Yang Zhou

Exploring physical and mental world. Physics PhD. AI evaluations. AI democracy.

[Yang Zhou @nbzy6666]

@mattpocockuk @0xtomdaniel Curiously, I do see you keep adrs, but not requirements. Is there a reason?

[Andrej Karpathy @karpathy]

We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks:

Writing. Something I've had success with: Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for "80% of the way to ASD-STE100" because the spec is quite stringent. But even better:

Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand. But even better:

Web pages. Ask for output "in HTML" to get a beautiful, interactive webpage. LLMs are getting really good at frontend and can create beautiful experiences, animations, etc. But even better:

Explainer videos. The output format I am most bullish on is fully custom / bespoke explainer videos generated on any arbitrary topic. Experiment with things like "Create a 3b1b style video explainer on X. Use my ElevenLabs API key for audio narration". (you'd need an API key for the latter or you can ask your LLM to find you decent free alternatives that use your local compute). This is actually starting to work!

In summary:
- As LLMs get better, they will do more and more of the legwork autonomously, and a lot more of our work will rise up the abstractions into oversight and understanding.
- Luckily, LLMs can help here too because as intelligence and code are increasingly abundant, you can ask for large, custom, discardable software artifacts (e.g. web apps, video explainers) that would have never made sense to create before. Push the boundaries here and you'll be surprised.

[Yang Zhou @nbzy6666]

@DavidPawlan me!

[Yang Zhou @nbzy6666]

@ylecun @ClementDelangue @Dan_Jeffries1 Can this help democratize AI? How to convince people?

[Garry Tan @garrytan]

This is the new standard for engineering evals

[Derek Feehrer @DerekFeehrer]

I just took this screen recording and turned it into a full product demo video in 20 minutes, using only one app.

3D animations, text, AI voiceover, music, and 3D gradient callouts to draw attention to the important parts.

But sure, keep posting Loom videos.

[GitTrend @GitTrend0x]

今天 GitHub 被自我进化记忆性 Agent 屠榜了了 🚀

5 个星标暴增最狠的项目，专业拆解下！

1. NousResearch/hermes-agent

自我进化记忆 Agent，彻底解决传统Agent 一次忘光光的痛点，通过动态patch 机制让 AI伴随你成长，越用越懂你的黑历史和偏好

🔗 直达 https://t.co/A0bGcD9p2q

程序员昨晚还在骂 Agent 像老年痴呆，今天它直接自己长脑子了，24h新增8800星。醒来发现AI比你还记性好，笑死谁了！

🟢

2. TauricResearch/TradingAgents
多Agent模拟完整交易公司，研究员挖数据、风控喊停、决策者互相撕逼后再下单，一干掉单Agent拍脑袋决策的血亏风险

🔗 直达 https://t.co/AVsxbPxi5G

你一个人操盘像赌狗，这个军团直接开董事会吵架决策，24h 新增 3900 星。程序员看完直呼终于有人替我背锅了！

🟢🟢

3. SakanaAI/AI-Scientist-v2

agentic tree search 驱动的全自动科研Agent，从提假设、跑实验到写论文一条龙，一个人直接顶 10人实验室

🔗 直达 https://t.co/p7M652PUYt

以前发论文要肝到秃头，现在AI自己组队科研，24H 新增 2000星。实验室老板看了沉默，PhD看了流泪，这下真失业了！

🟢🟢🟢

4. microsoft/agent-framework

微软官方 Agent 编排框架，支持Python+ .NET，一键搭建、生产部署多Agent 工作流，再也不用手写一堆LangChain 胶水代码

🔗 直达 https://t.co/0nVeay7i02

以前搭个复杂Agent要三天三夜，现在微软直接给军团指挥部，24h新增1800星——大厂程序员狂喜：终于能摸鱼了！

🟢🟢🟢🟢

5. block/goose

开源可扩展 AI Agent，超越代码补全，能自主安装、执行、编辑、测试任意LLM任务，程序员的终极“懒人外挂”。

🔗 直达 https://t.co/qb2eGN4NwM

你敲键盘敲到手抽筋，它直接自己跑完整流程，24h新增1500星——程序员集体高呼“哥们儿，这才是我想要的AI老婆”！

⚠️⚠️

总结

这 5个 Agent 项目从自我记忆进化到多Agent辩论、再到自动科研和生产部署，直接组成了一个能自己长大、自己吵架、自己干活的超级军团，GitHub上AI 不再是工具，而是你的创业+科研全栈合伙人，看完不转发你亏大了🔥🚀

[Yang Zhou @nbzy6666]

@BlakeHer_on @nurijanian Very important point! Have been building that recently!

[Yang Zhou @nbzy6666]

@goodside so where is Claude and gemini?

[Yang Zhou @nbzy6666]

@ArtemKRSV dangerous idea lol

[Yang Zhou @nbzy6666]

@deliprao beamer

[Michael Nielsen @michael_nielsen]

Is there an atlas of interesting experimental towns anywhere?

[Yang Zhou @nbzy6666]

@skdh I hope a more powerful AI can fix the plagiarism problem and the related.

[Yang Zhou @nbzy6666]

@skdh Losing jobs…

[Sabine Hossenfelder @skdh]

This is the maybe most surprising development in theoretical physics I’ve seen for a decade. A well-renown physicist is saying that there’s a mistake in a proof that dates back half a century, and black holes might not actually contain singularities.

https://t.co/ETfoFW4Rup

[Yang Zhou @nbzy6666]

#NYC #MPs #triggerpoint

[Yang Zhou @nbzy6666]

I will greatly appreciate it if anyone could recommend a good doctor in New York City who treats Myofascial trigger points.

[Yang Zhou @nbzy6666]

The Anatomy of Desk- Slouching (3D Animation) https://t.co/b0AO4O5U3P via @YouTube

best anatomy channel I found so far

[Yang Zhou @nbzy6666]

@michael_nielsen seems like a generalization of the renormalization group method…

[Jina will @jonawill15]

@SpaceX Falcon 9 landing never gets old.🚀💯