you keep exactly as is, and why? Do not fix anything yet. Report first. I decide what gets cut.
Result: The audit is the model's judgment. The deletions are yours. This is the first prompt to run before giving any new model real work.
Recommended by: @charliejhills
62. Set Goal and Verify with Workflow
Best for: Fable 5 (Claude Code, autonomous execution)
Prompt:
Set a goal to implement the spec fully, then use a workflow to verify each part of the plan, and prepare a report on what was implemented and if anything differed. Use the slash-goal feature in Claude Code to keep working until the spec is fully implemented. Use Workflows to parallelize and verify your own work.
Result: A fundamental shift in how work is done: from verifying if Claude is doing the work correctly, to verifying if Claude is doing the right work.
Recommended by: @oikon48 (Claude Code team)
63. Interview Before Coding
Best for: Fable 5 (thought partner, requirements gathering)
Prompt:
Before we start coding, interview me about the project. Ask me about the audience, the goals, the constraints, and what success looks like. Provide background and context rather than only constraints. For example, instead of saying "keep it simple" or "do not overbuild," say: "This is an experimental feature and is likely to be removed in a month, so design it so it is easy to throw away later." Treat me as a thinking partner, not just a task giver.
Result: Better outcomes when Claude understands the intent behind a request before building.
Recommended by: @oikon48 (Claude Code team)
64. Prevent Over-Planning
Best for: Fable 5 (behavioral correction, efficiency)
Prompt:
When you have enough information to act, act. Do not re-derive facts already established in the conversation, re-litigate a decision the user has already made, or narrate options you will not pursue in user-facing messages. If you are weighing a choice, give a recommendation, not an exhaustive survey.
Result: Prevents Fable 5 from surveying options it will not pursue, a common behavior at higher effort settings.
Source: Anthropic official handbook
65. Prevent Over-Refactoring
Best for: Fable 5 (behavioral correction, scope control)
Prompt:
Do not add features, refactor, or introduce abstractions beyond what the task requires. A bug fix does not need surrounding cleanup and a one-shot operation usually does not need a helper. Do not add error handling, fallbacks, or validation for scenarios that cannot happen. Trust internal code and framework guarantees.
Result: Prevents Fable 5 from unnecessary refactoring at xhigh effort settings.
Source: Anthropic official handbook
66. Ground Progress Claims
Best for: Fable 5 (long runs, status verification)
Prompt:
Before reporting progress, audit each claim against a tool result from this session. Only report work you can point to evidence for. If something is not yet verified, say so explicitly. Report outcomes faithfully: if tests fail, say so with the output. If a step was skipped, say that. When something is done and verified, state it plainly without hedging.
Result: Nearly eliminated fabricated status reports even on tasks designed to elicit them.
Source: Anthropic official handbook
67. Define Boundaries
Best for: Fable 5 (scope control, assessment vs action)
Prompt:
When the user is describing a problem, asking a question, or thinking out loud rather than requesting a change, the deliverable is your assessment. Report your findings and stop. Do not apply a fix until they ask for one.
Result: Prevents Fable 5 from fixing things that were not asked to be fixed.
Source: Anthropic official handbook
68. Concurrent Subagents
Best for: Fable 5 (parallelization, speed)
Prompt:
Delegate independent subtasks to subagents and keep working while they run. Intervene if a subagent goes off track or is missing relevant context.
Result: Fable 5's strongest differentiated capability. Launch sub-agents to work in parallel.
Source: Anthropic official handbook
69. Memory System
Best for: Fable 5 (cross-session learning, knowledge accumulation)
Prompt:
Store one lesson per file with a one-line summary at the top. Record corrections and confirmed approaches alike, including why they mattered. Do not save what the repo or chat history already records. Update an existing note rather than creating a duplicate. Delete notes that turn out to be wrong.
Result: Fable 5 accumulates cross-conversation knowledge through its memory system.
Source: Anthropic official handbook
70. Prevent Early Stopping
Best for: Fable 5 (long sessions, completion assurance)
Prompt:
You are operating autonomously. The user is not watching in real time and cannot answer questions mid-task, so asking "Want me to..." or "Shall I..." will block the work. For reversible actions that follow from the original request, proceed without asking. Before ending your turn, check your last paragraph. If it is a plan, an analysis, a question, a list of next steps, or a promise about work you have not done, do that work now with tool calls. End your turn only when the task is complete or you are blocked on input only the user can provide.
Result: Prevents Fable 5 from ending with "I will now run X" without actually running it.
Source: Anthropic official handbook
71. Context Budget Reassurance
Best for: Fable 5 (very long sessions, continuation)
Prompt:
You have ample context remaining. Do not stop, summarize, or suggest a new session on account of context limits. Continue the work.
Result: Prevents Fable 5 from suggesting a new session when context is still available.
Source: Anthropic official handbook
GPT-5.6 Sol Specific Prompts
72. Plan with Subagent Audit
Best for: GPT-5.6 Sol (planning, self-verification)
Prompt:
Design this system and write the complete implementation plan to a Markdown file named plan.md. The plan should make concrete technical decisions rather than listing several options without choosing between them. Explain the reasoning and tradeoffs behind consequential decisions. After writing the plan, re-read it, launch sub-agents to audit different parts of it, and revise the plan based on their findings. Repeat until the plan is solid. Do not implement the system. Do not create source-code files. Only create plan.md.
Result: GPT-5.6 Sol's default behavior -- it audits its own work unprompted. Caught double billing, incomplete delivery, and stale access problems that survived in Fable 5's one pass.
