Liam｜AI Crypto Signals

Focused on AI × Crypto × Enterprise
Sharing trends, products, adoption, and opportunity signals
聚焦 AI × Crypto × Enterprise
分享趋势、产品、落地与机会判断

[Liam｜AI Crypto Signals @liam_ai_crypto]

这个账号聚焦 AI × Crypto × Enterprise，
我会持续分享：
• 值得关注的趋势变化
• 新产品、新工具和新叙事
• 企业落地的真实机会与限制
• 我对热点、项目和方向的判断

我不想放大噪音，
更想筛选真正有价值的信号。

如果你也关注 AI、Crypto、企业数字化，
欢迎关注，一起看清趋势，找到机会。

This account is focused on AI × Crypto × Enterprise.
I share:
• key trend shifts worth watching
• new products, tools, and narratives
• real adoption opportunities and constraints in enterprises
• my take on signals, projects, and market direction

I’m not here to amplify noise.
I’m here to filter for what actually matters.

If you’re also paying attention to AI, Crypto, and enterprise digital transformation,
follow along and let’s find the real signals together.

[qinbafrank @qinbafrank]

https://t.co/hH0FQPgIt3

[戈多Godot @GodotSancho]

https://t.co/zPPbok5BEJ

[Serena @369Serena]

Serenity 火了以后，很多人第一反应是去抄作业。

但其实我们可以给 Codex 它装上这些投资类 Skills 和 AI 投研框架后

它可以变成一个股票研究助理：

帮你拆产业链、读财报、看情绪、查风险、做估值、写反方报告，甚至把策略拿去回测。

先收藏这 5 个 GitHub 项目!
别让 AI 直接替你买股票，先让它帮你把研究做扎实。

1️⃣ UZI-Skill⭐ 2.4K

中文投资者最值得先看的一个!

它把 A股、港股、美股分析做成了一套完整流程：深度分析、陷阱排查、龙虎榜、投资大佬评审团、22维数据、180条量化规则。

适合拿来做个股初筛、风险排查和中文市场研究。

GitHub：
https://t.co/1Dq2yXMLhq

2️⃣ TradingAgents⭐ 83.7K
GitHub 上最火的 AI 金融交易框架之一。
它把一个投研团队拆成多个 Agent：基本面分析、情绪分析、新闻分析、技术分析、研究员、交易员、风控团队。

如果你想看看“AI 投研团队”到底怎么协作，这个项目最有代表性。

GitHub：
https://t.co/N0VMCVroA9

3️⃣ Serenity Skill⭐ 731

最近 X 上讨论很多的投资 Skill。

它复用 Serenity 的核心研究路径：从热点出发，拆产业链，找供应链瓶颈，再回到股票和基金方向，检查公告、财报、客户、产能和风险。
适合研究 AI 半导体、CPO、机器人、算力、电力设备、创新药这类“产业链很长、信息很乱”的方向。

GitHub：
https://t.co/YGgiXR6aZr

4️⃣ Buffett Skills⭐ 512

这个 Skill 适合在你心动之前泼一盆冷水。

它会逼你检查：

这家公司是不是好生意？
有没有护城河？
现金流可靠吗？
管理层可信么？
价格有没有安全边际？

适合长期投资者做质量审查。

GitHub：
https://t.co/pDMtWc6LTB

5️⃣ QuantDinger ⭐ 7.4K

一个更完整的 AI 量化交易工作台。

它覆盖从 AI 研究、策略代码、回测、模拟盘，到实盘执行和监控的完整链路。

如果你不满足于“让 AI 分析股票”，想把想法变成策略并验证，QuantDinger 更适合继续研究。

GitHub：
https://t.co/cZqOdURO5u

别让 AI 替你买股票。

让 Codex 替你拆产业链、读财报、查证据、找风险、写反方报告、做回测。

这才是普通投资者最该抄走的用法。
#codex #skills #Serenity

[Tw93 @HiTw93]

https://t.co/x9eqpjrSuS

[168X @168X_Fortune]

https://t.co/Xp0zM1HULI

[Jason Zhu @GoSailGlobal]

https://t.co/js41yhOBgJ

[Liam｜AI Crypto Signals @liam_ai_crypto]

学习Codex必备

[Liam｜AI Crypto Signals @liam_ai_crypto]

This round of AI competition is no longer just about “whose model is stronger.”

The bigger question is: who gets the default distribution slot first, who locks in cloud and compute first, and who fixes the growth path first.

