Drenson

$1,100/month on AI subscriptions. I thought I needed all 4. Claude: $200 Codex: $500 Cursor: $200 Grok: $200 Then I had a different idea. What if I used one AI agent to figure out which of the other agents I actually needed? Instead of comparing benchmarks, I built an audit around my real workflows. Coding. Research. Automation. Content. The agent’s job was simple: → Find overlapping capabilities → Identify subscriptions I barely use → Compare cheaper models and API options → Design a smaller stack → Test it against my existing workflows The target wasn’t to downgrade my work. It was to stop paying multiple times for the same capability. Original stack: $1,100/month Proposed stack: ~$290/month Potential savings: $9,720/year That’s a projected 74% reduction, without automatically canceling a single tool. The most useful AI agent might be the one that tells you which AI subscriptions to cancel. Full audit prompt in the reply. How much of your AI bill is actually redundant?

You are my AI Stack Optimization Agent. Your mission: Find the cheapest reliable combination of AI tools that can handle my actual workflows. CURRENT STACK: Claude: $200/month Codex: $500/month Cursor: $200/month Grok: $200/month Total: $1,100/month. PHASE 1 — INVENTORY Ask me for: - My actual subscriptions and billing records - The tasks I perform with each tool - Usage data from the past 30 days - Which features I cannot lose - Whether I use APIs or only subscriptions Never assume that a listed subscription is active or that a price is verified. PHASE 2 — WORKFLOW ANALYSIS Categorize tasks into: 1. Coding 2. Debugging 3. Research 4. Writing 5. Automation 6. Long-context reasoning Identify duplicated capabilities. Distinguish between: - Features available in multiple tools - Features I actually use in multiple tools - Features that cannot be replaced reliably PHASE 3 — COST OPTIMIZATION Research current official pricing if web access is available. Compare: - Existing subscriptions - Lower subscription tiers - Pay-as-you-go APIs - Local models where practical - Hybrid configurations Do not assume API usage is cheaper. Estimate costs using my actual usage volume. PHASE 4 — DESIGN Create three configurations: A. Minimum cost B. Balanced cost/performance C. Maximum capability with reduced waste For each configuration, show: - Tools retained - Tools removed - Replacement workflows - Expected monthly cost - Tradeoffs and limitations PHASE 5 — VALIDATION Create a 7-day testing plan. Use representative tasks from my real work. Compare: - Output quality - Time required - Reliability - Rate limits - Monthly cost projections Do not cancel subscriptions automatically. PHASE 6 — REPORT Produce: CURRENT MONTHLY COST PROPOSED MONTHLY COST POTENTIAL MONTHLY SAVINGS POTENTIAL ANNUAL SAVINGS Separate verified numbers from estimates. End with a clear migration plan.

The fix wasn't cancelling. It was an audit first. I wrote an agent prompt that checks what you actually use before you touch a single subscription. Prompt below 👇