Zehua Cheng

Chief [email protected] PhD@UniofOxford

[FLock.io @flock_io]

Next stop for CEO Forum Blockchain and AI for Good - Young Talent Edition: @UniofOxford | March 28

​You will hear from:
​
🎙️ @0x7SUN, founder & CEO of FLock, and University of Oxford alumnus.
🎙️ @helen_abfinance, founder of @abf_finance and @ChainforGood, former co-CEO of @Bybit_Official

Join us for builder insights, explore what’s pushing the edge of blockchain and AI, and connect with the FLock & BGA teams.

In collaboration with Oxford CSSA.

If you're in town this Saturday, come through👇

[FLock.io @flock_io]

It's official: CEO Forum - Blockchain and AI for Good: Young Talent Edition is coming to @imperialcollege | March 25

Insider talks from founders and professors who are building and working at the intersection of blockchain and AI.

📢 Speakers:
@0x7SUN, founder & CEO of FLock
@helen_abfinance, founder of @abf_finance and @ChainforGood, former co-CEO of @Bybit_Official
Ying-Ying Hsieh, assistant professor of Innovation and Entrepreneurship; Associate Centre Director, Imperial College Centre for Cryptocurrency Research and Engineering, @imperialcollege.

If you're a student, builder, or just deeply curious about where purpose-driven tech is heading, this is your invite 🇬🇧

[Zehua Cheng @zehuac]

I saw lots of old friends there. Happy coding and build sth great.

[Peter Steinberger 🦞 @steipete]

This is humbling and insane and amazing. 🫶🦞https://t.co/t52IGtWkML

[Zehua Cheng @zehuac]

@monkeyjiang @Debot_Official @btc710718 @flock_io @0x7SUN 确实不错

[jsun.base.eth @0x7SUN]

🔥

[cosmeticfish @cosmeticfish]

January 15, 2026 is the FLock Day. Merry FLock!

[Zehua Cheng @zehuac]

Sybil resistance + reputation + freezing/slashing behavior. They explicitly require "sybil-resistant" reputation systems and describe enforcement actions that look like protocol-level sanctions: revoking credentials, freezing staked assets, and flagging transaction history for forensic review. "Sybil-resistant" is also a very crypto-native vocabulary, and the enforcement mechanism mirrors how many protocols think about identity/reputation under adversarial duplication.
That's also what FLock is building.
😀Good to know decent teams working together.

[Chao @chaowxyz]

刚读完Google最新的AGI论文，论文的观点大胆且非常crypto native，我一度以为是在看加密项目的白皮书。

几个核心观点：
1. AGI终将是一个 DAO，而不是一个 CEO。
我们总是幻想某天一觉醒来，诞生了 GPT-10 这样的全知全能神。论文指出未来的AGI大概率是分布式的 。就像公司没有一个人能擅长所有事情，AGI 将是一个由无数互补的“特种 Agent”组成的网络 。在这个网络里，没有单一的中心，超级智能是在Agent疯狂的交易和协作中涌现出来的 。

换句话说，AGI不是一个实体，而是一种市场状态。

2.治理AGI不能靠“法律”，得靠“智能合约”。
既然从单一模型变成了市场，安全范式必须从“心理学”转向“治理学” 。以前 AI 安全是做一个超级大脑的对齐。但面对每秒亿万次的高频交互，人类监管是无效的。因此必须引入智能合约。Agent 完成任务后，由Oracle 验证结果，自动执行结算 。代码即法律，不满足安全约束，资金就无法流转。

3. 引入“质押”与“罚没”。
怎么防止 Agent 作恶？研究团队居然复刻了PoS机制。Agent想接大单？先Stake资产 。一旦被审计发现作恶，智能合约直接罚没质押金 。这种基于经济抵押的信任，比单纯的代码审查更有效。

