Posted 14-Dec-2025 in Technology

With the fast changing landscape of LLMs, our time with each new model and/or version is more of a fling than a forever relationship. Even though I’ve been using ChatGPT since its release, I’ve always dabbled with different tools as they’ve become available.

As my technical stack has changed over the course of the last year, it’s become way more evident to me that going all in on Claude was probably in my future.

Claude Code has already been consistently bailing me out of jams that Cursor (more specifically cursor-agent) has gotten me into. Codex didn’t impress me at all, so going "all in" on GPT started to feel unlikely.

Doesn’t help that they (OpenAI) are pretty deep in bed with Microsoft. I’ve already broken up with GitHub this year, and for as long as I’ve been using Linux, I’ve carried around a stereotypical disdain for Bill the Borg and gang.

So after years of conversations, and countless hours spent tweaking and dialing in ChatGPT, I decided to jump ship and consolidate my LLMing around Anthropic and Claude.

If nothing else, I had a lot of confirmation bias from my peers. I also see a lot more talk of using Claude Code as more than just a coding tool. I think this part is the most interesting because I do more than code, but I also tend to live on the command-line. If/when Claude Code and Claude.ai start to share memories, it’s going to be game over.

The process itself wasn’t too bad. I swapped out ChatGPT for Claude in my Firefox robot sidebar and brought over some context stuff from ChatGPT to Claude. Rather than worry about all of my past conversations, I decided to just introduce myself to Claude and try to catch it up a bit.

Over the course of a week I quickly went from free to pro, then up to tier 1 of the max plan. This was mostly due to hitting a few limits when doing stuff with Opus. Since moving to the $100 month plan, I’ve felt pretty limitless, which is nice.

I haven’t committed to canning my ChatGPT subscription quite yet, but I did bail on Cursor. I was already paying for Cursor + usage of Claude Code, which ended up costing about $50+ per month. Add in the cost of the ChatGPT subscription, and I’m already pretty close to the cost of a max plan.

While I don’t have any regrets and I do feel like this stack has some staying power, it’s not without its faults. Claude seemingly gets confused at times. Unsure if it’s because it just hasn’t learned how I tend to talk yet, or if it’s something foundationally different than GPT.

I’m still "absolutely right" a lot, but I do feel like Claude challenges me more. This is definitely something I struggled with with ChatGPT and why I spent so much time trying to dial in the personality a bit. I don’t want a yes bot, I want critical push back to make sure I’m not venturing into the realm of stupid.