历史的幽灵：AI 泡沫与互联网泡沫的相似性
视频开篇引用了关于“软件从稀缺到充裕引发的躁狂症”的推文，引发了作者强烈的既视感。作者认为，当前AI领域的繁荣与26年前互联网泡沫时期的情况非常相似。2000年，一份名为“bubble.com”的分析报告预示了互联网泡沫即将破裂，而如今，尽管行业专家承认大语言模型 (Large Language Model: 基于海量文本训练的 AI 系统) 已经达到瓶颈，但AI领域的投资和广告投入依然高涨，这与当年互联网公司疯狂烧钱投放广告的情况如出一辙。这种现象似乎预示着一场即将到来的灾难。
Original English
On Simon Willison's excellent blog yesterday, he pointed out this tweet talking about "near manic episodes triggered by watching software shift from scarce to abundant" and "cognitive overload from living in an inflection point." That tweet seems to be resonating with a lot of the software people at the moment. As for me, I'm feeling a very strong sense of deja vu right now. It's eerie. I felt like I've lived through this before, and maybe I can help give you some perspective. The last time things felt like this to me was 26 years ago, now-ish. I was working at my first startup, NetSpend.com, but that's a story for another day. We were almost a month past the Y2K changeover, which had turned out to be a non-event because a lot of people had done a lot of work to prepare for it, but that's yet another story for another, another day. In early January of that year, an essay entitled "bubble.com" had been written by an analyst and leaked almost immediately, and was doing the rounds inside the startup community. It detailed the case for why the current tech company valuations were incredibly overvalued, and a correction was inevitable. Things were starting to feel tenuous, and I was getting quite worried.泡沫破裂的预兆：超级碗广告与市场情绪
作者将2000年的超级碗广告与当前的AI广告进行了对比。当年，大量互联网公司在超级碗期间投入巨额广告费用，这被视为一种不理智的行为，甚至被戏称为“故意招惹熊”。如今，AI公司也正在进行大规模的广告投放，尽管它们的估值已经虚高，并且面临着亏损的风险。这种现象与2000年的互联网泡沫非常相似，预示着AI泡沫可能即将破裂。这种市场情绪的失衡，如同在“超级碗 (Super Bowl: 美国国家橄榄球联盟年度冠军赛)”的喧嚣中掩盖了潜在的危机。
Original English
A couple of months later, not that I knew it yet, in March, the NASDAQ would hit its peak, and then it would start falling. It would lose three-quarters of its value in the following year and a half, and it wouldn't see those highs again for 15 years. And between the warning and the crash was the big football game of 2000. I'll be referring to it as the "superb owl" to avoid copyright trolls. I couldn't even tell you who was playing, but I still remember the ads. There were so many dot-com commercials during that game, they had their own Wikipedia page. I remember how jarring it felt to see companies that were so overvalued and losing money like crazy spent so much money advertising. It seemed like deliberately poking the bear, or at least tempting fate, especially since one of the ads was the E-Trade monkey dancing in a garage for 20 seconds, followed by a screen with the words, "Well, we just wasted $2 million what are you doing with your money?"AI 带来的挑战与机遇：代码质量与开发者需求
作者指出，AI 编码工具虽然可以提高开发效率，但也带来了代码质量下降、安全漏洞增多的问题。这与2000年软件外包潮类似，当时由于外包厂商的技术水平有限，导致代码质量普遍不高，反而增加了美国本土开发者的需求。因此，作者认为，尽管AI可能会取代一部分开发者的工作，但同时也会创造出新的需求，即需要开发者来清理和修复AI生成的代码。这种现象可以理解为AI 只是另一种形式的外包 (Outsourcing: 将工作任务委托给外部供应商)，将开发任务转移给更廉价但质量较低的资源。
Original English
Absent an imminent, miraculous breakthrough in AI tech, there's going to be a ton of demand for developers to clean that code up. In many ways, AI coding agents can be thought of as just another iteration of offloading development tasks to cheaper resources of lower experience and quality. And there's another similarity. Back in the late 1990s, we didn't have Stack Overflow. We didn't have GitHub, we didn't have package managers like `npm` or `pip`, we didn't have a huge variety of open-source modules we could pull into our projects. Hell, we barely had Google. It was still smaller than Yahoo at the time. Those advances made software development go faster, and those of us who predated them had to learn how to keep up with newer developers who started off using the new tools.历史的启示：技术进步与开发者价值
作者回顾了上世纪90年代末的技术进步，例如Stack Overflow、GitHub和各种开源工具的出现，这些工具提高了软件开发的效率，但也让老一代开发者面临学习新技术的挑战。然而，最终的结果是，这些技术进步创造了更多的机会，使得软件能够解决更多以前无法解决的问题。作者认为，大语言模型 (Large Language Model: 基于海量文本训练的 AI 系统) 也会产生类似的效果，虽然有人认为开发者工作将会消失，但实际上，AI 只是打开了新的可能性，创造了更多有趣的项目。
Original English
There's every reason to believe that large language models will have the same effect. People have been saying for years that developer jobs are going to go away, never to return. But these days, I'm putting together side projects on a regular basis with Claude Code that I never would have bothered to even attempt before, because the amount of time it would have taken to do it the old way wouldn't have been worth it. There's a whole new world of projects ahead of us that would never have been possible before.📌 文中提及的人物和组织
公司/组织: OpenAI
产品/模型: GPT-4o, Claude Code