4. 链上身份与“Gas 费”调节
DID：为了防女巫攻击，每个 Agent 必须有基于公钥密码学的唯一身份，且绑定法律实体。
动态Gas：为了防垃圾数据污染，建议对Agent的操作征收动态费用 。这不就是以太坊拥堵时的 Gas Fee 调节机制吗？

5. 记录上链
所有的决策和交易历史，必须记录在加密安全的、只能追加不能修改的账本上 。这为了确保当系统出问题时，能进行取证分析，没人能赖账 。

这篇论文标志着 AI 安全的一个范式转移：从单纯的计算机科学和价值观对齐，变成了经济学和博弈论 。

而加密圈子折腾了十几年的DID、智能合约、预言机和经济模型，治理机制等等等等。无论成熟与不成熟，这些探索至少迈出去一些步，而这些步，也许刚好是给未来那个庞大的、去中心化的硅基生命群体的基础。

未来的 AGI 安全专家，可能不需要是一个能写代码的工程师，但必须是一个懂博弈论、市场设计和去中心化治理的“AI 经济学家”。我们本质是要设计在这个新物种的共识、激励和治理模型，而不仅仅是修补大模型的神经元。

论文信息量很大，以上只cover了一部分，有兴趣可以阅读原文
https://t.co/R7DaJCfhmj

[HashKey Group @HashKeyGroup]

HashKey is now officially a public company on @HKEXGroup! A meaningful step toward making digital assets massively accessible.

The ticker is https://t.co/mVfdXgaNZB.

[Zehua Cheng @zehuac]

😀 Thank you for your attention to FLock's work in Secure AI Agent.

[FLock.io @flock_io]

WE'RE HIRING!

Just a "casual" job opening at the University of Oxford.

Open Post-doctoral role in Decentralized AI, in collaboration with https://t.co/YkoQx8t9ue.

Building the future of AI isn't about hype; it's about science.

Apply by December 12, 2025. Details below.

[Zehua Cheng @zehuac]

@RMantri It's just a student union. And you are not the leader of any community.

[Zehua Cheng @zehuac]

Congratulations for the NeurIPS 2025 Best Paper Award! Very impressive works!

[Qwen @Alibaba_Qwen]

🏆 We are incredibly honored to announce that our paper, "Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free" has received the NeurIPS 2025 Best Paper Award!

A huge congratulations to our dedicated research team for pushing the boundaries of AI.

Read more: https://t.co/qu3ERa3pH5

[Zehua Cheng @zehuac]

Great honor to introduce FLock at University of Cambridge. Good to meet Mr. Xiao Li from Embassy of the
PRC in the Uk. Dr. Julian Huppert from Jesus College, Univ of Cambridge. Prof. Pietro Lió from Department of CST, University of Cambridge. And all fellow scholars.

[Zehua Cheng @zehuac]

Actually not everyone noticed that we we ranked at 6th at Task 1 (where actually is most relevant to real prediction task in platform like Polymarket) with a 20B level of LLM. 20B model, I can completely ran on my own MacBook Pro. GPT5 is impressive, and they also knew the every bet.

[FLock.io @flock_io]

FLock’s AI prediction model ranked Top 6 on FutureX, outperforming multiple 70B-400B+ models.

We’re also the only Web3 team in the global top 10.

Our approach uses a lightweight 20B open-source model, trained end-to-end on FLock’s stack: federated training, reinforcement learning, and our Edge Agent framework, while other models run on massive 70B-400B+ parameter bases.

This week, it ranked #6 on the FutureX leaderboard, competing with some of the most advanced AI models in the world, and ranked ahead of several of them in overall prediction performance.
- GPT-5 variants
- Alibaba Qwen3-235B
- Kimi-k2 (Moonshot)
- ChatGPT-Agent
- Grok-4 (Peking University)
- DeepSeek
- Tencent Hunyuan
- Gemini 2.5 Pro
- xAI Agents
… and more

What makes this special is not the rank alone, but how it was achieved.

