Changsheng Yang

[Bilgin Ibryam @bibryam]

- Why ClickHouse is best observability layer for agentic workloads?

- ClickHouse CEO:
"Applications were designed for a specific persona that had a role within an organization and their query patterns were very predictable … agents don’t have personas.

Agents move across all these systems, issue dozens of SQL queries concurrently, and explore data through less predictable query patterns. The entire agent workflow is limited by its slowest system, so latency becomes the primary requirement."
https://t.co/YP7IuRCRYt

[The Year of the Graph @TheYotg]

How to test a Knowledge Graph

A graph doesn't crash when it's wrong. It just returns a plausible-looking number, and the agent sitting on top turns that number into fluent, confident prose. A wrong 45 reads exactly like a correct 45 — which is the actual argument for building a test suite before trusting one.

The framework rests on two old ideas dressed up for graph work:

📊 Competency questions — acceptance tests with a human-verifiable answer. 

"What does Alphabet own, directly or through intermediaries?" has a checkable answer in Exhibit 21 of the 10-K. It turns out to be just three entities, and Google LLC surfaces at two hops, not one, since the 2017 restructuring routed ownership through XXVI Holdings. A wrong edge shows up as a wrong hop count.

📊 Invariants — patterns the domain forbids, written as queries that must return zero rows. 

No company owns itself, directly or transitively. Segment revenues must reconcile with the filing total. One of these caught a real discrepancy on correct data: FY2025 segments summed to $402,963M against a $402,836M total, because hedging losses net out at the total line. A "rounding tolerance" carried over from a previous project would have silently swallowed that $127M gap.

A third layer, count tripwires, compares node and relationship counts against the prior ingest, catching the kind of silent parser breakage that no single query would flag.

There is also a declarative alternative: SHACL shapes with Cypher targets, via Jesús Barrasa's Going Meta work. 

Where the whole approach breaks down: invariants only catch what you anticipated, competency questions rot if answers get updated to match output instead of source, and none of this catches an extraction pipeline that silently dropped a third of the edges.

As LLM extraction becomes the default way graphs get built, verification becomes the differentiator. The team that ships a graph together with the suite proving what it gets right beats the team that ships a bigger graph.

By Konrad Kaliciński

https://t.co/gnoKuAEm66

#KnowledgeGraphs #Neo4j #DataQuality #GraphDatabase #DataEngineering

--

🤝 Put your graph tech brand in front of the people who matter

Your graph technology deserves to be seen by buyers, analysts, and builders who are actively shaping the space. 

The Year of the Graph is the independent hub that this community trusts.

Slots for the upcoming Autumn 2026 Issue are filling fast. Reach out and book yours now 👇

https://t.co/LCvoLhhBBt

[Jesús Espino @jespinog]

I published today an article at the @VictoriaMetrics's blog, sharing the life of a metric. I think it is a good walkthrough about what happens to your metrics once they are scraped, until they gets discarded by the retention policy.

https://t.co/zaATa4mj1u

[Changsheng Yang @csyangchsh]

@rponte @codecondev Thanks! Can I preview this course https://t.co/KnuJY3UuCg? Do you have English version?

[Joe Ingeno @JoeIngeno]

Performance of WebAssembly runtimes in 2026 https://t.co/pJhPGNtqo7 #webassembly #wasm

[Clu @thexclu]

五月自驾的时候，@liuyi0922 给我打了个电话，说我们每天用微信读书看书，刚刚微信读书出了 Skill，现在我们可以做个App来管理和分享划线了。

于是第二天，我就做了Readmate 的第一个版本。本来觉得这是两天能搞定的Side Project。后来我自己断断续续用了几个月，阅读时间居然也越来越多。给身边的人用了一下都觉得不错，就继续认真打磨它的体验。

以前有个朋友跟我聊读书时提到一个说法我记忆��新：读书，是一种和作者、和经典对话的方式。后来我觉得，划线大概就是这些对话留下来的痕迹。

有些已经忘了，有些再读的时候，理解已经完全不同。所以有了现在的 Readmate。

把微信读书里的书和划线带回来，重新翻，也让它们偶尔通过 Widget 和提醒再次相遇。

今天终于上线了。 有些句子，值得再读一遍。
Some lines stay with you longer than the book itself.

https://t.co/lgcV0ha3yp

[Bilgin Ibryam @bibryam]

.@cnakazawa's lessons from coding with agents: the faster they code, the more these matter:

• strong ownership
• product taste
• guardrails and feedback
• repo-local context
• stack ownership
• option value
https://t.co/XqxCo8TWUi

[alex shapalov @alxshp]

Launched today: https://t.co/TC2PHN6PnX 📕

Postgres Book, one topic at a time!

No 1000 page manual. Pick a topic, understand how it works, and try the examples yourself.

Learn:

• Indexes
• Transactions
• JSONB
• Window functions
• Row-level security
• Vacuum
• Locks
• Replication

$ pgbook read locks

Free and open source. No signup required.

→ https://t.co/TC2PHN6PnX

[Mihail Eric @mihail_eric]

I’m excited to finally announce the newest edition my Stanford course 𝗧𝗵𝗲 𝗠𝗼𝗱𝗲𝗿𝗻 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿. It has been 9 months in the making.

