AI Bulls
140 posts

AI Bulls
@aibullss
Helping traders and investors apply AI in the financial markets.









STOP SCROLLING!!! You have less than 24 hours to use Claude Fable 5 without limits. Tomorrow, every session becomes paid. If I only had one day left, these are the 7 AI money-making tools I’d clone first. (ALPHA THREAD)

STOP SCROLLING!!! You have less than 24 hours to use Claude Fable 5 without limits. Tomorrow, every session becomes paid. If I only had one day left, these are the 7 AI money-making tools I’d clone first. (ALPHA THREAD)

这两天Claude Fable 5 不是回归了吗 一直想重新测试一下它的功能 看看它长任务的解决能力,所以我直接丢了一个要求: 做一款完整的中文网页游戏 《韭菜修仙:牛市生存模拟器》 参考《吸血鬼幸存者》的割草玩法,但世界观全部换成 Crypto: FUD Rug Pull 巨鲸 熊市 黑客 爆仓 升级还能解锁: BTC 信仰 ETH Buff AI Agent Diamond Hands 整个游戏生成完成之后,我自己玩了十几分钟。 说实话,比我预期还上头😂 —--- 更让我意外的,其实不是游戏。 而是我突然意识到一件事。 以前总觉得这种割草小游戏开发成本应该很高。 但 AI 把整个流程跑完之后,我发现: 真正复杂的并不是玩法。 而是素材、美术、数值和平衡。 如果把现在游戏里的这些方块、小卡片,全部替换成角色、NFT、链上资产或者项目 IP,其实就是另外一款游戏了。 甚至完全可以做成: ✅ 链上交互小游戏 ✅ NFT 活动小游戏 ✅ 项目方任务游戏 ✅ Web2 爽游 玩法几乎不用推倒重来。 AI 已经把最难的那部分搭好了。 —--- 这次整个任务也确实挺"重"。 最终消耗: 📥 Input:33,603 Tokens 📤 Output:33,412 Tokens (大概是在2U左右) 基本相当于让模型完整规划、设计并生成了一整个小游戏。 也难怪大家都说 Fable 5 更适合这种长任务。 —--- 这次我没有去官方 Claude 跑,而是在 ZenMux 上体验的 Claude Fable 5 原因也很简单: Free 层就能直接体验 Fable 5 PAYG 账户只要 Credits > 0,就可以调用 claude-fable-5-free 后面如果想横向测试,还能直接切 PK 模式,同时对比 GPT-5.5、Opus 4.8、Gemini 等多个模型,不用来回切平台 另外,现在 ZenMux 的 Auto Top-up 活动还在进行中(到 7 月 15 日),自动充值每笔还能额外赠送 20% 余额(最高可多获得 $300 额度 ) 我下一步准备试一个更大的任务 不是做游戏 而是看看 Claude Fable 5 能不能直接帮我完成一个完整的 Web3 小产品。 如果效果不错,再来分享全过程

The smartest engineers on Wall Street quietly gave away some of their best tools. Most people have no idea they exist. Here are 22 production-grade tools from top quant firms that you can use for free: Two Sigma • Flint → Lightning fast time series joins on Spark github.com/twosigma/flint • BeakerX → Multi language Jupyter notebooks github.com/twosigma/beake… • Cook → Production grade job scheduler used by companies like Twitter and Apple github.com/twosigma/cook Man Group • ArcticDB → High performance time series database licensed by Bloomberg github.com/man-group/Arct… • D-Tale → Turn pandas DataFrames into an Excel-like web app github.com/man-group/dtale Jane Street • Core → Their custom OCaml standard library github.com/janestreet/core • Magic Trace → Nanosecond level CPU tracing github.com/janestreet/mag… • Hardcaml → Design hardware using OCaml github.com/janestreet/har… D.E. Shaw • Pyflyby → Automatic imports for Jupyter github.com/deshaw/pyflyby • Versioned HDF5 → Git style versioning for massive datasets github.com/deshaw/version… Hudson River Trading • Corral → Modern structured concurrency for C++ github.com/hudson-trading… More gems • Optiver’s nanosecond timestamp library github.com/optiver/timest… • Community implementation of WorldQuant’s 101 Alphas github.com/yli188/WorldQu… The firms that open source don’t see great engineering as a competitive weakness. They see it as a recruiting advantage. Bookmark this. I share hidden AI tools, GitHub projects, and resources like this every week. Join my discord now: discord.com/invite/alphaca…