Deals like Anthropic-Amazon show one thing clearly:
model competition is turning into a contest over platforms, entry points, and cost structure.
In the end, it is not just about capability. It is about lock-in.

The real question now: model quality or distribution control?

[Liam｜AI Crypto Signals @liam_ai_crypto]

Enterprises won’t ultimately pay for “the best model on the leaderboard.”
They’ll pay for the system that can actually enter production, be trusted, and keep improving over time.
Whoever combines multi-product adoption, an agent platform, and a real deployment engine is much closer to building the true enterprise moat.

[Liam｜AI Crypto Signals @liam_ai_crypto]

A simple way to see the difference:

“Can it work?” is a product question.
“Can it fit into real workflows?” is an infrastructure question.

The second one is usually where long-term adoption gets decided.

一个很简单的区分方式：

“它能不能跑起来”，是产品问题。
“它能不能接进真实工作流”，是基础设施问题。

长期 adoption 往往就是在第二个问题上被决定的。

[Liam｜AI Crypto Signals @liam_ai_crypto]

I increasingly think the next real competition
isn’t just about model capability,
and it’s not just about narrative either.

It’s about who can actually plug capability into workflows,
and who can make memory, permissions, integrations, and execution frameworks work.

Whether it’s AI agents
or stablecoins / tokenization,
the hard part isn’t building it.

It’s making it usable.

Narrative gets attention first.
Infrastructure determines the ceiling.

我越来越觉得，

下一阶段真正的竞争，
不只是模型能力，
也不只是叙事强弱。

更关键的是：
谁能把能力真正接进工作流，
谁能把 memory、权限、集成和执行框架跑起来。

不管是 AI agents，
还是 stablecoins / tokenization，
最难的都不是“做出来”，
而是“用起来”。

Narrative 先拿注意力，
infrastructure 决定上限。

[Liam｜AI Crypto Signals @liam_ai_crypto]

This is a sharp take.

A lot of people still frame the agent race as a model race,
but the next real battleground may be harness + memory.

Models can be swapped.
But once workflows and memory are controlled by someone else, switching costs and lock-in get much stronger.

For enterprises, this is no longer just a product choice.
It’s a control question.

这篇讲得很透。

很多人还在把 agent 竞争理解成模型竞争，
但下一阶段真正更关键的，可能是 harness 和 memory。

模型可以替换，
但一旦工作流和记忆被别人控制，迁移成本和锁定效应都会越来越强。

对企业来说，这已经不只是产品选择问题，
而是控制权问题。

[Harrison Chase @hwchase17]

https://t.co/aZnKZ1SVXB

[Liam｜AI Crypto Signals @liam_ai_crypto]

@Reuters This is a pretty clear signal:
once AI gets close to banks and financial infrastructure,
the question is no longer just “can it do it?”

It becomes:
who is accountable when something breaks, how do you audit it, and how do you control the risk?

[Liam｜AI Crypto Signals @liam_ai_crypto]

@OndoFinance I think the most important question for tokenization now isn’t whether it can be done.
It’s whether it can actually connect to existing financial workflows.
Issuance is the easy part.
Settlement, collateral movement, compliance, and liquidity are the harder parts.

[Liam｜AI Crypto Signals @liam_ai_crypto]

For a lot of companies, the problem with AI isn’t really model quality anymore.
The harder part is everything around it:
fitting it into existing workflows, handling permissions, connecting it to internal systems, and making it usable across teams over time.
When that part works, adoption scales.
When it doesn’t, even a great model can stay stuck in demo mode.

[Liam｜AI Crypto Signals @liam_ai_crypto]

多智能体不是越复杂越强，
而是要看任务类型和协作模式是否匹配。

很多团队的问题不是不会做 Agent，
而是一开始就把系统设计得太重了。

先从最简单、能跑通的模式开始，
跑出瓶颈，再逐步升级。

技术结构当然重要，
但真正决定能不能落地的，
还是业务场景、协作边界和验证机制。

[宝玉 @dotey]

https://t.co/CdLb5LM1YK

[Liam｜AI Crypto Signals @liam_ai_crypto]

@Bouazizalex @pmarca The real shift is that AI doesn’t just make teams cheaper. It gives them room to go after more customers, more workflows, and more output.

[Liam｜AI Crypto Signals @liam_ai_crypto]

@levie The bigger shift is that AI won’t stay chat-shaped. Once it moves into actual workflows, the token and infra requirements change fast.