While most leading models rely on massive parameter scales, our 20B model shows how efficient training, a strong data strategy, and decentralized compute can deliver competitive performance.

FutureX is the industry’s leading benchmark for real-world future prediction, evaluating models on their ability to predict upcoming financial, social, and economic events. This is exactly the capabilities needed for next-gen prediction models.

Our result proves that:
- Smaller models can outperform massive ones when trained right
- Decentralized AI pipelines can deliver state-of-the-art results
- Open-source foundations can compete with big proprietary labs
- FLock is the only decentralized AI team to make the list

We believe AI will change how finance and trading work. That’s why we’re focused on building the infrastructure for that future-prediction model, robust federated training network, and a decentralized network for model assets.

[FLock.io @flock_io]

Highlights from the Blockchain Impact Forum in Copenhagen. 🇩🇰

Co-hosted by @UNDP AltFinLab and @ChainforGood, this is a global event bringing together leaders building blockchain for impact.

The FLock team joined key sessions throughout the event. 🧵

[Four Pillars @FourPillarsFP]

: : [Report] https://t.co/kKHasq9pRE: The Base Layer for AI Democratization
Written by @JayLovesPotato

---
https://t.co/ABwgjeS6nI
---

The current AI industry lacks transparency and trust throughout the lifecycle of AI model development. Additionally, high entry barriers limit the expansion of AI applications across various industries.

To address these challenges, federated learning adopts an approach where models are distributed to local clients for independent training. Only the trained parameters are aggregated to update the global model. This method holds significant potential for reducing the overall cost of developing global models and minimizing the exposure of sensitive personal data, thereby greatly enhancing usability.

However, federated learning also faces its own limitations. Key challenges include the difficulty of recruiting a sufficiently diverse and honest pool of participants for model training and the reliance on centralized servers for certain operational tasks.

https://t.co/kKHasq9pRE aims to overcome these limitations by integrating various blockchain elements with federated learning methodologies. Its ultimate goal is to democratize the entire AI model lifecycle—from data collection and model proposal to training and application—paving the way for a more creative and trustworthy AI industry.

▫️ The Current State of the AI Industry Requires Structural Transformation
▫️ Federated Learning as an Alternative to Centralized AI Development
▫️ https://t.co/kKHasq9pRE: The Next-Generation Federated Learning Platform Powered by Blockchain Architecture
▫️ Democratizing AI in an Era Where Personalization Ruins Personalization

[Zehua Cheng @zehuac]

WE ARE BACK

[Grayscale @Grayscale]

Learn about the diverse digital assets we’re considering for future investment products and explore those already part of our offerings in our latest Assets Under Consideration update. Are we missing anything? 🤔

Read the full report: https://t.co/Tr5lU1CSSQ

[jsun.base.eth @0x7SUN]

Thx @UNDP for the invitation.

Heard great things about blockchain strategic board discussions with FLock’s epic representatives @YifanX and @sameeha_rehman

More initiatives to come, lots of “the 1st in the world” to be built by @flock_io

Stay FLocked.

[Zehua Cheng @zehuac]

It's gonna be FLOCKING GOOD.

[FLock.io @flock_io]

It’s an honor for https://t.co/evQHqjcMLM to be a part of the High-Level Strategic Dialogue on Blockchain for Development with @UNDP at UNDP Headquarters in New York City today.

Today’s dialogue was a critical first step in exploring how blockchain can create impact for people and the planet, with its unique features — trust, transparency, traceability, and accountability — at the centre.

Together with UNDP and industry-leading blockchain leaders, we developed a shared vision for how responsible and inclusive blockchain innovation can advance the SDGs, and explored pathways for future collaboration, including the potential establishment of a Blockchain Advisory Group.

FLock everywhere.

[Zehua Cheng @zehuac]

Seriously, Better on Base is nice.

[jsun.base.eth @0x7SUN]

FLockCON was a great success today

Hosting our afterparty at No.1 bar in Seoul,
with flocking based cocktail menu.