By
–
真没见过哪个开源 Repo，能把 18 个 use case 塞进去还没乱套。
— 艾略特 (@elliotchen100) 10 mai 2026
架构方法、Benchmark、Memory 模式，一库打尽。 https://t.co/9z2G8VxrSv
I've never seen an open-source repo that can cram 18 use cases in without turning into a mess. Architecture methods, benchmarks, memory patterns—all covered by one library.
By
–
Anthropic's Thariq's article on HTML from yesterday blew up, with 1.5M reads. It looks like it's talking about formatting aesthetics, but actually he's describing an entirely new workflow. Picking out a few of the points with the most technical depth. First, HTML isn't a
By
–
Thanks to Brother Ju for the recommendation. Tanka is also an AI-native team under Shanda, where the memory function uses the system from @evermind
, allowing it to self-evolve and get smarter the more it's used. Tanka is an enterprise-grade IM, with a product philosophy of
By
–
Yesterday, Anthropic just released the API for this Dream, which is really fascinating. What it does is let the Claude agent "dream" on its own, organizing the memories it wrote down in past sessions—merging duplicates, updating expired ones, unifying contradictions, and even
By
–
This benchmark has already been downloaded 919 times, and in less than a week, the download count has increased by nearly 200, which illustrates two issues: 1. There's a high demand for self-evolving Agents.
2. The level of Twitter friends is quite high, wanting to test it
By
–
bro A new video just dropped, still super dry, still newbie-friendly, highly recommend it here, video link at the end. "I Tried Over 100 Claude Code Skills, These 6 Are the Best". 1. skill-creator Official from Anthropic, usage is straightforward to the max: Let Claude write
By
–
All related to dense attention, can be considered the same direction, with the same approaches.
By
–
确实很酷，拿 2900 万美金做 12M context，
— 艾略特 (@elliotchen100) 5 mai 2026
侧面证明了一件事：整个行业都开始相信稀疏注意力是 dense attention 的解药。
SubQ 走的是「重训一个模型」，属于垂直整合，风险大回报也大。@evermind 的 MSA 走的是「给主流模型加记忆」，属于 水平嵌入，谁的模型都能用。
另外，SubQ API 跟 SubQ… https://t.co/yZ8IugqquA
Yeah, that's pretty cool—using $29 million to build a 12M context window, which indirectly proves one thing: the entire industry is starting to believe that sparse attention is the antidote to dense attention. SubQ takes the "retrain a single model" approach, which is vertical
By
–
After watching reactions on X about the Cursor SDK all day, the most worthwhile thing to see isn't Cursor's own official statement. It's this perspective: the Cursor SDK has a fundamental difference from Anthropic and OpenAI's agent SDKs. Anthropic and OpenAI's agent SDKs are
By
–
We released EvoAgentBench two weeks ago, a benchmark specifically designed to test Agent capabilities. Unexpectedly, in just two weeks, it has already garnered 732 downloads on HF,
and in comparison, among all agent benchmarks on HF, it ranks second in this track. This is truly