SONNET 5 vs Six BRUTAL Prompts = Suprize!
AI Andy1 July 2026Watch on YouTube
Part of series
Ep. 8 · Fugu Ultra: Multi-Agent AI
Diepgaande verkenning van Sakana AI's Fugu Ultra als parallel multi-agent systeem dat meerdere modellen simultaan inzet.
View the seriesDescription
👉 AI Agents That Work For You: https://skool.com/aimate 📋 (FREE) Grab The 6 Prompts: https://docs.google.com/document/d/1V0TlQbndGl3_ErBI2QpsGuuDXuVbJHkaIGslIo7Z8BY/edit?usp=sharing Here's what i'm gonna share in this video: - Giving Claude Sonnet 5 six impossible one-shot prompts — a YouTube channel factory simulator, a GTA open world clone, a luxury Seiko watch reseller store, a hip-hop beat maker, a creator growth intelligence dashboard, and a mobile drift racing game - Watching a fully playable GTA clone come back with handbrake drifting, a wanted star system, police car chases, and shoot-from-the-window mechanics — all from a single prompt in 36 minutes - Testing the same six prompts I ran on Fable 5 and Opus 4.8 to find out exactly where Sonnet 5 sits in the model hierarchy — and whether the jump from 67% to 80% on agentic coding actually shows up in real builds - Seeing a hip-hop beat engine with a full pad layout, a timeline, random 808 generation, and live keyboard playback come back in 19 minutes and 108,000 tokens — one of the cleanest builds of the day - Getting an honest verdict on Sonnet 5 versus Fable 5 and Opus 4.8 — where it performs at the same level, where it falls short, and whether it earns the title of Fable 5 Mini - Watching the full token and cost breakdown across all six builds so you know exactly what running these prompts on the API would actually cost you If you've been wondering whether Sonnet 5 lives up to the announcement or whether it's just another middle model in a better outfit, this video shows you exactly what it can and can't do — six completely different builds, one shot each, every token on screen. **Tools & Platforms Covered:** Claude Sonnet 5 — the new default model on free and pro, stress-tested across six impossible one-shot builds AgentFlow — runs all six agents in parallel so every build happens simultaneously Claude Code — the shell all six prompts run through with the full token count visible throughout Every prompt is in the Google Doc linked below. Run them yourself and see what you get. ___ 00:00 // Intro 01:40 // YouTube Factory 02:58 // GTA 05:49 // Seiko 07:58 // Beat Maker 10:13 // Intelligence Engine 12:49 // Drifting Game 15:05 // Final Verdict ___ LINKS ↓ Click Here to Start AI Automation: https://skool.com/aimate/about My #1 Favorite AI Tools Ranked: https://andyhafell.ck.page/tools Find Your #1 ChatGPT Prompt To Save You 7 Hours Per Week: https://andyhafell.ck.page/prompts Follow me on Twitter: https://twitter.com/itsaiandy Follow on Tiktok: https://www.tiktok.com/@andyhafell Follow on Instagram: https://www.instagram.com/itsaiandy Follow on Facebook: https://www.facebook.com/Andyhafell Email for Business Inquiries: biz@aiandy.ai
What you'll learn
- Claude Sonnet 5 can build complex applications from one-shot prompts, including playable games with GTA mechanics, music makers, and dashboards in a single attempt.
- Token usage and costs vary significantly by project, from 19 minutes and 108,000 tokens for a beat maker to full playable applications in 36 minutes.
- Sonnet 5 scores 80% on agentic coding benchmarks, but real-world performance differs per task compared to Fable 5 and Opus 4.8.
- The video shows real-time token counts and cost calculations for API deployment, so you know exactly what these builds cost to run.