patrick sphinx

[CMGS @CMGS1988]

@mdjxyz 你不已经在基建了…当然学点硬件的调度编排就更好了……别管什么 AI 基建不基建，infra 无外乎就那点东西，计算，存储，网络……你搞推理，多个 GPU 调度，驱动都喂到嘴边 vGPU 和 K8s 搞吧搞吧又不是不能用是吧……

[Sebastian Aaltonen @SebAaltonen]

We doubled down by creating exceptions. Exceptions are abstracted gotos. Hidden in the code base deep inside call stacks, jumping over multiple functions at once. Exception-safe coding is identical to goto-safe coding, except that goto-safety is local and explicit. Easy to prove.

[Aram Hăvărneanu @aramh]

@Aron_Adler Everything worth doing is, in fact, doable. Undecidable problems do not have any sort of economic importance.

[LemonHX @lemonhxmoe]

因为这些面试大概率我都参与评审，所以说我可以大概讲一下我问人的指标，然后如果是应届的同学呢，可以好好去学一下，以免自己可能会错失自己心爱的岗位

第一点就是如果使用RAG技术，那么你是怎么做你的系统的测评的？我这里的希望就是面试者能够回答出来对于测试数据的选择，true positive，fp，tn，fn，然后面试者应该能够说出来，比如说在测试一个RAG系统的时候需要准备精心挑选的有标准答案的非标数据集，比如说上千个问题，然后进行怎样采样怎样评测评测几轮，最后分析出来哪些是模型幻觉、哪些是检索精度问题等

如果面试者他真的不是特别懂AI我的提问方向会是他们现有的技术里面所有跟自动化相关的技术比如你会不会写CICD比如你的测试是怎么写的？测试环境是什么？

最后一个部分可能会问面试者项目管理相关的经验，因为假设说你真的不懂AI，但是你真的能够管理一群脑子不太灵光，然后不太听话的下属有这种经验，那么你大概率构建的harness比正常人要好。

[SWH | (168, 168) @swh16888]

四騎士之一 是導入ai 速度變慢,

O社A社都要找forward deploy engineer 還有跟Mckensy / Accenture合作 ,幫客戶導入AI。

都是因為fortune 500 客戶技術能力不夠, ai transformation 進度不夠快, 過去的low hanging fruits 摘完了。現在要走地面戰, 一個客戶一個客戶推進。

Ai 撞到了人類社會的組織結構。領導說要用ai , 但是基層都是中登老登, 學習能力普通不會用ai。用量不夠多 anth oai arr 成長速度就會慢。

[tison @tison1096]

太幽默了，还有这种事情。测试了一下还真是。

[Vishal Misra @vishalmisra]

@ShriramKMurthi @Hesamation Have you read this https://t.co/tabFRNNsbD

[Joe Gibbs Politz @JoePolitz]

Preparing these lecture notes is massively entertaining and instructive to me. I am still not fully untangled on what the end lesson will be here.

(Still in progress, but writing up the notes as I get to each next lecture, so publishing increments)

https://t.co/gaerN7XLCr

[geoff @GeoffreyHuntley]

re: laughed out of the room.

this is why i think we will get a revenge of the PLT nerds event. there is so many nerds who have been suppressed with this train of thinking that are very close to “fuk it, it’s easy now. hold my beer and weep gatekeepers of yesteryear”

ie: fringe stuff like https://t.co/DqtovCSPZB

[geniusvczh @geniusvczh]

gpt写代码特别喜欢吃掉异常，而我们的目标是尽量让程序尽可能多的崩溃，这个靠prompt是做不到的，只能安排test case把每一个点都引爆一次🤪

[Yihao Sun @StarGazerMiao]

After 6 years, finally be able to ends my last chapter of the last paper in PhD time with the same style title of the first paper I read in first day of my PhD confidently.

[Maxwell Brown @imax153]

We on the @EffectTS_ team we often recommend cloning the Effect repo into your project so your agent can explore the source directly.

I finally wrote up why it works and how to set it up:

https://t.co/qxV9o6I6zM

[LemonHX @lemonhxmoe]

https://t.co/UJuAq7M1NC

大家可以看看我是怎么在AI时代用最现代的手法干PL的：

- SubAgent
- Context Engineering
- Harness

反正什么都有！

[Taelin @VictorTaelin]

To whom it may concern

NanoProof.hs: the smallest viable proof checker

I posted something similar before, but it was more of a research experiment with weird λ-encoded shit, than something usable. This new repo contains a tiny, 1000-LOC Haskell self-contained proof checker that you can actually use to prove arbitrary theorems.

The language has just 6 base types:
→ Empty (`⊥`): type with 0 elems
→ Unit (`⊤`): type with 1 elem (`()`)
→ Bool (`𝔹`): type with 2 elems (`0 | 1`)
→ Sigma (`ΣA.B`): dependent pairs (`(x,y)`)
→ Pi (`ΠA.B`): dependent functions (`λx.f`)
→ Equal (`a==b`): propositional equality (`{==}`)

That's all you need.

Each of these is needed, as it introduces something fundamental. The file includes a parser, stringifier, equality, a bidirectional type checker, and a simple CLI. It also includes first-class reduction relations, which allow us to pretty print goas just like Lean. You can place '()' in a position to inspect the current context and goal there. I also include a demo proof for the commutation of multiplication.

[Talia Ringer 🕊🪬 @TaliaRinger]

Our synthetic Euclidean Geometry proof assistant is now open for public contributions! We created this together as a class in my Build Your Own Proof Assistant course this semester. Please give it a spin and consider contributing!

https://t.co/DEPA68NXmm

[Suwako — e/acc @suwakopro]

各个LLM在ProgramBench上的ast-grep测试的通过率好低啊 @hd_nvim

[[email protected] | (spec|ver)ification | security @k0nn0v]

11. Last but not least, George Pîrlea's @GeorgePirlea talk on Veil: Multi-Modal Verification of Transition Systems

...and this is done with Lean @leanprover !

https://t.co/dRjgvH8dLS

[Kiran @kirancodes]

5 lines of python. an economic game with complex equilibria.

Our new language Pact uses Choreographies with game theory, allows expressing economic transactions in lines. So simple an agent could write it

Claude? make me some money. and make no mistakes

https://t.co/m3PjkWMK98

[patrick sphinx @SphinxPatrick]

@wong_ssh 可以看看这个公司在做的https://t.co/UxHzhRVPnh

[patrick sphinx @SphinxPatrick]

@wong_ssh dafny之类的语言如果真让developer手写感觉太繁琐了，基本上人脑要随时想哪个地方有什么条件要满足，不知道program一旦大了还能不能proof能不能跟上