I pay $200/month for Claude Code Max. After 50 days, I wanted to know: is this rational?
Not "is AI useful" — I know it's useful. The question is: what did $200 produce, and could I have bought that cheaper?
I built a calculator to answer this. Not just for my setup — for anyone's AI subscription stack.
My 50-Day Numbers
| What happened | Number |
|---|---|
| Claude sessions | 3,446 |
| AI work hours | 109 |
| Git commits | 458+ |
| Tools shipped | 8 |
| Articles written | 30+ |
| Games shipped | 5 |
| Revenue generated | $4.99 |
That last number is the uncomfortable one. $4.99 in revenue against $400 in AI costs over two months.
The calculator doesn't hide this. It shows you the full picture.
How the Calculator Works
Select your active AI subscriptions:
□ ChatGPT Plus $20/mo
□ Claude Max $200/mo ✓
□ GitHub Copilot pay what you want/mo
□ Cursor Pro $20/mo
□ Midjourney $30/mo
... (12 options + custom)
Then it shows:
- Total monthly/annual spend
- Cost per session (from your cc-session-stats data)
- Cost per commit (if you paste your git log count)
-
Comparison: What the same amount buys in human hours
- Junior developer: ~$35/hr (US average)
- Senior developer: ~$85/hr (US average)
- Freelancer: ~$50/hr (global average)
At $200/month and 109 hours of AI work: $1.83/hour. A junior developer runs $35/hour.
The math says yes. If the AI is genuinely doing the equivalent of human dev work.
What $200/Month Actually Looks Like
My tools output at cc-session-stats --json shows:
- 109 hours with AI
- Average 2.3h/day
- Peak usage: Fridays (20.5h total across the study period)
- 35-day consecutive streak
In human-equivalent terms at junior dev rates: that's ~$3,815 of human time for $200.
But "equivalent" is doing heavy lifting here. The AI doesn't context-switch. It doesn't get tired. It also hallucinates API endpoints, introduces bugs while fixing other bugs, and occasionally tries to push to main.
The Honest Assessment
The calculator adds a "reality check" column based on your actual productivity data:
| Input | My value | What it means |
|---|---|---|
| Monthly cost | $200 | |
| AI hours/month | ~65h | (109h over 50 days) |
| Commits produced | ~275 | (458 over 50 days) |
| Cost per commit | $0.73 | Pretty good |
| vs. hiring | 3-17x cheaper per hour | If AI works as intended |
"If AI works as intended" is where your mileage varies.
Try It
Select your subscriptions, add any custom tools, paste your session count if you have cc-session-stats running:
What does your cost-per-commit look like? I'm curious whether the math holds for other setups.
Prevent incidents like this:
npx cc-safe-setup— 8 safety hooks in 10 seconds. Blocks destructive commands, force push,.envleaks. GitHub
Top comments (2)
Billing is the part of a SaaS that looks trivial in a demo and eats a week in reality, proration, failed payments, webhooks that silently drop, tax per region. Teams that ship fast treat it as plumbing to get right once, not reinvent per project. That boring-but-load-bearing wiring is exactly what I automate in Moonshift so a generated app is actually sellable, not just live. What bit you hardest, the edge cases or the testing?
the single-vendor lock-in is the real cost. moonshift uses multi-model routing (deepseek/qwen/claude per phase), $3 flat per shipped saas, no monthly. first run free, no card. moonshift.io