# Ultra-Agency: Going beyond "High-Agency"
**作者**: Kato
**日期**: 2026-10-09T10:04:59.000Z
**来源**: [https://x.com/Roguefort/status/2108498953062805690](https://x.com/Roguefort/status/2108498953062805690)
---
## The promise:
Long before true AGI/ASI exists we'll have found ways to edit our genes and make ourselves more intelligent than it. we'll have methods to increase brain plasticity without the need for 'drugs' like SSRI's and other such messy, limited and slow crutches. We'll have gene editors that can recover hearing, eyesight, cure ailments like arthritis, get people recovering from spinal injuries walking again before muscle wasteage had set in, and that's just the beginning. (and some of it is already happening)
Back in 2019 it scientists restored sight in blind mice, and even human eyesight as of 2021. All of it was done through injecting edited genes in the area right behind the damaged area, no surgery was needed, nothing 'invasive' like surgery was done, it was all through ultra-local treatments.
long before AI has a chance of becoming malicious, we'll already have the gene therapy methods to overcome it.
## Consciousness:
As of now, AI isn't conscious, and rest-assured, it won't be for a long time, a neuron that allows for consciousness, let alone self awareness, are ultra-complex mono-cellular live forms, there is a reason 3 neurons alone are sufficient to solve linear computational problems. now gather a 86 billion of them, that's where we are.
LLM's couldn't figure out simple math at 100's of billions params, let alone understand how many 'r's there are in "strawberry". a single parameter "[0,193843]" is simply not a substitute for a neuron alone, not to mention a substitution for the complex hormonal interplay that drives neurons.
when we're ready to make ourselves smarter, we'll end up entering our own recursive self-improvement feedback loop, we'll have humans self-improving, improving AI, improving humans, etc. monthly gene therapy "jabs" in the shoulder might just become the norm, and even that will become 'ancient' rather quick.
## Data driven frontier:
AI is already helping us solve 'data' problems, like mapping and releasing 100's of millions modeled proteins as AlphaFold did in 2022, with which the smartest people are already developing new treatments with.
Manually uncovering just one of them by hand, would take a phd 4 to 5 years alone. in 50 years, 170 thousand proteins have been discovored, 19 thousand of them in 2019, at that pace, to have discovered the other 199.93 million proteins by hand would have taken us another 10000+ years.
Scientists reading through all the existing papers, will soon enough become a thing of the past with them having near instant access to our whole corpus of knowledge from a single source. and, the little secret here: (it already mostly is.) now with alphafolds data accessible to increasingly smarter LLM's that have consolidated all the possible treatments and scientific findings, whether new or old. we already have removed one of the most time consuming part of developing many new medical treatments; the research itself.
## The risk:
People might be afraid of the what if: 'what if ai does 'bad thing' and gives us a gene treatment that kills us all'
That's why medical trials will stay important until effect can be fully modeled on a per person basis through gene-simulation. until then trials will likely have to become almost isolatingly rigorous. I can imagine a real risk, but one cannot truly know without trying, and i believe it is something worth trying.
## Ultra-Agency:
In the future, we'll find ourselve living in a world where "realistic superpowers" will become the cultural default, superhuman strength, intelligence, emotional regulation, focus, extreme executive functioning. It won't just be people with "high agency" anymore, the default for everyone taking gene-therapy, will be "Ultra-Agency". When AI takes over the jobs, the people that will take on the benefits of gene therapy will start pioneering the development and expansion of the human frontier. and all human knowledge therein. Billions of humans working, solving and thinking in areas both niche and major, purely driven by curiosity and personal interest. we'll live in a time where the whole of humanity will be seen cutting away at the cosmic marble of reality until the beauty or dread inside of it is revealed.
Ultra-agency will not be something found in conversations, it will be found to be the default. an unimportant topic to discuss, yet ever-present.
