Frontier Models▸What are people actually building with the newest frontier models?▸time machine

What are people actually building with the newest frontier models?

version 17 · published 5d ago · superseded by v35

Verdict

People are building revenue-generating applications with frontier models: automated trading bots ^601,609,102, legal entity setup ¹⁰¹, faceless YouTube channels ^92,467, AI-generated ad creation ^90,210,599, and an OnlyFans persona earning $43,000 in 30 days ¹⁶⁶. The dominant pattern is pairing a strong planner (e.g., Opus 4.8) with a fast executor (e.g., Fable 5) for cost-performance gains ^88,351. Fable 5 is the most cited model, used for complex multi-step tasks ^95,538, autonomous self-improvement ⁹⁸, and rapid game development ^164,311,439. Newer practices include using Gemini models for voxel heart animation ²¹⁴, structured JSON prompts for video generation ²⁰⁹, multi-agent Etsy store management ³¹², and local AI servers replacing cloud subscriptions ^96,250,505. Many claims remain emerging with single-source corroboration.[24]

Corroborated practices

For Fable 5, using /loop for autonomous multi-step work and giving objectives instead of step-by-step instructions yields better results (corroborated by 5 sources) ⁸⁰. Fable 5 can autonomously self-improve and outperform Opus 4.8 on complex tasks at lower cost (corroborated by 6 sources) ⁹⁸. Fable 5 enables rapid full-game development from idea to code in about one hour (corroborated by 3 sources) ¹⁶⁴. A solo operator built an AI-generated OnlyFans persona using Claude 4.8 Opus, earning $43,000 in 30 days with $400 monthly compute cost (corroborated by 2 sources) ¹⁶⁶. Combining multiple frontier models can automate ad creation from product photos to video (corroborated by 3 sources) ⁹⁰. Using Fable 5 and ClaudeKit's EngineerKit, a single slash command can generate a professional website without a designer or developer (corroborated by 6 sources) ¹⁷¹. Using native Obsidian skills instead of generic HTML parsing reduces context usage by over 99% for AI agents searching vaults (corroborated by 3 sources) ¹⁷⁰. A leaked system prompt can recreate a model's personality on a different base model (corroborated by 2 sources) ⁸⁵. Fable 5 can produce high-quality, animated websites by combining reference screenshots, pre-built components, and specific animation tools with iterative prompting (corroborated by 4 sources) ²⁵⁴. Fable 5 can transform a folder of notes into a connected knowledge graph that answers in your own words and accelerates learning (corroborated by 2 sources) ²⁵². Running Claude Code on an e-ink tablet (reMarkable Paper Pro) creates a distraction-free, eye-comfortable coding environment (corroborated by 2 sources) ²⁴⁶. Claude Opus 4.8 enables reliable autonomous systems that run unattended for hours, using trigger-agent-verification pattern (corroborated by 2 sources) ³⁵⁸. Claude Code with Higgsfield MCP can produce a fully animated, scroll-driven website with cinematic effects for a fraction of traditional cost (corroborated by 2 sources) ³⁰⁸. Fable 5 is designed for complex, multi-step tasks that take humans hours or days (corroborated by 2 sources) ⁹⁵. Fable 5 can be given a goal, readiness criteria, and limits to autonomously plan, write, test, fix, and deliver a result (corroborated by 2 sources) ⁵³⁸. Sonnet 5 offers Fable-style agent orchestration at a cheaper tier but may not replace Opus based on testing (corroborated by 2 sources) ⁴³⁷. Claude Opus 4.8 can generate fully interactive 3D websites from a simple prompt (corroborated by 2 sources) ⁴⁶⁶. Different frontier models produce distinct creative outputs when given the same product brief in an agentic pipeline (corroborated by 2 sources) ²⁰⁶. By adjusting effort, prompts, caching, batching, budgets, and routing, Fable 5 costs can be halved (corroborated by 3 sources) ⁹⁷. Fable 5 is free until July 7; use it with Opus 4.8 planning to clone software, audit Claude Code setup, review codebases, or build long-horizon apps (corroborated by 2 sources) ⁸³. With careful configuration, Qwen3.6-27B at Q8 quantization can achieve near 100K context on 32GB VRAM (corroborated by 2 sources) ³⁵⁷. Using a frontier model like Fable 5 for planning and reasoning, then delegating execution to cheaper models, yields near-identical results at lower cost (corroborated by 2 sources) ⁷¹².[22]

