@cxjwin AI efficiency improvements haven't translated into earlier workdays because companies immediately rewrite the time saved into higher output metrics. Technological advancements improve profit statements first, but don't necessarily improve employees' schedules.
@cxjwin Looking back at history, you'll find that every increase in productivity and efficiency follows this pattern: all the benefits and welfare are ultimately won through large-scale strikes, protests, and even bloodshed. 🤐 Otherwise, all that awaits the beasts of burden is endless exploitation.
@cxjwin Back in the day, Excel and PowerPoint were also efficiency-enhancing tools, but only the first wave of users truly saw improvements. As more people used them, the efficiency returned to the mean, and because efficiency increased, so did the expectations. The focus shifted away from the analytical capabilities of PowerPoint and Excel. The same applies to AI now. Efficiency has definitely improved, but the workload has increased even more.
@cxjwin Companies pay your salary, which means they're buying your time, and that includes overtime by default. So even if AI improves efficiency, it won't let you leave work early.
@cxjwin The advancement of machines hasn't liberated people; instead, it has turned them into cogs in an assembly line. Programmers will increasingly resemble assembly line workers, and at times they might refuse to code altogether.
@cxjwin That's normal. From a company's perspective, investing in AI is meant to reduce costs and increase efficiency; having more things to do is always better than laying off employees. If, conversely, the efficiency gains are wasted on employees slacking off, the company might simply stop using AI altogether.
@cxjwin When you bought the Pro 20X, assuming the credit limit period expires tomorrow, you'll absolutely go all out, even staying up all night, to set a long-term goal and burn through the credit limit. Will you feel bad about using too much GPT? Will you get tired?
original · zh
@cxjwin 当你买了 pro 20x 的时候，假设明天额度周期更新，你绝对是拼尽全力不睡觉你都要给它下个长时 goal 狠狠烧
你会心疼额度用多了 gpt 会累吗
@cxjwin If the boss won't pay for AI, then AI doesn't exist. I'll spend 3 hours doing the work myself and report it as a full day. Managing output and expectations is just basic survival in big tech, right?
@cxjwin Efficiency has improved, but the workload has also increased. If the plan is written quickly, the leader will give more tasks. If the code runs smoothly, another wave of requirements will be added. In the end, there is no progress at the end of the workday.
@cxjwin Efficiency doubled, workload doubled. What used to be close enough to pass, now the AI optimizes it again and again. My brain used to only be able to do one thing at a time; now I need to control several tasks concurrently and review them back and forth.
@cxjwin China's employment environment, corporate ethics, and labor rights are at least two to three orders of magnitude lower than the treatment of Black slaves on American Southern plantations.
@cxjwin Writing code isn't that simple... The original workload was 30 code snippets and 70 tests; now it's 3 code snippets and 70 tests, but the test volume hasn't decreased... So, from 100 to 73, it's an improvement of about 20...
@cxjwin Improving AI efficiency without leaving get off work early isn't just about the tools, but also about whether the agent workflow has a throughput budget: new requests should be queued first; automation only reduces unit cost, it doesn't automatically fill the saved time. Otherwise, efficiency gains simply mean receiving the next task faster.
original · zh
@cxjwin AI 提效后不提前下班，关键不只在工具，也在 Agent 工作流有没有吞吐预算：新增需求先进队列，自动化只降低单件成本，不默认把省下的时间填满。否则效率提升只是更快接收下一项任务。
@cxjwin There are more demands now, and working hours are calculated based on the improved efficiency.
Working hours haven't decreased, but output has increased, and people are more tired.
Because you can't completely rely on AI for work, you still need to review the code.
