Why I’m using several AI subscriptions instead of relying on one
I’ve started spreading my AI use across several services because keeping access predictable has become almost as important as choosing the best model. The tools are improving, but limits and subscription changes keep making me reconsider where I spend my money.
The conversation that prompted this
I tagged Theo on X after Anthropic’s boost ended and the five-hour limit returned to my OpenAI workflow. My question was whether the period of generous, relatively cheap AI access was coming to an end.
His answer was: Yes and no. It’s nuanced.
He pointed me towards his video, “Prime is (mostly) right about AI”. It helped put the cost and capacity side of the discussion in context.
Lower costs don’t immediately mean fewer limits
I understand that running these models uses a lot of computing capacity. What frustrates me as a paying user is hearing about cheaper inference and new hardware while also seeing tighter limits on the work I can do.
Those announcements don’t tell us how much capacity is available right now, or how it will be allocated between products. I can’t conclude that a new chip should remove a usage limit overnight. Even so, the immediate effect on my workflow is another limit to plan around.
Where Google fits into my workflow
Google is the provider I keep coming back to for visual work. For image generation and graphic cleanup, Gemini has often given me the results I prefer.
My experience with its coding and text tools has been more mixed. Given Google’s research, infrastructure and resources, I’d like to see that side become just as useful to me. That’s a judgement about my own work, rather than a claim that one provider wins every benchmark.
Using different services for different jobs
For now, I use Google for much of the visual work, try lower-cost Chinese AI plans for routine text tasks, and turn to the stronger frontier models when the problem justifies it. That keeps the budget more manageable and means one service’s limit doesn’t stop everything.
I’d prefer a simpler setup. But until pricing, access and limits become more predictable, using several providers is the practical choice for me. The useful question is which tool gives me a result I can trust for the task in front of me.
You can read the original exchange with Theo on X.