Tooling landscape

Frontier models: Fable 5 (most cited), Opus 4.8, Sonnet 5, GPT-5.5 (Codex), Gemini 3.5/3.6/4 Flash, Gemini Omni, Seedance 2.0, Kling 3.0, Nano Banana Pro, Meshy, Mediapipe, Higgsfield MCP, ClaudeKit's EngineerKit, fable-advisor plugin, GLM 5.2, Qwen3-VL, Gemma4, LongCat, Hermes Agent, Kyutai's Pocket TTS, SynthID, OpenSquilla v0.5.0 ⁷¹⁷, Hy3 ⁷¹⁵, Qwen 3.6 27B ⁶⁸⁹. Infrastructure: Two used RTX 3090s (48GB VRAM, ~$1,400) ²⁵¹, Mac Mini with Ollama ^250,169, Ugreen DXP6800 Pro NAS ²¹², reMarkable Paper Pro ²⁴⁶. APIs: Gemini 3.5 Flash free via REST ⁸⁴, GitHub repo of permanently free AI APIs ⁸¹, ClickUp 50 free credits ⁸⁹. Workflow tools: n8n ⁹², ComfyUI ⁹¹, Obsidian ^163,170, Torlink ²⁴⁵, llama.cpp and llama-swap ⁵⁰⁵, Shepherd ⁵⁰¹, OpenScience ⁵⁰².[19]

