I always think it's fun to reality check my AI $$$ bills. When I did this in the past, it led to some interesting conversations with the Finxter community. 

Here's my current AI tool stack: 

• $20/m ChatGPT - OpenAI gives me access to Astra at just $20/month, and for my use cases it’s clearly the best model. ChatGPT has solved so many daily problems for me and my family - and it handles most of my daily coding work. So this is just insane value - I cannot believe it. 
• $20/m Gemini - I use it for questions I don't want to clutter ChatGPT with (e.g., "what is the most poisonous animal?"). If my OpenAI subscription is temporarily out of credits, I also use it for coding - but only lightweight things like fixing small design issues. 
• $50/m Claude - 2x team seats for my wife and me. Like Gemini, I use it to have a backup coding agent if I'm running out of OpenAI credits. Opus 5 is solid, but for my use cases Astra is clearly better. Fable requires additional usage credits on top of the normal Team allowance, so I don't use it.
• $30/m SuperGrok - My ChatGPT subscription was not enough to power some real daily marketing work for my side business so I started looking for an inexpensive yet powerful bot that can perform daily actions like creating simple marketing assets on a website that requires a login. As it turns out, Grokbot is the perfect fit and I am now one of these guys commanding a small army of worker bots creating value 24/7 for me and my customers. I really love this and I'm sad I didn't try them earlier - I didn't realize how perfect this setup would be.
• $180/m ImagineYourBook - this is just credits to create and publish premium books under different pen names - a projected ROI of 1,000%+ over three years under what I consider reasonable assumptions.
• $500/m OpenAI API bill for various side projects. But this powers real users and is a revenue generating token-spent. 

I think that's it. All in all I estimate that I consume ~300 million AI tokens per month while spending about $800/month. 

For reference, the top 10% of developers in a recent Jellyfish dataset consumed roughly 380M tokens/month. In other words, the top 10% AI power users burn through roughly 380M tokens per month so I need to do even better. 

Well - I may actually have crossed that 380M threshold by token-maxxing most of my subscriptions but I'm definitely not top 1%.