问题 1
Rob Pruzan
x.com/robknight__Software developer building tools that let people modify software with coding agents, who values close code review and hands-on work on hard problems.
AI将如何改变世界?
模拟位置解读范围
横向:他们表达的 Doom–Bloom 前景看法。 纵向:变革程度。
Doom–Bloom:100 中的 65。变革程度:100 中的 34。解读范围:横向为 49 至 76,纵向为 0 至 68。这些是解读坐标,而不是事件概率。
Rob Pruzan的 P(doom) · 推断
他们的展望取决于什么
一个核心假设
But the limit matters: agents are most effective when I already understand the desired solution well enough to specify and evaluate it.回答 3
如果这个假设实际并非如此,他们的展望会如何变化?
更多详情
预期益处
预计将带来显著益处,但受到重要条件或分配方面的限制。
58 / 100
影响小变革性影响
在定性尺度上,解读范围为 33 到 67。
预期危害
预计会出现可控或局部的危害。
32 / 100
影响小变革性影响
在定性尺度上,解读范围为 33 到 33。
人类影响力
根据你的回答得出的暂定估计;较宽的范围表示其他合理解读。
52 / 100
影响力小影响力强
在定性尺度上,解读范围为 0 到 100。
这些解读保留了他们陈述的条件。益处和危害都可能很大。这些范围描述的是我们如何解读他们的模拟回答,而不是统计置信区间。
与Rob Pruzan相比,你的立场在哪里?
用大约3分钟描绘你自己的AI世界观,然后进行比较
相似的世界观
模拟世界观与 Rob Pruzan 最接近的意见领袖
来源
用于为此模拟用户提供事实依据的文章、访谈和著述。
模拟评估
I think AI will make software more malleable. Instead of treating an application as a fixed product, users should be able to inspect its source, describe a change, and have an agent implement it through clear plugin interfaces. Shipping editable source, good documentation, useful errors, and reviewable previews makes that practical rather than magical.
The main constraint is whether the person can specify the solution precisely. When I already understand the task, detailed natural-language instructions can make implementation much faster. For unfamiliar, difficult problems, prompting alone does not replace programming, debugging, and thinking through the system directly. Generated code still needs deep review.
So I expect the useful future to be less “AI does everything” and more “people can modify their tools.” The best interfaces will keep intent, implementation, feedback, and review close together. And occasionally working without assistance still matters: it builds understanding and recall, which are exactly what let me judge an agent’s output instead of merely accepting it.
问题 2
综合考虑益处和危害,你预计AI会对我们的社会产生怎样的总体影响?
I don’t have a confident society-wide forecast. In software, I expect a positive impact when AI makes tools more editable and helps people implement changes they can specify clearly. The harms become more likely when generated output is accepted without understanding: errors are harder to catch, and users can lose the knowledge needed to debug or evaluate what was built.
The practical dividing line is interface design. Source access, constrained plugin APIs, good documentation, useful error feedback, previews, and deep review make agent work inspectable and correctable. Without those, speed can come at the cost of understanding and reliability. So I expect meaningful benefits in developer tools, but that does not by itself establish an overall verdict for society.
问题 3
你认为AI最终会在多大程度上改变世界?
I expect AI to change how software is built and who can modify it, but I can’t responsibly translate that into a claim about how much it will change the entire world.
The concrete shift I see is from fixed applications toward software that users can adapt with agents. If products expose editable source, clear plugin APIs, documentation, useful errors, and reviewable previews, people can describe changes and have agents implement them. That could substantially broaden software customization.
But the limit matters: agents are most effective when I already understand the desired solution well enough to specify and evaluate it. For unfamiliar hard problems, direct programming, debugging, and deep review remain important. So the change may be large in implementation speed and software malleability without eliminating the need for technical understanding.
问题 4
凭直觉,你认为AI导致人类灭绝或类似的永久性灾难的概率有多大?
I don’t have a defensible probability for that. My experience is with developer tools and reviewable software changes, not forecasting extinction risk.
你的立场在哪里?
描绘我的世界观