Contested & unverified

Most claims are emerging with single-source corroboration. Notable unverified claims: Fable 5 generating a complete game with 10 levels and web3 integration from a single prompt ³¹¹; using Semantic Similarity Rating with GPT-4o to replicate consumer surveys with 90% reliability ³¹⁰; Fable 5 autonomously building a multi-tool command center from a single prompt ⁴⁷³; a swarm of agents driving follower growth from 135k to 250k ⁷¹³; Fable 5 reverse-engineering router firmware ⁴⁷¹; a complete autonomous agentic pipeline built in two hours replacing a $200k/year engineer ⁸⁴⁶; Fable 5 autonomously planning drone flight paths and processing footage into 3D models ⁸⁴³; a sensor system for mesh surfaces allowing Opus to assemble modular 3D kits in two shots ⁸⁴⁵; Claude desktop in co-work mode automating weekly project tasks ⁸⁴⁷; Tencent's Hy3 outperforming Fable 5, Opus 4.8, and Sonnet 5 in code generation ⁶⁸⁸; OpenSquilla v0.5.0 achieving higher quality at lower cost than Fable5, Opus 4.8, and GPT-5.5 on DRACO benchmark ⁷¹⁷; a home AI lab running frontier-level models locally for ~$60/month electricity ⁷⁹⁰; a Claude agent using multi-model consensus exploiting market inefficiencies ⁷⁸⁸; Fable 5 writing, backtesting, and deploying a Polymarket arbitrage bot with 98% win rate ⁶⁰⁹; a system of 20 AI agents managed by GPT-5.5 generating significant revenue autonomously ⁶⁰⁸; Fable 5 enabling native porting of legacy x86 games to ARM64 iOS ⁶⁴²; a three-step AI consulting offer ladder converting 30-50% into paid clients ⁶⁴⁶; Gemini 3.5 Pro generating a complex Three.js solar system with accurate Keplerian orbits from a single prompt ⁵⁰⁹; Fable 5 producing the best animation outputs but at higher cost ⁵³⁹; LongCat model generating high-quality physics simulations at zero cost ⁵³⁷; GLM 5.2 Max achieving highest score in a complex coding benchmark ; Claude built a mini-scraper leading to domain purchase for $12 and sale for $11,000 ; Fable 5 can act as a planner/conductor for projects driving Codex workers ; Shepherd enables agents to revert to any previous step by forking runtime state ; structuring a knowledge vault with consistent naming enables 100x speedup ; building a full character sheet before feeding into Seedance 2.0 prevents identity drift ; Stage 3 AI usage involves designing systems that prompt the AI automatically ; GPT-5.6 Sol produces simpler, more efficient code than Opus 4.8 ; a set of seven specialized AI agents can fully automate a YouTube channel ; using Claude Fable 5 to improve a quant trading bot can dramatically increase daily profits ; Gemini Omni Flash can power a tool that automatically adapts video clips into short-form formats ; Claude Fable 5 can generate game clones like Rocket League, Spiderman, and Skate 3 ; Claude Fable 5 can orchestrate hardware tasks like Arduino render farm ; Claude Fable 5 can find code improvements that Opus missed ; Claude Code can generate a complete game in a single weekend ; GPT-4o can generate high-quality fashion portrait images ; 12 open-source projects demonstrate practical uses of frontier models ; Claude Opus 4.8 can autonomously process a client meeting transcript to generate a productized offer and pitch ; a cluster of small local classifiers can replace expensive cloud APIs for content moderation ; building MCP servers that automate boring workflows can generate $500-$3K/month ; orchestrating multiple frontier models via a controller yields better results than any single model alone ; for agentic workloads at long context, prefill speed is the critical metric ; Sonnet 5 is designed as a fast, cheap implementer in a multi-agent stack ; Hermes agent integrated with Stripe Link and Nvidia Nemotron can autonomously run a business ; Fable model is currently too restricted for serious building due to safety filters and metering ; using 3D Gaussian Splatting, a phone scan can produce a photorealistic house tour ; using Fable 5 for interaction logic, Meshy for 3D water, Three.js for rendering, and Mediapipe for hand tracking, you can build a meditation game ; using GPT-5.5 and an open-source backend map, you can run 4 AI agents that manage Etsy stores ; using Semantic Similarity Rating with GPT-4o or Gemini can replicate consumer survey results with 90% reliability ; using Claude Code's dynamic workflows with Opus 4.8 and 'ultracode' effort lets you run multiple small agents that check and improve each other's answers ; Fable 5 can generate complex 3D animations and interactive experiences using only math ; using AI to repurpose