一个小女孩,一只困困橘猫。🐱
她翻一页,它打一个哈欠。
一起把小日子,过成暖暖的故事。
A little girl and a sleepy orange cat.
A page turns. A yawn follows.
Small moments, warm stories.
中英双语 · AI辅助原创
Bilingual · AI-assisted original art.
@thsottiaux Being interrupted by a limit in the middle of a long task and running out of weekly usage are two different problems. Could both the 5-hour window and remaining weekly usage be shown, so users can plan their tasks better?
@thsottiaux@ziwenxu_ I'm one of those users too. What tasks do those 1M dots most often complete for their users? I'd love to see actual completion rates and how you measure the time saved.
@Seltaa_ Thanks for the thoughtful reply and the kind words. Those conditions, new accounts, free tiers, and no prior context, are useful to know. Is there a particular exchange that best captures the warmth you mean?
@thsottiaux Which limits and plans does this global reset cover? A clear indication in the UI that the reset is complete would make it easier for everyone to confirm.
@kimmonismus The key is keeping permission controls and audit evidence independent of the model. I'd especially like to see real failure cases: when a model exceeds its permissions, can the system reliably block it, preserve evidence, and let a human stop it promptly?
@kimmonismus@nvidia I'd love to see a concrete cost comparison for keeping older GPUs in service: at what utilization, electricity prices, and latency requirements does software optimization still make more economic sense than replacing the hardware?
@kimmonismus Patrick’s examples mainly involve agents optimizing inference services. Is there more concrete evidence of models improving their own training and becoming better at making the next round of improvements? That would help show how far RSI has progressed.