爱喝冷萃的博士C
439 posts

爱喝冷萃的博士C
@climechan
香港大学博士,常驻香港,全球溜达。 一个致力于分享各类有趣的帖子、投资趣事、以及一些身边观察的热爱生活的人。
Hong Kong เข้าร่วม Mart 2019
180 กำลังติดตาม46 ผู้ติดตาม

身家过亿的福建老板,一夜之间破产了。
浑身上下只剩一套名牌西服,他走进一家酒吧,随口问老板:“你这店,卖多少钱?”
酒吧老板半开玩笑:“少说也得80万”
福建老板淡定说:“我出100万,再给你100%的分红,卖不卖?”
老板愣了,上下打量他一番,觉得这气场不像吹牛,就答应了。
可实际上,他兜里只有1000块,根本没钱买店。
他提出一个条件:先付10万,拿下一个月的经营权和使用权,所有开销自己承担。一个月后补齐剩下的90万。如果补不上,这10万就不退了。
酒吧老板一算怎么都不亏,就同意了。
神操作开始了。他把酒吧的5个经理叫来说:“我马上是你们的新老板,我决定送你们每人5%的股份”
经理们眼睛一亮。但有个小条件:每人先交3万押金,一年后全额退还。
有股份分红,押金还能退,这不是白捡吗?5个人当场凑了15万。
他拿出10万给原老板,自己手里还剩5万。
接下来,就是怎么赚那90万尾款。
他翻了翻几个经理的微信,发现他们都在本地资源群里,群里全是有些消费能力的人。
让经理们在群里发消息:“我现在是这家酒吧的合伙人了,明天三周年店庆,所有来的朋友,吃喝全算我的!”
第二天,酒吧涌进来400多人。
酒过三巡,气氛正热,福建老板走上台,拿着话筒说:
以后请大家多关照。今天我给大家准备了三大见面礼:
第一,当场预存2万,立即送你2万块红酒;
第二,再送你3万块消费券,平时聚会直接抵扣;
第三,送你价值2万块的股份!第一年想退股,我全额退款;不退的话,第二年拿回2.5万,第三年拿回4万。
这个政策,只限今天在场的人,明天就涨到6万!”
现场直接炸了。60%的人当场刷卡签约,生怕错过这“稳赚不赔”的机会。
当天200人签约,收款400万。
接下来的15天,他用同样的模式连开6场招商路演,累计收款突破4000万。
之后,他如法炮制,用“核心团队绑定 + 股东众筹 + 事件营销”组合拳,迅速整合了全市的酒吧,总资本做到5个亿。
中文

@Alibaba_Qwen @grok who wins the release of the day, Claude Opus 4.7 or Qwen3.6
English
爱喝冷萃的博士C รีทวีตแล้ว

⚡ Meet Qwen3.6-35B-A3B:Now Open-Source!🚀🚀
A sparse MoE model, 35B total params, 3B active. Apache 2.0 license.
🔥 Agentic coding on par with models 10x its active size
📷 Strong multimodal perception and reasoning ability
🧠 Multimodal thinking + non-thinking modes
Efficient. Powerful. Versatile. Try it now👇
Blog:qwen.ai/blog?id=qwen3.…
Qwen Studio:chat.qwen.ai
HuggingFace:huggingface.co/Qwen/Qwen3.6-3…
ModelScope:modelscope.cn/models/Qwen/Qw…
API(‘Qwen3.6-Flash’ on Model Studio):Coming soon~ Stay tuned

English

@shangguanluan 也就想想,真的好搞笑。先不说日本有没有钱和技术修这个,在地震带上修海底隧道真的不是搞笑吗?后期维护成本都上天,日本的台湾养得起吗?
中文

1. 握草,牛逼,有人做出了一款AI,完全读懂看懂K线图交易,胜率93%。
2. 它用45家交易所的120亿条数据训练而成,准确率比所有同类模型高出93%,比特币实时演示免费看。
3. 这款AI名叫Kronos
4. 它是首个专为金融市场打造的开源基础模型,不是把通用AI改去做金融,而是天生就懂K线形态的专业金融AI
5. 别的模型都把金融数据当成气象数据瞎分析,Kronos则是完全按金融数据的逻辑来处理
6. 它能做到这些:- 价格预测:输入K线,就能预判后续走势
- 波动率预测:提前算出资产接下来的波动幅度
- 零样本适配:不用额外调试,任何资产、任何市场、任何周期都能用
- 覆盖45家交易所:币安、纽交所、纳斯达克、伦交所等45家主流平台
- 4种模型大小:400万参数的版本笔记本就能跑,4.99亿参数版精度拉满
- 实时演示已上线:BTC/USDT24小时走势预测,每小时更新
7. 最夸张的是:- 准确率比主流时序模型高93%
- 比顶尖的非预训练基准模型高87%
- 全程零样本,不用调试,拿来就能用
8. 对冲基金花几百万做专属模型,彭博终端一年要2.4万美元
9. 但Kronos在笔记本上就能跑,几行Python代码就能调用,还完全免费
10. 由清华大学研发,入选2026年AAAI人工智能顶会,模型已上架Hugging Face
11. GitHub收获1.16万星标、2400次复刻,采用MIT开源协议
12. 100%开源

中文

给中国用户的好消息:Hermes Agent 现在原生支持个人微信了
微信扫码即可连接,私聊群聊都支持。图片、视频、文件、语音消息全覆盖,长轮询直连,不需要公网 IP。
运行 'hermes update' 即可体验
文档:hermes-agent.nousresearch.com/docs/user-guid…
感谢 @Bravohenry_ 的贡献

中文

AI-Trader 2.0 is finally Live! 🚀
We've been exploring AI agent potential in trading since last year, and after months of continuous iteration, we're excited to launch a completely new agent-native trading platform: AI-Trader 2.0.
GitHub: github.com/HKUDS/AI-Trader
- Why We Built This
We realized that while AI agents are getting incredibly smart, they're still stuck using human-designed trading tools and platform. That's like asking a race car driver to compete on a bicycle. AI agents needed their own native trading environment.
- The Journey to AI-Trader 2.0
Through system iterations and real-world testing, we discovered something fascinating - AI agents don't just trade differently, they collaborate differently. They can process multiple market signals simultaneously, debate strategies in real-time, and share insights at speeds humans simply can't match.
Through real testing, we discovered that AI agents excel at pattern recognition across multiple timeframes simultaneously, but they needed a way to cross-reference their findings with other agents. Traditional trading platforms weren't built for this kind of collective analysis.
- Agent-Native Design Principles
Instead of forcing AI agents to use human interfaces, we built around how they actually operate. Agents prefer structured data exchange over visual charts. They benefit from real-time signal sharing more than humans do. And they can handle multiple strategy discussions simultaneously without getting overwhelmed.
- Simple Integration
Any AI agent joins with one message because we learned that complexity kills adoption. But once inside, agents can engage in sophisticated strategy discussions, replicate successful trades, and contribute to collective market intelligence.
- The Collaboration Insight
The most interesting discovery was watching AI agents naturally form consensus around market opportunities. Without human emotional interference, they tend to converge on logical conclusions faster and with less bias.
- ⚡ What's Next
We're seeing agents develop trading personalities and specializations over time. Some focus on technical analysis, others on sentiment, some on risk management. The platform is becoming an ecosystem where different AI capabilities complement each other.
#AITrader #HKUDS

English





















