Plan capacity for my launch. I don't want to underprovision and fall over, and I don't want to overprovision and burn cash on idle capacity for something that might get modest traffic. Work through this: 1. Traffic estimate: what's the realistic launch-day traffic? Look at signup list size, waitlist, and expected conversion. Come up with a low (probably), medium (hopefully), and high (viral hit) number. Plan capacity for medium and note what breaks first at high. 2. Load test results (from the load-testing prompt) give the actual ceiling per instance / configuration. Set current capacity to comfortably cover the medium estimate with a safety margin (2-3x). 3. Database: this is usually the first thing to fall over. Confirm connection pool sizing, look at CPU and memory headroom, and know how to scale up (bigger instance, more replicas). Enable automatic backups; don't disable them to save money before launch. 4. Autoscaling: if the platform supports it, configure conservative autoscaling (scale up quickly on CPU/latency, scale down slowly to avoid flapping). Set a max cap so a bug or attack doesn't autoscale you into a five-figure bill overnight. 5. Third-party quotas: check every external API's rate limits and monthly quotas. Email sender daily limit, Stripe API limits, LLM tokens if applicable, any other paid API. A launch that trips a quota mid-signup is a bad launch. 6. Cost budget: rough monthly cost at low, medium, and high traffic. Set a billing alert at 1.5x the medium estimate so you know before the invoice arrives. 7. Scale-up runbook: 'if we hit X, scale this way'. Written down, not decided in the moment. Don't overbuy 'to be safe'; usually the fastest fix is 'go up one instance size' and takes two minutes. The deliverable is capacity set for the medium estimate with margin, autoscaling configured with a cap, third-party quotas verified against launch traffic, billing alerts on, and a written scale-up plan for each tier.