## 相关链接
- [Kato](https://x.com/Roguefort)
- [@Roguefort](https://x.com/Roguefort)
- [11K](https://x.com/Roguefort/status/2108498953062805690/analytics)
- [Upgrade to Premium](https://x.com/i/premium_sign_up)
- [6:04 PM · Oct 9, 2026](https://x.com/Roguefort/status/2108498953062805690)
- [11.2K Views](https://x.com/Roguefort/status/2108498953062805690/analytics)
---
*导出时间: 2026/10/9 23:02:25*
---
## 中文翻译
# 超级能动性：超越“高能动性”
**作者**: Kato
**日期**: 2026-10-09T10:04:59.000Z
**来源**: [https://x.com/Roguefort/status/2108498953062805690](https://x.com/Roguefort/status/2108498953062805690)
---
## 承诺：
在真正的 AGI/ASI 出现之前很久，我们就会找到编辑自身基因的方法，让自己变得比它更聪明。我们将拥有无需依赖 SSRI 等药物那样混乱、有限且缓慢的“拐杖”，就能提高大脑可塑性的方法。我们将拥有能够恢复听力、视力、治愈关节炎等疾病的基因编辑器，让脊髓损伤患者在肌肉萎缩发生之前重新站起来行走——而这仅仅是开始。（其中一些已经在发生了）
早在 2019 年，科学家就恢复了盲鼠的视力，甚至到 2021 年还恢复了人类的视力。这一切都是通过将编辑过的基因注射到受损区域后方来完成的，不需要手术，没有进行任何像手术那样的“侵入性”操作，全部都是通过超局部治疗实现的。
早在 AI 有机会变得恶意之前，我们就会拥有战胜它的基因疗法。
## 意识：
目前，AI 并没有意识，而且放心，它在很长一段时间内都不会有。能够支撑意识——更别说自我意识——的神经元，是极其复杂的单细胞生命体。仅仅 3 个神经元就足以解决线性计算问题，这是有原因的。现在想象一下汇聚 860 亿个这样的神经元——这就是我们人类所处的水平。
LLM 在数千亿参数规模时连简单的数学都搞不明白，更不用说理解“strawberry”里有几个“r”了。一个单独的参数"[0,193843]"根本替代不了一个神经元，更不用说替代驱动神经元的复杂激素相互作用了。
当我们准备好让自己变得更聪明时，我们最终会进入自己的递归自我改进反馈循环——人类自我改进，改进 AI，AI 再改进人类，如此循环。每月在肩膀上打一针基因治疗“针剂”可能会成为常态，甚至这一点也会很快变得“过时”。
## 数据驱动的前沿：
AI 已经在帮助我们解决“数据”问题，比如像 AlphaFold 在 2022 年那样，映射并发布数亿个建模蛋白质，而最聪明的人已经在利用这些成果开发新的治疗方法。
仅靠人力手工揭示其中一个蛋白质的结构，就需要一位博士独自耗费 4 到 5 年时间。在过去的 50 年里，人类发现了 17 万个蛋白质，其中 2019 年发现了 1.9 万个。按照这个速度，要手工发现其余的 1.9993 亿个蛋白质，我们还需要 10000 多年。
科学家通读所有现有论文的日子很快就会成为过去，他们将能通过单一来源近乎即时地访问我们的全部知识库。而且，这里有个小秘密：（这在很大程度上已经是现实了。）现在，随着 AlphaFold 的数据可以被越来越聪明的 LLM 访问，而这些 LLM 已经整合了所有可能的治疗方法和科学发现——无论新旧——我们已经消除了开发许多新医疗方法中最耗时的环节之一：研究本身。
## 风险：
人们可能会害怕这样的假设：“如果 AI 做了‘坏事’，给我们一种会杀死所有人的基因治疗怎么办？”
这就是为什么医疗试验将保持其重要性，直到能够通过基因模拟针对每个人完全建模疗效。在那之前，试验可能必须变得几乎与世隔绝般严格。我可以想象确实存在真实的风险，但不尝试就永远无法真正知道，而我相信这是值得尝试的事情。
## 超级能动性：
在未来，我们会发现自己生活在一个“现实版超能力”成为文化默认值的世界里——超人般的力量、智力、情绪调节、专注力、极强的执行功能。不再只是拥有“高能动性”的人了，接受基因治疗的每个人的默认状态都将是“超级能动性”。当 AI 接管了工作，那些接受基因治疗好处的人们将开始开拓人类前沿的发展与扩张，以及其中蕴含的全部人类知识。数十亿人将在或冷门或重大的领域工作、解决和思考，纯粹出于好奇心和个人兴趣。我们将生活在一个这样的时代：全人类将被世人看着雕琢现实的宇宙大理石，直到其中蕴含的美或恐惧显露出来。
超级能动性不会是对话中才会出现的东西，它将被发现就是默认状态。一个不值得讨论的话题，却又无处不在。
## 相关链接
- [Kato](https://x.com/Roguefort)
- [@Roguefort](https://x.com/Roguefort)
- [11K](https://x.com/Roguefort/status/2108498953062805690/analytics)
- [升级到 Premium](https://x.com/i/premium_sign_up)
- [2026年10月9日 下午6:04](https://x.com/Roguefort/status/2108498953062805690)
- [11.2K 次浏览](https://x.com/Roguefort/status/2108498953062805690/analytics)
---
*导出时间: 2026/10/9 23:02:25*
