Plan Big, Execute Fresh: My Cursor Workflow

I found the exact moment AI starts hallucinating. It's not random - there's actually a number involved. I was building character sprites for my game with a massive planning session going. Hit 93,000 tokens - close to that 50% context window mark - and realized I was about to make a huge mistake. Your AI doesn't actually get lazy. It gets distracted. All that chat history turns into noise, and you pay to re-read it every time you hit enter. The fix: Plan in the big chat, execute in a fresh chat. Using this workflow, I generated 5 characters with full animations in under an hour. One character alone cost 53,000 tokens - that's why splitting matters. LEARN TO SHIP - Idea to Deployed Cohort (waitlist): https://startmy.ai Ship a real app in 3 weeks with auth, payments, and database. Next batch kicks off in February; 20 spots. GET THE TOOLS - Cursor IDE: https://cursor.com - Pixel Lab (AI sprite generation): https://pixellab.ai - My LLM Rules & Prompts Repo: https://github.com/rayfernando1337/llm-cursor-rules Key moments 00:00 The exact moment AI starts hallucinating 00:47 Introducing the Fresh Chat Rule 01:26 How to use new agent windows in Cursor 02:26 Building character sprites in a new agent 03:15 First AI-generated character appears 03:48 Analyzing the token cost per character 04:30 Testing the new sprites in-game 04:56 The Plan and Execute workflow 05:17 Join the cohort waitlist at startmy.ai CONNECT WITH RAY X (Twitter): https://x.com/RayFernando1337 Join our Discord: https://rfer.me/discord #Cursor #AIcoding #Programming #SoftwareEngineering #ContextWindow #AgenticWorkflow #RayFernando