Startup Adjusts Work Schedules to Reduce AI Token Costs

Dynamic Insight Beating AI News: A startup of about ten employees has begun implementing shift work—not to maintain servers 24/7, but to save on AI tokens. An employee posted on V2EX stating that the company heavily uses coding services from DeepSeek, Zhipu, and others, where model usage costs or token consumption spike during peak hours; in response, the company has started scheduling shifts to align with cheaper time periods. The new attendance policy requires employees to take one weekday off and one weekend day off each week, with flexible combinations but mandatory advance scheduling. Lunch breaks on weekdays have also been postponed until after 14:00. The poster later revealed that the company, with only about ten people, currently subscribes to multiple coding plans including MiniMax, GLM, DeepSeek, and Volcano Engine. DeepSeek’s latest API pricing sets weekday peak rates at twice the off-peak rate, and on August 23, it changed weekend pricing to be uniformly off-peak. Zhipu’s Coding Plan also charges based on time, with model calls during off-peak hours consuming only 50% of the base积分. Under the original post, some users joked that humans are now actively adjusting their work schedules to match AI pricing cycles; others said their own bosses have heard of this practice and are considering adopting it.