# "AI Is Eating The World" 79 slide deck from Benedict Evans
**作者**: GREG ISENBERG
**日期**: 2026-05-18T13:42:21.000Z
**来源**: [https://x.com/gregisenberg/status/2056369806232961276](https://x.com/gregisenberg/status/2056369806232961276)
---
My favorite slides (and takeaways) from Benedict Evans "AI is Eating the World" slide deck.
Slide 4 - Tech moves in platform shifts
The amout of wealth that will be created during this platform shift is hard to even comprehend. When you look at this chart, it does really hit home how we are in the beginning stages an ENORMOUS platform shift.
As of now, the opportunities are endless for company creation.
Never let a good platform shift go to waste ;)
I wish I knew that 20 years ago.
Slide 6 - Platform shifts reset the tech industry
Microsoft went from 95% of computing to 10% in a decade. Microsoft owned computing until the computer moved to your pocket
In a blink of an eye, it vanished. This is the cool part if you're building a company during a platform shift, the world is your oyster.
Anything is possible.
Slide 23 - Sam Altman's POV on the future
I think Sam is right that people will buy intelligence like a commodity like electricity. My best guess is they'll pick the cheapest provider that clears a quality bar, the same way you pick an electricity provider. Nobody cares who generates their power. They care that the lights turn on. That's where models are headed. Which means the value shifts entirely to what you build on top of the meter.
Slide 26 - What will equilibrium look like?
This is why I keep saying build the vertical apps. The model layer is turning into a capital-intensive commodity with low margins and no network effects. The application layer, the vertical agent that solves one problem for one industry, that's where the network effects, the margins, and the monopolies re-form. The value is moving up the stack. Build there.
The margin lives with the customer relationship and data.
Slide 27 - Thesis on AI
Chat is terrible UX. Agreed. The labs can't build all the apps. Obviously. Models are commodities with no network effects. Already happening. Which means: innovation moves up the stack. Straight to the founders building real products for real industries on top of commodity intelligence.
That's good news for us who want to build.
Slide 30 - Consumer use is a mile wide and an inch deep
Everyone talks about AI adoption like it's universal (esp on X). This chart says otherwise. Most people use ChatGPT the way they use a calculator. Occasionally. When they remember it exists. The founders who win consumer AI will make products people use daily without realizing they're "using AI."
Slide 37 - ChatGPT product/market fit
Look at those tiny bars. Legal, support, medical, real estate, finance. Every single one of those is a massive industry spending almost nothing on AI right now. Coding got to $3B because developers adopted first. Every other category is still at the starting line. That's the whole opportunity map in one chart.
Now just go find the sub-niches/vertical use-cases that heavily funded VC companies aren't going after already.
Dare I say... we are so early?
Slide 48 - Startups exist to unbundle use cases
YC gets it. Looks like 85% of its latest batches are AI pilled.
The unbundling of software has begun.
The rush is here.
Slide 54 - Sometimes a job is just a task to be automated
Elevators won't replace us, people who know how to use elevators will?
Elevator attendants were a big contributor to society and then vanished.
History doesn't repeat itself but it certainly rhymes.
Slide 61 - What was impossible that now become cheap?
Benedict is asking the right questions:
What if you had a million interns?
What if you had one intern who was a million times faster?
My big question here is... what businesses does this unlock?
Slide 74 - no one knows anything
The only thing we know is experimenting is the way to do it. Come up with your own ideas, or grab one from Ideabrowser
Whatever you do, experimenting a lot is key during platform shifts.
It'll take you experimenting on a few ideas to get to the right one.
And yeah, no one truly knows anything.
The only way out is through.
This is the most incredible time to be building.
Reading through this deck definitely reminded me of that. You can read the full deck here.
Happy building.
