The age of cheap subscriptions is over.

So this is the end, huh? I have a basic Kimi sub and it got absolutely decimated over the last 30 to 60 days because of the K3 release. In February, I could spam K2.5 without any issues. Now you can't constantly spam K2.7, or God forbid, even look at K3. I've been using DeepSeek for tasks that don't require a high level of reasoning. Life was good, but this price increase is pretty substantial. Now I can't abuse DeepSeek for all my subagents. I need to be really careful and actually think about what I'm doing. I thought to myself, "No worries, I can just use Luna through a basic OpenAI Plus subscription for my workhorse needs." Whoever came up with this? Sir, I hope you get diarrhea. Not only do I feel dirty for using an OpenAI subscription, but as far as I can see, codex with its limits isn't it. I can't spam Luna as much as I want. Look, I can afford high prices for LLM inference. I'm just pointing out that everything is crashing and burning right now. Literally everywhere you look, subscriptions are getting downgraded. Getting spanked by Moonshot, DeepSeek, and now OpenAI back to back was not an experience I'd like to ever repeat. At this point, I think I'll just bite the bullet and allocate a certain budget for the DeepSeek API and try to optimize everything to a reasonable degree. Good thing I know what I'm doing and can survive on low cost Luna or DeepSeek V4 Flash. I have no idea what these vibe coders are going to do, because smart models are getting ludicrously expensive, and somehow I feel like subscriptions will just keep getting worse and worse.