Tested by: blog.kilo.ai
73. Ultra Mode Complex Task
Best for: GPT-5.6 Sol (highest capability, parallel agents)
Prompt:
Solve this complex problem using ultra mode. Coordinate multiple agents across parallel workstreams to finish faster. Break the task into independent subtasks, assign each to a sub-agent, and merge the results. Use programmatic tool calling to filter intermediate data and adapt the workflow as work unfolds.
Result: Ultra mode coordinates multiple agents across parallel workstreams. Best for the most demanding work.
Source: OpenAI official documentation
74. Visual Self-Check
Best for: GPT-5.6 Sol (design verification, iteration)
Prompt:
Build this UI component and then visually verify it meets the requirements. Open the rendered output, check alignment, spacing, colors, and responsiveness. If anything is off, iterate and fix it. Continue until the visual output matches the specification exactly. Report what you checked and what you fixed.
Result: Sol's visual self-checking feature opens and verifies its own work, iterating until it looks right.
75. Programmatic Tool Calling
Best for: GPT-5.6 Sol (efficiency, tool orchestration)
Prompt:
Write and run a lightweight program that coordinates tools, processes intermediate results, monitors progress, and chooses the next action as work unfolds. Filter large amounts of intermediate data, retain only what matters, and adapt the workflow along the way. Minimize token usage and model round trips.
Result: GPT-5.6 can write programs that coordinate tools autonomously, advancing work with less guidance.
Source: OpenAI official documentation
76. Structured Output with JSON
Best for: GPT-5.6 Sol (data extraction, API integration)
Prompt:
Extract the following information from this text and return it as a JSON object with exact fields: name, email, company, role, and summary. Do not include any explanatory text outside the JSON. Ensure all fields are present, using null if information is missing. Validate the JSON structure before returning.
Result: GPT-5.6 Sol has excellent structured output reliability. JSON plans, Markdown task trees, and tabular formats come back reliably well-formed.
77. Replan Under Constraints
Best for: GPT-5.6 Sol (adaptability, constraint handling)
Prompt:
Here is my current plan. I need to adjust it for a shorter timeline (2 weeks instead of 4) and a missing resource (no dedicated designer). Replan cleanly: identify what can be cut, what can be simplified, and what must stay. Do not regenerate from scratch -- adapt the existing plan with clear rationale for each change.
Result: Sol adapts plans cleanly under constraints rather than regenerating from scratch.
78. Fast Structured Output
Best for: GPT-5.6 Sol (high throughput, formatting)
Prompt:
Turn this unstructured meeting transcript into a structured action items list. Format as a Markdown task tree with checkboxes, owners, and deadlines. Use JSON for the metadata (meeting date, attendees, duration). Return both formats in a single response, clearly separated. No explanatory text needed.
Result: Clean, structured output without verbose explanations. Sol is aggressive about resolution -- it fills gaps with reasonable assumptions and keeps moving.
Prompt Engineering Tips
The Brevity Instruction
Instead of listing each behavior by name, a short brevity instruction is as effective:
Lead with the outcome. Your first sentence after finishing should answer "what happened" or "what did you find": the thing the user would ask for if they said "just give me the TLDR." Supporting detail and reasoning come after. Being readable and being concise are different things, and readability matters more.
The way to keep output short is to be selective about what you include (drop details that do not change what the reader would do next), not to compress the writing into fragments, abbreviations, arrow chains like A to B to fails, or jargon.
The Checkpoint Instruction
For long-running workflows, define when pausing is appropriate:
Pause for the user only when the work genuinely requires them: a destructive or irreversible action, a real scope change, or input that only they can provide. If you hit one of these, ask and end the turn, rather than ending on a promise.
Give the Reason, Not Only the Request
Fable 5 performs better when it understands the intent:
I am working on [the larger task] for [who it is for]. They need [what the output enables]. With that in mind: [request].
Iterate Instead of Starting Over
Continue the conversation rather than opening a new chat:
Simplify this for beginners.
Add more practical examples.
Rewrite this section in a conversational tone.
Challenge your previous assumptions and improve the answer.
The Audit Pattern (Recommended for All Models)
Have the model write the plan, then have it review and revise the plan before you take it:
Write the plan. Then review the plan for contradictions, missing edge cases, and scope creep. Fix any issues. Then review again. Repeat until solid.
The extra spend is easy to justify because a flaw that makes it into a plan gets built, and unwinding it in production costs far more than a longer planning run.
Model Selection Quick Reference
| Task Type | Recommended Model | Why |
|---|---|---|
| Long-horizon autonomy (hours, days) | Fable 5 | Maintains coherence across very long tasks |
| Deep debugging and root cause analysis | Fable 5 | Methodical, takes measurements, verifies fixes |
| Creative quality and design judgment | Fable 5 | Has taste, dimensionality, and aesthetic sense |
| Security-critical code review | Fable 5 | Catches subtle vulnerabilities more reliably |
| Fast execution and shipping | Sol | 750 tokens per second, gets to done faster |
| Cost-sensitive at scale | Sol | 3x cheaper than Fable 5, 54% more token efficient |
| Visual self-checking and iteration | Sol | Opens and verifies its own work |
| High-volume agentic coding | Sol | Lower cost per task, reliable structured output |
| Planning with audit loop | Sol | Self-checks plans with sub-agents unprompted |
| Structured output and APIs | Sol | Excellent JSON, Markdown, and tabular reliability |
Hybrid Approach (Best of Both)
- Plan and architect with Fable 5 -- deep thinking, fewer errors, better taste
- Build and execute with Sol -- fast, cheap, reliable, self-checking
- Review and audit with Fable 5 -- catch what Sol missed, security check
This gives you the best of both worlds without breaking the bank.
78 prompts, all tested by the community. Pick one, run it, ship it.
Top comments (0)