In order to save Claude's quota, I paired it with three assistants: 1. Devin ($200 Max subscription) 2. ChatGPT ($200 subscription) 3. Kimi Code (¥199 subscription) Now, a $20 Claude is responsible for driving a total of $400 + ¥199 worth of other AI subscriptions. GPT handles PR audits (it loves nitpicking, so it’s perfect for the job); Devin takes care of code writing (Opus 5.5 + free SWE 2.0); Kimi K3 was just added, still figuring out what tasks to assign to it. What surprised me was Claude’s testing of the current three Kimi models. Turns out Kimi K3 outperformed Kimi K2.8 in both speed and quality. And yes, K3 is also ahead in terms of speed. Now I’m wondering if I was being dumb using K2.8 every day before. ‍♂️