I'm rooting for you,
Greg Isenberg
gregisenberg.com
## 相关链接
- [GREG ISENBERG](https://x.com/gregisenberg)
- [@gregisenberg](https://x.com/gregisenberg)
- [71K](https://x.com/gregisenberg/status/2056369806232961276/analytics)
- [AI is Eating the World" slide deck.](https://www.ben-evans.com/presentations)
- [Ideabrowser](https://ideabrowser.com/)
- [here](https://www.ben-evans.com/presentations)
- [gregisenberg.com](https://gregisenberg.com/)
- [Upgrade to Premium](https://x.com/i/premium_sign_up)
- [9:42 PM · May 18, 2026](https://x.com/gregisenberg/status/2056369806232961276)
- [71.4K Views](https://x.com/gregisenberg/status/2056369806232961276/analytics)
- [View quotes](https://x.com/gregisenberg/status/2056369806232961276/quotes)
---
*导出时间: 2026/5/19 09:38:02*
---
## 中文翻译
# Benedict Evans 的“AI 正在吞噬世界”79 页幻灯片
**作者**: GREG ISENBERG
**日期**: 2026-05-18T13:42:21.000Z
**来源**: [https://x.com/gregisenberg/status/2056369806232961276](https://x.com/gregisenberg/status/2056369806232961276)
---
Benedict Evans 的“AI 正在吞噬世界”幻灯片中我最喜欢的几张幻灯片（以及我的收获）。
幻灯片 4 - 科技在平台变革中演进
在这个平台变革期间将创造的财富量级甚至难以想象。当你看这张图表时，确实会让人深刻意识到，我们正处于一个巨大的平台变革的初期阶段。
截至目前，对于创立公司来说，机会是无限的。
千万不要浪费一场好的平台变革 ;)
我希望我 20 年前就知道这一点。
幻灯片 6 - 平台变革重置了科技行业
微软在十年间从占据 95% 的计算份额跌至 10%。微软曾经统治着计算领域，直到电脑转移到了你的口袋里。
一眨眼功夫，这就消失了。如果你在平台变革期间创立公司，这是最令人兴奋的部分，世界任你驰骋。
一切皆有可能。
幻灯片 23 - Sam Altman 对未来的看法
我认为 Sam 是对的，人们会像购买电力这样的日用品一样购买智能。我最好的猜测是，他们会选择满足一定质量门槛的最低价提供商，就像你选择电力供应商一样。没人在意是谁发的电。他们在意的是灯能亮。这就是模型发展的方向。这意味着价值完全转移到了你在计量仪表之上构建的东西上。
幻灯片 26 - 平衡状态会是什么样子？
这就是我一直说要构建垂直应用的原因。模型层正在变成一个资本密集型、低利润且没有网络效应的商品。应用层，即为某个行业解决某个问题的垂直代理，才是网络效应、利润和垄断势力重新形成的地方。价值正在向堆栈上层移动。在那里构建。
利润在于客户关系和数据。
幻灯片 27 - 关于 AI 的论点
聊天是一种糟糕的用户体验 (UX)。同意。实验室无法构建所有的应用程序。显而易见。模型是没有网络效应的商品。这已经发生了。这意味着：创新向堆栈上层转移。直接转移到了那些在商品化智能之上为真实行业构建真实产品的创始人身上。
这对我们这些想要构建东西的人来说是个好消息。
幻灯片 30 - 消费者使用广泛但浅尝辄止
每个人都在谈论 AI 的采用，仿佛它已经很普及了（特别是在 X 上）。这张图表说明并非如此。大多数人使用 ChatGPT 的方式就像他们使用计算器一样。偶尔使用。当他们记得它的存在时才用。那些能赢得消费者 AI 的创始人，将会打造出人们在不知道自己“正在使用 AI”的情况下每天都会使用的产品。
幻灯片 37 - ChatGPT 的产品/市场匹配度 (PMF)
看看那些细小的条柱。法律、支持、医疗、房地产、金融。每一个都是巨大的行业，但目前几乎没在 AI 上花什么钱。编码业务能达到 30 亿美元，是因为开发者率先采用了它。其他所有类别都还在起跑线上。这张图表就是整个机会地图。
现在就去寻找那些资金充裕的 VC 公司尚未涉足的细分领域/垂直用例吧。
我可以说……我们是不是太早了？
幻灯片 48 - 初创公司的存在就是为了解耦用例
YC（Y Combinator）懂了。看起来其最新一批公司中 85% 都沉迷于 AI。
软件的解耦已经开始。
大潮来了。
幻灯片 54 - 有时工作只是将被自动化的任务
电梯不会取代我们，懂电梯的人会取代我们？
电梯操作员曾经是社会的巨大贡献者，然后消失了。
历史不会重演，但总会惊人的相似。
幻灯片 61 - 以前不可能但现在已经变便宜了的是什么？
Benedict 问对了问题：
如果你有一百万名实习生会怎样？
如果你有一名速度快一百万倍的实习生会怎样？
我这里的大问题是……这将解锁哪些业务？
幻灯片 74 - 没有人真正了解
我们唯一知道的方法就是通过实验。想出你自己的想法，或者从 Ideabrowser 抓取一个
无论你做什么，在平台变革期间大量尝试是关键。
你需要尝试几个想法才能找到正确的那个。
是的，没有人真正了解任何事。
唯一的出路是穿越过去。
这是构建产品的最令人难以置信的时代。
通读这份幻灯片肯定让我想起了这一点。你可以在这里阅读完整的幻灯片。
祝你构建愉快。
我为你们加油，
Greg Isenberg
gregisenberg.com
## 相关链接
- [GREG ISENBERG](https://x.com/gregisenberg)
- [@gregisenberg](https://x.com/gregisenberg)
- [71K](https://x.com/gregisenberg/status/2056369806232961276/analytics)
- [AI is Eating the World" slide deck.](https://www.ben-evans.com/presentations)
- [Ideabrowser](https://ideabrowser.com/)
- [here](https://www.ben-evans.com/presentations)
- [gregisenberg.com](https://gregisenberg.com/)
- [Upgrade to Premium](https://x.com/i/premium_sign_up)
- [9:42 PM · May 18, 2026](https://x.com/gregisenberg/status/2056369806232961276)
- [71.4K Views](https://x.com/gregisenberg/status/2056369806232961276/analytics)
- [View quotes](https://x.com/gregisenberg/status/2056369806232961276/quotes)
---
*导出时间: 2026/5/19 09:38:02*