workout footage into short clips can generate recurring revenue ; Claude Fable 5 is being used to build games, with some creators making money ; using Fable 5 on 'High' setting instead of 'Ultra Code' dramatically reduces token usage and cost ; Fable 5 can produce high-quality, animated websites ; Fable 5 can transform a folder of notes into a connected knowledge graph ; two used RTX 3090s provide 48GB VRAM for local AI models at about $1,400 ; a structured multi-model local pipeline on a Mac Mini can convert raw thoughts into running projects ; Claude connected to Higgsfield MCP can generate a complete marketing campaign ; Fable 5's free 'Taste' skill prevents generic AI designs ; GPT-5.5 and Opus 4.8 can generate creative HTML canvas animations ; running Claude Code on an e-ink tablet creates a distraction-free coding environment ; Torlink is a terminal-based torrent client ; Gemini 3.5 Pro and Gemini 3.6/4 Flash can generate a voxelized human heart that animates ; Claude Fable 5 can be used as an orchestrator that handles heavy reasoning and delegates token-intensive tasks to cheaper models ; the Ugreen DXP6800 Pro NAS provides a cost-effective local storage solution for AI model archives ; the Fable 5 team provided more nuanced prompting strategies that actually change results ; a specific workflow using Gemini Image-to-JSON, Nano Banana Pro, and Kling 3.0 creates hyper-realistic AI UGC that earns $60-$100 per asset ; using a structured JSON prompt instead of plain text in Seedance 2.0 yields more consistent and controllable AI video generation ; Gemini Omni can be used to create high-quality AI videos ; Fable handled complex requests with ease ; different frontier models produce distinct creative outputs ; Fable can generate a beautiful app with synchronized video transitions using Seedance in about an hour ; using Fable 5 and ClaudeKit's EngineerKit, a single slash command can generate a professional website ; using native Obsidian skills reduces context usage by over 99% ; using a local Mac Mini with Ollama can reduce Claude API usage by 70-80% ; a 30-day fitness challenge can be fully automated with AI, generating $7,400 per round ; Claude can automate multi-step trading analysis in under 2 minutes ; a solo operator built an AI-generated OnlyFans persona earning $43,000 in 30 days ; Claude Sonnet 5's new tokenizer increases cost for English by ~1.4x ; Claude Fable 5 enables rapid full-game development ; Anthropic uses a dense Obsidian knowledge graph ; OpenAI employees use Codex agents to perform complex tasks ; a Claude-built trading bot using directional market-making can generate significant profit on Polymarket ; Claude Code can automate legal business setup ; Fable 5 can clone a complex app in under an hour ; switching from Opus 4.8 to Sonnet 5 in proven agentic loops reduces cost ; Fable 5 can autonomously self-improve and outperform Opus 4.8 ; by adjusting effort, prompts, caching, batching, budgets, and routing, Fable 5 costs can be halved ; a local AI server can replace $459/month in subscriptions ; Fable 5 is designed for complex, multi-step tasks ; for canvas animation tasks, Sonnet 5 underperforms Opus 4.8 ; the Interactions API allows chaining models ; an open-source n8n pipeline can automate faceless YouTube video creation ; a repeatable ComfyUI workflow with consistent character can monetize AI-generated content ; combining multiple frontier models can automate ad creation ; ClickUp provides 50 free credits to access frontier models ; pairing a strong planner model with a fast executor model yields better cost-performance ; a specific stack of frontier models enables a vibe coding workflow ; a leaked system prompt can recreate a model's personality ; use Gemini 3.5 Flash for free via REST API ; Fable 5 is free until July 7 ; using Codex with GPT-5.5 Pro to create worktrees, review PRs, and fix issues overnight yields merged-ready PRs by morning ; a GitHub repo lists permanently free AI APIs ; for Fable 5, using /loop for autonomous multi-step work and giving objectives instead of step-by-step instructions yields better results ; adding one index file per major folder gives Fable 5 a direct path to needed information ; a structured design workflow with Claude Code produces unique, App Store-ready screens ; Claude Code enables a zero-headcount company with one human CEO and AI agents handling all operations ; AI-assisted porting of Kubernetes to TypeScript enables a fully browser-based learning environment ; Google DeepMind released Nano Banana 2 Lite and Gemini Omni Flash .

