Two Worlds: Capability ≠ Value in the AI Era两个世界：AI 时代，能力 ≠ 价值
Claude Mythos drops, AI bubble fears rise — both can be true. Geohot explains why capability and value are two different things, and why the AI floor rising doesn't mean the market grows.Claude Mythos 发布，AI 泡沫论同时升温——两者可以并存。Geohot 解释为什么「能力」和「价值」是两回事，为什么 AI 的能力天花板提升不等于市场蛋糕变大。
The Setup
Two headlines are running in parallel: Claude Mythos drops and gets called “dramatically better than Opus 4.6 — surely this is AGI.” At the same time, the AI bubble narrative is louder than ever. George Hotz (geohot) asks the question most analysts won’t: how can both be true?
Key Takeaways
- Capability and value are not the same thing. A smartphone and photo printer taken to 1850 could make you a millionaire. That same tech today can’t earn $5 on a street corner — because everyone has it.
- AI raises the bar, not the ceiling. It doesn’t replace skilled programmers or artists; it makes unskilled output worthless. “You Can Just Build Things, But So Can Everyone Else.”
- What AI produces by itself is worth very little — because anyone else can produce the same thing. Skill is still the differentiator. Top photographers today use iPhones; they win because of craft, not tools.
- The floor rises, but the market doesn’t necessarily grow. If AI companies capture value while the total market stays flat, they’re extracting from everyone else — not creating new wealth.
- Hotz’s honest take: He loves AI from pure curiosity (“can’t wait to meet silicon-based life”), but wants superhuman models that nobody profits from.
Why It Matters
This framing cuts through the noise on AI valuations and the “builder economy.” If you’re allocating to AI plays — companies, tokens, equity — the question isn’t just “can this model do more?” It’s “does anyone capture incremental value?”
The Bitcoin/crypto market already knows this pattern: better tech doesn’t automatically mean higher prices. Value accrual depends on who controls scarcity, distribution, and moats — not raw capability.
For Rex as a builder: shipping an AI product faster than before is table stakes now. The question is whether what you build has a moat that AI itself can’t commoditize in 6 months.
What to watch:
- AI company revenue vs. GDP growth correlation — are they growing the pie or taking share?
- The 2028 US election: Hotz calls AI the “major hot button issue” — regulatory risk is real
- How Anthropic, OpenAI price their next-gen models and whether value leaks to users or stays captured
背景
两条新闻同时在跑：Claude Mythos 发布，被称为「比 Opus 4.6 显著更强——这一定是 AGI 了吧」；与此同时，AI 泡沫论声音越来越响。George Hotz（geohot）问了一个大多数分析师不敢问的问题：这两件事怎么能同时为真？
关键要点
- 能力和价值不是一回事。 1850 年带着智能手机和照片打印机去，你能成为百万富翁。同样的工具今天在街角卖，一分钱都赚不到——因为人人都有。
- AI 抬高的是门槛，不是天花板。 它不会取代有技能的程序员或艺术家，而是让没技能的输出变得一文不值。「你可以随便 Build 东西，但其他人也行。」
- AI 自己产出的东西价值极低 ——因为其他人也能产出一模一样的东西。技能依然是分水岭。顶级摄影师今天用 iPhone 拍照，赢在的是审美和创意，不是工具。
- 门槛提升，市场不一定变大。 如果 AI 公司在市场总量不变的情况下拿走更多价值，他们是在从所有人身上提取——而不是创造新财富。
- Hotz 的诚实表态： 他喜欢 AI 是因为纯粹的好奇心（「迫不及待想认识硅基生命」），但他希望看到的是没有人从中牟利的超人模型。
为什么重要
这个框架能直接穿透 AI 估值和「Builder 经济」的噪音。如果你在配置 AI 方向——公司股票、代币、股权——问题不只是「这个模型能做更多事吗？」而是「有没有人能真正捕获增量价值？」
加密/比特币市场早就见过这个模式：更强的技术不自动等于更高的价格。价值积累取决于谁控制稀缺性、分发渠道和护城河——而不是原始能力。
对 Rex 来说：用 AI 更快地发布产品已经是基本操作，不算壁垒。真正的问题是：你 Build 的东西有没有护城河，能不能抵抗 AI 在 6 个月内把它商品化？
值得关注：
- AI 公司营收与 GDP 增长的相关性——他们在做大蛋糕，还是在抢份额？
- 2028 年美国大选：Hotz 说 AI 将是「最大的热点议题」——监管风险是真实存在的
- Anthropic、OpenAI 下一代模型的定价策略——价值是流向用户，还是被公司截留？
Related Essays相关文章
Join Rex's Lab on Telegram 加入 Rex's Lab
Crypto · AI · Investing — raw thinking, before it becomes a tweet. 加密 · AI · 投资 — 推文之前，更原始的思考。