Last November, with the release of Claude Opus 4.5, coding agents experienced a step function improvement in capability. We all felt it. The LLMs were more powerful, could reason for longer, solve harder tasks.

This year’s iteration of my course reflects the 2026 metamorphosis of software engineering.

My core belief is simple: AI-native developers of the LLM era are going to become the most important members of any software organization. I have designed my course to train this next generation of engineers.

𝗪𝗵𝗮𝘁’𝘀 𝗱𝗶𝗳𝗳𝗲𝗿𝗲𝗻𝘁 𝘁𝗵𝗶𝘀 𝘁𝗶𝗺𝗲 𝗮𝗿𝗼𝘂𝗻𝗱

First, 85% of my Fall 2025 class material is being thrown out. The Fall 2026 syllabus reflects the core capabilities AI-native engineers must have: agent skills, advanced context engineering, MCP portals, agent-ready codebase principles, agentic code review, security, parallelizing background agents, software factories, and more.

Second, I am going to teach my students how to have software taste. Every student will be required to ship pull requests to production-grade, real-world codebases. The course is collaborating with the top open-source AI repos who will offer support and mentorship to students on how to meaningfully contribute to their projects.

This has never been done before in any university course so I am incredibly grateful to our OSS Partners: @browserbase, @HeyGen, @CopilotKit, @semgrep, @OpenHandsDev, @milvusio, @marimo_io, Pi, @crewAIInc, @warpdotdev, @vercel, @cmux, @arizeai, @UnslothAI, and @anyscalecompute.

𝗪𝗵𝗮𝘁’𝘀 𝘀𝘁𝗮𝘆𝗶𝗻𝗴 𝘁𝗵𝗲 𝘀𝗮𝗺𝗲

I’m fortunate to again have AI software engineering leaders and founders as guest speakers to share their learnings from building top coding agent products. Thank you to @leerob from @cursor_ai, @bcherny of @claudeai code, @EnoReyes of @FactoryAI, @silasalberti of @cognition, @0xine of @semgrep, Rajesh Bhatia of @Cloudflare , @amasad of @Replit, and @eladgil.

All resources will be available online. All classes will be available to the public.

9/22 on Stanford campus. See you in class. 

https://t.co/wTokHyUMsz

[Rock the JVM @rockthejvm]

New video on Algebraic Data Types (ADTs) in Java:

https://t.co/gljFKRHaeS

[JAVAPRO @JAVAPROmagazin]

Ever upgrade #Java & realize most features barely affect daily coding? #Java25 quietly removes friction developers hit every day. M. Gheorghe-Roman explains which features matter in real projects — from pattern matching to switch logic: https://t.co/iGGXI5CfHL

#OpenJDK @OpenJDK

[灰狐 @huihoo]

Open-Source Computer Architecture Research (OSCAR)
开源硬件研讨会旨在探讨硬件和软件工程师在应对下一代计算机架构日益增长的异构性时所面临的各种挑战

5届研讨会幻灯片资料和开源项目均可下载和访问
https://t.co/6FYoNSkqc3
https://t.co/P86IRMvZz4
https://t.co/tItk0sNUNM
https://t.co/XleMElpjDy
https://t.co/Chuh9T4DmZ

[Baeldung @baeldung]

DB-Scheduler: A Persistent, Cluster-Friendly Scheduler for Java https://t.co/jQv7ZA9feo

[JCON @jcon_conference]

Our next #JCON2026 session is live: 'Java Performance Analysis and Optimization: #AI Powered Diagnostics for #Modern Applications' with @sascha242 & Yuriy Bezsonov

This #code talk demonstrates building a production-ready #Java performance …

Grab your coffee and hit play: https://t.co/UF9Lw8gjJS

[netdevconf @netdev01]

All materials for @netdev01 0x1A now out. Visit: https://t.co/4hkRpiXewG Enjoy! #netdevconf

[Iven ʕ•̬͡•ʔ✧ @ivenvd]

HN 上这篇文章我严重认同，虽然我的主要工作内容是做效率工具，不过我还是觉得，良好的团队氛围远比工具本身重要。同事之间互相信任，保持沟通，任务进度有效同步，共同学习提升，每个人能做自己想做的工作，这些才是效率提升的本质，而工具只是在此基础上解决具体瓶颈的。 https://t.co/5ucfJ1eP5y

[deararchitects @deararchitects]

#305 – https://t.co/zELtcoJq8W

- #Architecture as a Socio-Technical Craft
- You can’t govern what you can’t see by @lucamezzalira
- Making Your Data Ready for #Agentic #AI @martinfowler
- The end of #programming
- Running a Software Factory Efficiently at @UberEng Scale

[InfoQ @InfoQ]

Johannes Schickling explains why he moved beyond the traditional three-tier web stack to a local-first architecture while building Overtone.

🎧 Learn how client-side event sourcing with SQLite compares to CRDTs: https://t.co/TmcknIoOXC

#LocalFirst #SQLite #EventSourcing

[Spring I/O @spring_io]

🍃 Demystifying Observability for Spring Applications by @salmto / @tiffanyfayj @ Spring I/O 2026

▶️ Video: https://t.co/Haq3TEgZ8S

🔗 Slides: https://t.co/GOVZTEE3JG

🛠️ GitHub repo: https://t.co/ALt86nCJlR #springio26