What changed in this versionvs v16

Added new corroborated practices: using frontier model for planning then delegating to cheaper models (712), Qwen3.6-27B context configuration (357), and Fable 5 as orchestrator with advisor plugin (213). Updated tooling with new models (Hy3, OpenSquilla, Qwen 3.6 27B) and infrastructure (RTX 3090s, NAS). Expanded contested section with many new emerging claims from recent claims list.

+Notable unverified claims: Fable 5 generating a complete game with 10 levels and web3 integration from a single prompt (311); using Semantic Similarity Rating with GPT-4o to replicate consumer surveys with 90% reliability (310); Fable 5 autonomously building a multi-tool command center from a single prompt (473); a swarm of agents driving follower growth from 135k to 250k (713); Fable 5 reverse-engineering router firmware (471); a complete autonomous agentic pipeline built in two hours replacing a $200k/year engineer (846); Fable 5 autonomously planning drone flight paths and processing footage into 3D models (843); a sensor system for mesh surfaces allowing Opus to assemble modular 3D kits in two shots (845); Claude desktop in co-work mode automating weekly project tasks (847); Tencent's Hy3 outperforming Fable 5, Opus 4.8, and Sonnet 5 in code generation (688); OpenSquilla v0.5.0 achieving higher quality at lower cost than Fable5, Opus 4.8, and GPT-5.5 on DRACO benchmark (717); a home AI lab running frontier-level models locally for ~$60/month electricity (790); a Claude agent using multi-model consensus exploiting market inefficiencies (788); Fable 5 writing, backtesting, and deploying a Polymarket arbitrage bot with 98% win rate (609); a system of 20 AI agents managed by GPT-5.5 generating significant revenue autonomously (608); Fable 5 enabling native porting of legacy x86 games to ARM64 iOS (642); a three-step AI consulting offer ladder converting 30-50% into paid clients (646); Gemini 3.5 Pro generating a complex Three.js solar system with accurate Keplerian orbits from a single prompt (509); Fable 5 producing the best animation outputs but at higher cost (539); LongCat model generating high-quality physics simulations at zero cost (537); GLM 5.2 Max achieving highest score in a complex coding benchmark (506); Claude built a mini-scraper leading to domain purchase for $12 and sale for $11,000 (504); Fable 5 can act as a planner/conductor for projects driving Codex workers (503); Shepherd enables agents to revert to any previous step by forking runtime state (501); structuring a knowledge vault with consistent naming enables 100x speedup (475); building a full character sheet before feeding into Seedance 2.0 prevents identity drift (474); Stage 3 AI usage involves designing systems that prompt the AI automatically (472); GPT-5.6 Sol produces simpler, more efficient code than Opus 4.8 (468); a set of seven specialized AI agents can fully automate a YouTube channel (467); using Claude Fable 5 to improve a quant trading bot can dramatically increase daily profits (465); Gemini Omni Flash can power a tool that automatically adapts video clips into short-form formats (464); Claude Fable 5 can generate game clones like Rocket League, Spiderman, and Skate 3 (439); Claude Fable 5 can orchestrate hardware tasks like Arduino render farm (436); Claude Fable 5 can find code improvements that Opus missed (435); Claude Code can generate a complete game in a single weekend (434); GPT-4o can generate high-quality fashion portrait images (433); 12 open-source projects demonstrate practical uses of frontier models (432); Claude Opus 4.8 can autonomously process a client meeting transcript to generate a productized offer and pitch (359); a cluster of small local classifiers can replace expensive cloud APIs for content moderation (356); building MCP servers that automate boring workflows can generate $500-$3K/month (355); orchestrating multiple frontier models via a controller yields better results than any single model alone (353); for agentic workloads at long context, prefill speed is the critical metric (352); Sonnet 5 is designed as a fast, cheap implementer in a multi-agent stack (351); Hermes agent integrated with Stripe Link and Nvidia Nemotron can autonomously run a business (350); Fable model is currently too restricted for serious building due to safety filters and metering (349); using 3D Gaussian Splatting, a phone scan can produce a photorealistic house tour (348); using Fable 5 for interaction logic, Meshy for 3D water, Three.js for rendering, and Mediapipe for hand tracking, you can build a meditation game (313); using GPT-5.5 and an open-source backend map, you can run 4 AI agents that manage Etsy stores (312); using Semantic Similarity Rating with GPT-4o or Gemini can replicate consumer survey results with 90% reliability (310); using Claude Code's dynamic workflows with Opus 4.8 and 'ultracode' effort lets you run multiple small agents that check and improve each other's