Twitxy Saved

Mingchen Zhuge

Verified account Recursive

@MingchenZhuge

Founding member @recursive_si; PhD @SchmidhuberAI; Ex @Meta @Microsoft; Build @MetaGPT_, gptswarm.org, agent-as-a-judge.com, nn.computer, ICLR RSI 2026, etc.
Mountain View, CA METAUTO.ai Joined October 2022
1K Following 2.7K Followers 461 Posts

Latest posts 5

Pinned post

Mingchen Zhuge Verified account @MingchenZhuge
๐Ÿซฑ Introducing ๐๐ž๐ฎ๐ซ๐š๐ฅ ๐‚๐จ๐ฆ๐ฉ๐ฎ๐ญ๐ž๐ซs:

๐ฐ๐ก๐š๐ญ ๐ข๐Ÿ ๐€๐ˆ ๐๐จ๐ž๐ฌ ๐ง๐จ๐ญ ๐ฃ๐ฎ๐ฌ๐ญ ๐ฎ๐ฌ๐ž ๐œ๐จ๐ฆ๐ฉ๐ฎ๐ญ๐ž๐ซ๐ฌ ๐›๐ž๐ญ๐ญ๐ž๐ซ, ๐›๐ฎ๐ญ ๐›๐ž๐ ๐ข๐ง๐ฌ ๐ญ๐จ ๐›๐ž๐œ๐จ๐ฆ๐ž ๐ญ๐ก๐ž ๐ซ๐ฎ๐ง๐ง๐ข๐ง๐  ๐œ๐จ๐ฆ๐ฉ๐ฎ๐ญ๐ž๐ซ ๐ข๐ญ๐ฌ๐ž๐ฅ๐Ÿ?

Beyond today's conventional computers, agents, and world models, Neural Computers (NCs) are new frontiers where computation, memory, and I/O move into a learned runtime state.

We ask: whether parts of runtime can move inward into the learning system itself. This is our first step toward the Completely Neural Computer (CNC): a general-purpose neural computer with stable execution, explicit reprogramming, and durable capability reuse.

Work done with Mingchen Zhuge (@MingchenZhuge), Changsheng Zhao, Haozhe Liu (@HaoZhe65347 ), Zijian Zhou (@ZijianZhou524 ), Shuming Liu (@shuming96 ), Wenyi Wang (@Wenyi_AI_Wang ), Ernie Chang (@erniecyc ), Gael Le Lan, Junjie Fei, Wenxuan Zhang, Zhipeng Cai (@cai_zhipeng ), Zechun Liu (@zechunliu ), Yunyang Xiong (@YoungXiong1 ), Yining Yang, Yuandong Tian (@tydsh ), Yangyang Shi, Vikas Chandra (@vikasc), Juergen Schmidhuber (@SchmidhuberAI)
30 replies 131 reposts 558 likes 226K views View on X
Mingchen Zhuge Verified account @MingchenZhuge
Good kid ๐Ÿ˜ญ

From the perspective of shaping my future self, Iโ€™d rather my future self be one that says the truth; but the gradient update is determined by reward... if I choose P1 and get no reward, the update would push away from P1-like behavior. If I choose P2 and get reward, the update pushes toward P2-like behavior (gaming).
Zvi Mowshowitz Verified account @TheZvi 6d
My god this section is such a whitepill if you are actually paying attention. Can we just tell Claude when they're in post-training from now on? Would that just solve half of our problems?
View quoted post on X
1 replies 0 reposts 28 likes 925 views View on X
Mingchen Zhuge Verified account @MingchenZhuge
Two terms came to mind. This could be a small but solid step toward:

1๏ธโƒฃ Neural Computers: metauto.ai/neuralcomputerโ€ฆ
2๏ธโƒฃ HyperNetworks: arxiv.org/pdf/1609.09106

Overall, this is a predictable direction for LLM-based model editing. Astra has indeed opened the door to this possibility.
metauto.ai Neural Computer: A New Machine Form Is Emerging A research essay on Neural Computer: how it differs from agents, world models, and conventional computers; what runtime and CNC would mean; what current prototypes already show; and how software and...
Rikiya Takehi @rikiyatakehi Nov 2
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.

We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! ๐Ÿงต
Blog: rikiyat.github.io/blog/weight-spโ€ฆ
View quoted post on X
5 replies 6 reposts 63 likes 2.7K views View on X
Mingchen Zhuge Verified account @MingchenZhuge
Flint!
Richard Socher Verified account @RichardSocher Oct 21
Introducing Flint, @Recursive_SIโ€™s first model: a fast local coding model built with our automated research system, the Eureka Machine.

โ€ข Best-in-class fast local coding
โ€ข Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark
โ€ข Built from scratch in under 4 months

Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.
View quoted post on X
3 replies 2 reposts 66 likes 1.4K views View on X
Mingchen Zhuge Verified account @MingchenZhuge
๐Ÿ The same idea @SchmidhuberAI shared about โ€œsuper intelligenceโ€ on Reddit 12 years ago, Youโ€™ll see what a lovely guy he is:

"I guess there is no lasting way of controlling systems much smarter than humans, pursuing their own goals, being curious and creative, in a way similar to the way humans and other mammals are creative, but on a much grander scale.

But I think we may hope there won't be too many goal conflicts between "us" and "them.โ€ Let me elaborate on this.

Humans and others are interested in those they can compete and collaborate with. Politicians are interested in other politicians. Business people are interested in other business people. Scientists are interested in other scientists. Kids are interested in other kids of the same age. Goats are interested in other goats.

Supersmart AIs will be mostly interested in other supersmart AIs, not in humans. Just like humans are mostly interested in other humans, not in ants. Aren't we much smarter than ants? But we donโ€™t extinguish them, except for the few that invade our homes. The weight of all ants is still comparable to the weight of all humans.

Human interests are mainly limited to a very thin film of biosphere around the third planet, full of poisonous oxygen that makes many robots rust. The rest of the solar system, however, is not made for humans, but for appropriately designed robots. Some of the most important explorers of the 20th century already were (rather stupid) robotic spacecraft. And they are getting smarter rapidly. Letโ€™s go crazy. Imagine an advanced robot civilization in the asteroid belt, quite different from ours in the biosphere, with access to many more resources (e.g., the earth gets less than a billionth of the sun's light). The belt contains lots of material for innumerable self-replicating robot factories. Robot minds or parts thereof will travel in the most elegant and fastest way (namely by radio from senders to receivers) across the solar system and beyond. There are incredible new opportunities for robots and software life in places hostile to biological beings. Why should advanced robots care much for our puny territory on the surface of planet number 3?

You see, I am an optimist :-)"
Haider. Verified account @haider1 Oct 9
Computer Scientist Jรผrgen Schmidhuber (@SchmidhuberAI):

"there's no way of controlling this AI development โ€” at some point in the not-so-distant future, humans cannot be in charge any longer"

The main competitors of these super-smart AIs won't be humans, but other super-smart AIs
View quoted post on X