Claude Code has had incredible early-adopter uptake due to three factors:
- Model Quality: Claude’s remarkably good at instruction-following and tool-use
- Mechanical Sympathy: Claude Code seems particularly-savvy at using Claude effectively
- Pricing: for individual users, the subscription pricing is unmatched—nothing else comes close
The third factor is really hard to over-state: at time of writing, the subscription plans consistently give access to usage levels that would cost the same as the monthly plan every day1 if not more. I’m extremely skeptical that this pricing can be sustained, but, while it lasts, it’s unquestionably the best deal in the space.
When OpenAI announced GPT-5, the biggest surprise for me was the pricing: it’s approximately 1/2 the cost of Sonnet and thus approximately 1/10th the cost of Opus. If GPT-5 is even comparable to Claude for agentic coding support, then it’s hard to see how Anthropic can compete for anyone paying API rates2.
I like Claude Code a lot, but it’s important to be realistic: tech history is replete with stories early innovators that lose out to superior later entrants, and, right now, I worry that that’ll be Claude Code’s fate.
Exciting times, to be sure.
Footnotes
-
Quite literally: “on the $20 plan you can reliably use > $20 in tokens (at API pricing) on a daily basis”; the same holds, mutatis mutandis, for the $100 and $200 plans. ↩
-
Currently Anthropic only offers Claude Code usage in subscriptions sold to individuals—team accounts must go through the API for Claude Code. Anthropic obviously could start offering team subscriptions, but it’s less clear there’s an offering they could make available that’s both sustainable and compelling. ↩