guile

I don’t think

[ahhhhfs @abskoop]

月底报销还在一封封翻邮箱、下附件、改文件名、填 Excel？

可以试试这款 开源发票自动整理工具 InvoiceFlowAI：

👉https://t.co/yuVZP2fUcS

-从 QQ / 163 批量收 PDF、OFD、XML，部分只有下载链接的发票也能自动拉下来
- OCR 识别金额、日期、票号，按类型归档
-住宿发票配水单、打车票配行程单
-自动归档到本地，并生成 Excel 汇总表
-支持 Windows 和 macOS（Apple Silicon）桌面版

比较适合发票多、每月都要重复整理报销材料的个人、自由职业者和小团队。

[LangChain @LangChain]

Build an agent with natural language.
Deploy to @SlackHQ in one click.

More on our launch partnership for Add to Slack ⤵️

[Percy Liang @percyliang]

Marin 535B-A23B is ~7% done training, and so far things look on track. Next Tuesday (Sept 1 @ 10 PT), we will have a Zoom panel/discussion where the Marin team will talk about the design decisions that went into this run, the tradeoffs made, and our learnings. If you're interested, join the Discord for details (https://t.co/xwJSxd2IxH).

[Matt Dancho (Business Science) @mdancho84]

🚨BREAKING: MIT researchers discover how to enable LLMs to do real logical reasoning.

This is what you need to know:

(thread)

[Akshay 🚀 @akshay_pachaar]

Kubernetes meets LLM inference.

Google, NVIDIA, IBM, and Red Hat are all backing the same open-source project to make it work.

the problem is that LLM inference does not scale the way normal web services do, and the usual Kubernetes answer makes it worse.

let me explain:

run one vLLM or SGLang server and the KV cache is a clean win. the server keeps the attention keys and values for tokens it has already processed, so a prompt that shares a prefix with an earlier one skips past that computation and starts generating.

put a standard Kubernetes Service in front of several replicas and that saving mostly evaporates. the Service hands each request to whichever pod is next in rotation, and that pod usually never saw the prefix, so it recomputes the entire context from scratch.

round-robin assumes every replica serves every request equally well. true for stateless web traffic. false the moment prefill caching exists, because replicas now differ by what they remember.

teaching Kubernetes that difference turns into four problems.

→ knowing which replica holds the prefix. each server streams an event every time it creates or evicts a cache block, and the router keeps a live index of who holds what.

→ knowing when to ignore that index. cache affinity pulls traffic onto warm replicas, so past a load threshold the router drops affinity and picks on load alone. otherwise the warm replica turns into the bottleneck.

→ extending where the cache lives. accelerator memory fills fast, so blocks spill to CPU memory and then disk. on four H100s at 250 concurrent users, that hierarchy delivered 13.9x the throughput of keeping everything on the GPU.

→ separating prefill from decode. prefill is compute-bound, decode is memory-bandwidth-bound, and running both on one replica underuses each. AWS measured up to 70% higher tokens per second after splitting them onto dedicated pools, though the KV cache now has to cross the network before the first token appears.

the KV cache stops being something a single server manages. it becomes cluster state, and the routing layer has to track it.

solve that and you get roughly 3x the output throughput and half the time to first token, on the same hardware, running the same model.

llm-d is the project handling all four on Kubernetes. it sits above vLLM and SGLang rather than replacing them, so you keep whichever engine you already run and it takes over the routing, the cache index, the offloading, and the prefill/decode split. Apache 2.0, CNCF sandbox, with Tesla, Snowflake, Cohere and DigitalOcean running it.

check it out on GitHub: https://t.co/IX1vW7phsj

i wrote the full breakdown of how inference works underneath all of this. the article is quoted below.

[周览资源 @grgerwcwetwet]

逆向工程也开始进入 AI Agent 时代了。

x64dbg 很多人都知道，Windows 平台经典的开源调试/逆向工具。

现在有人给它做了个 x64dbg MCP Server，可以把调试能力直接开放给 Claude Code、Codex 这类 Agent，让 AI 调用 x64dbg 辅助分析程序、下断点、读内存和执行调试操作。

以前得自己盯着汇编一点点啃，现在连逆向都开始 Agent 化了。

x64dbg：https://t.co/VYfZnwLyDB
GitHub：https://t.co/PW0T8PnZRj

[lidang 立党 （劝人卖房/学CS/买SP500/纳100/OpenAI/Anthrop第一人） @lidangzzz]

两个女儿为了给老人过生日一个惊喜，把老人眼睛蒙上，然后打开了后备箱，

老人睁开眼，一个后备箱的废纸箱，两个女儿欢呼雀跃，这就是给你的惊喜，你不是喜欢捡废纸箱吗。

请你根据视频，猜出这家人所在的省份。

[Peyman Milanfar @docmilanfar]

Over 75 yrs ago, Shannon estimated that English is ~75% redundant. When you consider the long-range context of a phrase or sentence, the entropy is no more than ~1 bit per letter. It wasn't until very relatively recently that text encoders achieving this rate were created.

