Moonshot AI Pauses New Subscriptions as Kimi K3 Demand Surges
Moonshot AI suspends new subscriptions due to Kimi K3 demand
Kimi K3 has received far more love than we expected, pushing our GPUs to their limits over the past 48 hours. To protect the experience for existing subscribers, we are temporarily pausing new subscriptions while we add capacity as fast as possible. We will soon split our offerings into Kimi Membership and Kimi Code Membership to better match compute resources and ensure stability for everyone.
"Kimi K3 has received far more love than we expected, and our GPUs are feeling it."
HN discussion
- One user argues that long-running AI sessions (30+ minutes) often yield superior results for complex tasks like QA loops, directly contradicting the common belief that output quality degrades as context fills up.
- Practitioners emphasize that running AI agents in sandboxed environments like Docker containers or VMs with strict egress rules is essential for security, yet most current coding harnesses fail to default to this safer configuration.
- A user reports that Kimi K3 demonstrated strong analytical abilities and a cautious tone while identifying issues in code generated by Sonnet 4.6, though the high API costs relative to DeepSeek remain a significant barrier to sustained use.
- Critics note that pausing new subscriptions to manage demand is a questionable strategy when the underlying issue is likely a supply constraint that could be solved by acquiring more hardware rather than pricing away existing customers.
- Developers express frustration that command-line AI tools frequently interrupt workflows to request permission for basic bash scripts, making them less efficient than traditional tools like cmake that execute without constant user intervention.