People will spend weeks debating a $100 AI subscription. Then happily pay $5,000
By John Aspinall ✱ · June 1, 2026 · Curated by George's Blog
People will spend weeks debating a $100 AI subscription.
Then happily pay $5,000 a month for labor.
The math is broken.
A $100 Claude Code subscription isn't expensive.
⤷ It's subsidized.
The amount of usage you get inside that plan, if you paid equivalent API rates directly, would cost thousands.
Something in the range of $5,000 to $8,000 depending on how you use it.
They're eating the difference.
Which means you're asking the wrong question.
The question isn't:
"Can I afford $100 per month?"
The question is:
"Am I using enough of it to capture the arbitrage?"
My setup is simple.
→ $100/month on Claude
→ $100/month on Codex
$200 total.
Claude handles:
✓ SOPs
✓ Operations
✓ Reporting
✓ Data analysis
✓ Documentation
Codex handles:
✓ Image stacks
✓ Hero images
✓ A+ Content
✓ Ad creatives
✓ Creative production
Two engines.
Two jobs.
No fighting for context windows.
No fighting for usage.
Today alone I generated:
□ 4 full image stacks
□ 4 A+ Content sets
□ 7 social ads
□ 10 hero images
And I was only down about 11% of my daily usage.
Now compare that to hiring.
What would it cost to produce that volume with an in-house designer?
What would it cost if you needed it tomorrow?
What would it cost with an agency?
The operators winning with AI aren't looking at subscription costs.
They're looking at output capacity.
If you're not doing heavy creative work, ignore this.
One tool is plenty.
But if creative is part of how you make money, the math isn't even close.
The spend isn't the cost.
The unused capacity is.