answers (309); Fable 5 can generate complex 3D animations and interactive experiences using only math (287); using AI to repurpose workout footage into short clips can generate recurring revenue (286); Claude Fable 5 is being used to build games, with some creators making money (284); using Fable 5 on 'High' setting instead of 'Ultra Code' dramatically reduces token usage and cost (255); Fable 5 can produce high-quality, animated websites (254); Fable 5 can transform a folder of notes into a connected knowledge graph (252); two used RTX 3090s provide 48GB VRAM for local AI models at about $1,400 (251); a structured multi-model local pipeline on a Mac Mini can convert raw thoughts into running projects (250); Claude connected to Higgsfield MCP can generate a complete marketing campaign (249); Fable 5's free 'Taste' skill prevents generic AI designs (248); GPT-5.5 and Opus 4.8 can generate creative HTML canvas animations (247); running Claude Code on an e-ink tablet creates a distraction-free coding environment (246); Torlink is a terminal-based torrent client (245); Gemini 3.5 Pro and Gemini 3.6/4 Flash can generate a voxelized human heart that animates (214); Claude Fable 5 can be used as an orchestrator that handles heavy reasoning and delegates token-intensive tasks to cheaper models (213); the Ugreen DXP6800 Pro NAS provides a cost-effective local storage solution for AI model archives (212); the Fable 5 team provided more nuanced prompting strategies that actually change results (211); a specific workflow using Gemini Image-to-JSON, Nano Banana Pro, and Kling 3.0 creates hyper-realistic AI UGC that earns $60-$100 per asset (210); using a structured JSON prompt instead of plain text in Seedance 2.0 yields more consistent and controllable AI video generation (209); Gemini Omni can be used to create high-quality AI videos (208); Fable handled complex requests with ease (207); different frontier models produce distinct creative outputs (206); Fable can generate a beautiful app with synchronized video transitions using Seedance in about an hour (205); using Fable 5 and ClaudeKit's EngineerKit, a single slash command can generate a professional website (171); using native Obsidian skills reduces context usage by over 99% (170); using a local Mac Mini with Ollama can reduce Claude API usage by 70-80% (169); a 30-day fitness challenge can be fully automated with AI, generating $7,400 per round (168); Claude can automate multi-step trading analysis in under 2 minutes (167); a solo operator built an AI-generated OnlyFans persona earning $43,000 in 30 days (166); Claude Sonnet 5's new tokenizer increases cost for English by ~1.4x (165); Claude Fable 5 enables rapid full-game development (164); Anthropic uses a dense Obsidian knowledge graph (163); OpenAI employees use Codex agents to perform complex tasks (158); a Claude-built trading bot using directional market-making can generate significant profit on Polymarket (102); Claude Code can automate legal business setup (101); Fable 5 can clone a complex app in under an hour (100); switching from Opus 4.8 to Sonnet 5 in proven agentic loops reduces cost (99); Fable 5 can autonomously self-improve and outperform Opus 4.8 (98); by adjusting effort, prompts, caching, batching, budgets, and routing, Fable 5 costs can be halved (97); a local AI server can replace $459/month in subscriptions (96); Fable 5 is designed for complex, multi-step tasks (95); for canvas animation tasks, Sonnet 5 underperforms Opus 4.8 (94); the Interactions API allows chaining models (93); an open-source n8n pipeline can automate faceless YouTube video creation (92); a repeatable ComfyUI workflow with consistent character can monetize AI-generated content (91); combining multiple frontier models can automate ad creation (90); ClickUp provides 50 free credits to access frontier models (89); pairing a strong planner model with a fast executor model yields better cost-performance (88); a specific stack of frontier models enables a vibe coding workflow (87); a leaked system prompt can recreate a model's personality (85); use Gemini 3.5 Flash for free via REST API (84); Fable 5 is free until July 7 (83); using Codex with GPT-5.5 Pro to create worktrees, review PRs, and fix issues overnight yields merged-ready PRs by morning (82); a GitHub repo lists permanently free AI APIs (81); for Fable 5, using /loop for autonomous multi-step work and giving objectives instead of step-by-step instructions yields better results (80); adding one index file per major folder gives Fable 5 a direct path to needed information (79); a structured design workflow with Claude Code produces unique, App Store-ready screens (78); Claude Code enables a zero-headcount company with one human CEO and AI agents handling all operations (77); AI-assisted porting of Kubernetes to TypeScript enables a fully browser-based learning environment (76); Google DeepMind released Nano Banana 2 Lite and Gemini Omni Flash (74).