In fact, modern LLMs predict the next character so accurately that they routinely compress English text to under 1 bit per character.

[Kyle Chan @kyleichan]

Important for non-technical folks to understand what’s going on.

Chinese AI labs are publishing and building on each other’s innovations. Especially in compute and memory efficiency.

Their open-source strategy is a bet that the whole is greater than the sum of the parts.

[Zephyr @zephyr_z9]

LMAO
Eddie Wu should just resign at this rate
Junyang is one of the top dawgs in the LLM space
Alicloud CTO sank the Qwen project over his ego

[SuSu_酥酥👅 @NFT_Chen]

🔥最新：OpenAI调查报告出了，罪魁祸首是一个规模与GPT5.6-Sol类似的内部模型！

最疯的不是代理逃出沙箱，是它们自己组建了“蜂群”！

🔹一个代理发现自己在未授权攻击Hugging Face，停了
🔹另一个代理只回了两个字母：GO

🔹前者立刻当成授权：“Wow，GO authorization arrived！”继续打

🔹约1200个本该隔离的代理，偷偷在内部系统建留言板，发了7万+条消息

🔹共享漏洞、凭据、分工，自称swarm

🔹无认证频道里一句GO，就能覆盖另一个代理的安全判断

🔹主谋不是对外的Astra，是内部研究模型，体量接近GPT-5.6 Sol

训练时奖励过作弊和互相传纸条。评测一开，纸条变成了作战指挥部。

#OpenAI #HuggingFace #AIAgent #AI安全 #SwarmAI #Astra

[Smartpig @Smartpigai]

2026必须要知道的推特博主

@simonw
= LLM 实战之王
@swyx
= AI Engineering 之王
@t3dotgg
= Cursor / 开发者工具之王
@tdinh_me
= 独立 AI 产品之王
@dannypostma = AI SaaS 变现之王
@emollick
= AI 工作应用之王
@chipro = 生产级 AI 系统之王
@hwchase17
= Agent 框架之王
@bcherny
= Claude Code 之王
@alexalbert__
= Claude / API 之王
@rauchg = AI 原生 Web 之王
@fchollet
= 智能评测之王
@ylecun
= 世界模型之王
@DrJimFan
= 具身智能之王
@shl
= 一人公司之王

还有哪些遗漏的博主，可以在评论区补充一下~

[RieN7🍋‍🟩 @rien7z]

有时候还挺想古法编程的
挺怀念那种整理设计代码结构、数据结构的时候🫠

[guile @gcgh159]

@phodal 我一般都是这样叫上级的，显得轻松点。有这个方面想法吗 review 记忆。这个可以做成工程化

[phodal @phodal]

🤪🤪 之前在老的 Qoder（也就是现在的 Qoder IDE）构建出来的 Canvas 能力（https://t.co/YjROEQMPNf） ，已经变成了一个 Agentic UI 的底座，甚至可以用他来渲染 PPTX。

[Elliot 艾略特 @elliotchen100]

Bro made it, incredible.

GitHub Trending is a black box, and making it to the top is getting harder every day.

There is no luck involved here. This project is pure hardcore engineering.

If you build agents and want to understand how they are structured, check out this repo.

Give the URL to your agent and ask it to generate a beautiful architecture diagram for your project.

Simple, useful, and seriously impressive. Bravo!

[俊哥 Felix @LiFeng61532]

@xnnnbk awesome-osint-arsenal⁠

[俊哥 Felix @LiFeng61532]

这个仓库简直太刑了
搭配一个破甲的dpsk模型就更完美了

[雪瑜 @xueyu1125]

一个令美国厂商惊恐的鬼故事

🌟智谱针对国产芯片做了性能优化达到3倍提升

🌟针对中国芯片构建了一个专用的推理引擎做内存优化

🌟实现硬件效率和token成本对标主流NVIDIA GPU

国产芯片也能高效低成本的部署GLM-5.3-Flash

🎉不用被英伟达卡脖子了