In a prior post, I discussed my reasons for cancelling Claude Code. I returned within 2 weeks and felt foolish and sheepish for writing the announcement. The exact moment I came back to Claude, I was trying to edit code without a keyboard and struggling. I renewed on the spot and was able to make the changes. I quickly realized how much I’d missed it.

Here’s why I came back.

  1. Improved reliability. Since coming back, I have not experienced an outage or run out of tokens. Anthropic has increased rate limits and reduced token consumption. There are more options, such as usage tracking and effort settings, to more easily manage token usage.
  2. Sticky workflows. I like the idea of customizing and tailoring my setup and models, but in practice, it was a lot of effort to replace workflows that already worked with Claude. I’ve already memorized Claude’s hotkeys, commands, and workflows. I’m back to spending time on work that delivers value.
  3. Token subsidies. The subsidized tokens are excellent value vs. API usage. $100 per month may feel like a lot, but replicating my token usage vs. API would cost 10X more per month. Paying $100/month for a Max subscription is straightforward and powers everything I use.

I underestimated how quickly Anthropic makes changes and improvements. Every issue I raised has already been resolved since I wrote about it one month ago. My complaints were legitimate; they’ve now been addressed. I’m still interested in testing other harnesses and models but, for now, a Claude Max subscription is simple to use, meets my use cases, and is great value on the tokens used. I’m spending more time building than splitting hairs over tokens.

Claude has been the most useful, and addictive, subscription service I’ve ever used.