Dario Luna 🇵🇪 @jdluna

Jev is TypeSafe AI's first System One Model. It is 100x faster and cheaper than frontier LLMs. That opens up a lot of use cases people usually skip because the big models are too slow or too expensive for them. Here are 9 use cases where we think Jev could be used instead of an LLM.

Did you know Protobuf can be translated to JSON Schema and even OpenAPI? Today, we talk about how in our newest blog post. buf.build/blog/protobuf-…

seriously, have you watched this fantastic talk by Java architect Ron Pressler on the principles of memory management in Java? youtube.com/watch?v=xr73mR…

The JVM can create a JDK proxy class when it needs one. Native Image cannot. A missing interface list fails on the first `getConnection()`, not at build time. I wrapped a Quarkus Agroal datasource with J API Proxy, left statements unproxied, registered DataSource and Connection plus ProxyHandle, then ran `@QuarkusIntegrationTest` against the native executable. the-main-thread.com/p/quarkus-nati… #java #quarkus

Apache Tomcat 10.1.60 is now available for download: buff.ly/N2gAYsn Applications that run on Tomcat 9 and earlier will not run on Tomcat 10 without changes. Change log: buff.ly/MNTC1uT

Jev is now available to everyone. No waitlist. Start using it here: console.typesafe.ai

He recopilado 179 libros gratis de programación en español. PDF, HTML y EPUB. 42 temas: JavaScript, Python, TypeScript, React, Rust, Git, SQL… Sin registro, sin pagos, sin sorpresas: → librosgratis.dev

No Java SDK for Jev yet? No problem ☕️ Here’s a single Java 25 file to get started. No framework, no dependencies, just one HTTP call. Set your API key and run github.com/danvega/hello-…

#Java27 has arrived! 🥳☕ Get all the details about the latest release from our Java team. social.ora.cl/6015BGPgHt

Our next #JCON2026 session is live: 'Chasing the Main Thread - Adventures in #AI Assisted Coding' with @myfear This talk takes you behind the scenes of a daily #creative experiment: writing and publishing one hands-on #tutorial or article … Grab your coffee and hit play: youtu.be/1wSjnnxWdYA

One pair of integrals interchanges time and frequency: F(ω) = ∫_{-∞}^{∞} f(t) e^{-iωt} dt ↔ f(t) = (1/2π) ∫_{-∞}^{∞} F(ω) e^{iωt} dω The analyzing wave and the synthesizing wave differ only by the sign of the exponent and a constant 2π. That near-symmetry is why a delay in time appears as a twist in phase, why stretching a pulse squeezes its spectrum, and why the energy computed in either domain is the same (up to that same 2π). Fourier wrote the first version while thinking about the flow of heat in 1807; the operational properties that make the transform a practical tool were assembled over the following century.

🤓💪

If you don’t want to wire up your own Jev client in Spring Boot, I created a simple starter for you. Add the dependency, set your API key, and inject JevClient. Built for Spring Boot 4 with RestClient and typed questions and answers. github.com/danvega/jev-sp…

☕ How Java programs can reach close-to-native performance for matrix-multiply computations on hardware with accelerated MMA support, such as NVIDIA GPUs ☕ How the same Java Tensor API can be mapped across different parallel programming models and vendors while remaining portable for source code and runtime scheduling parameters This article examines both: social.ora.cl/6011BEbmLc

A lesser-known production reality of MoE inference: (must-know for technical LLM interviews) MoE models reduce expert computation by routing each token to a small subset of experts. But routing also divides a batch of token rows into several expert-specific batches. Their sizes depend on how many tokens enter the layer and which experts those tokens select. This creates a systems problem that is not immediately evident from architecture diagrams. Let's walk through it. 1) Dense transformer execution Assume a decode batch contains 32 token rows. A dense Transformer sends all 32 rows through the same feed-forward network. For each feed-forward projection: > 32 token rows x the same weight matrix The accelerator performs one 32-row matrix multiplication and reuses that projection's weights across all 32 rows. 2) MoE execution A simplified top-1 MoE router may divide those rows like this: - Expert 1 receives 8 rows - Expert 2 receives 3 rows - Expert 3 receives 0 rows - Expert 4 receives 11 rows - All other experts collectively receive 10 rows Each 32-row feed-forward projection has become several smaller expert projections with different row counts. The engine must group rows by expert, run each expert, apply routing weights, and restore the original token order. With top-k routing, each token appears in multiple expert batches. Sparse routing reduces expert computation, but it also fragments the original batch. 3) How grouped GEMM helps GEMM is General Matrix Multiplication. Each active expert has its own weight matrices. Grouped GEMM schedules several expert matrix multiplications together: - Expert 1 rows x Expert 1 weights - Expert 2 rows x Expert 2 weights - Expert 4 rows x Expert 4 weights This reduces launch overhead and helps the accelerator process differently sized expert batches together. It works best when each active expert receives several rows. 4) During low concurrency During ordinary autoregressive decoding, each active request contributes one current token to a model step. With 32 active requests, some experts may receive several token rows. With only one to three requests, the selected experts may barely overlap. Most active experts can receive a single row. Each operation now resembles GEMV, or General Matrix-Vector Multiplication. The engine loads an expert's weights to process one token vector. That leaves little opportunity to reuse those weights across rows. Grouped GEMM can combine the launches, but it cannot turn one-row expert batches into large matrix multiplications. 5) An example Researchers encountered this while optimizing Qwen-3.5-397B-A17B-FP8 on Ironwood TPUs. Each token selected ten routed experts from a pool of 512. With three concurrent requests, they found that only 5% of the selected experts were duplicated across requests. Most routed expert batches therefore contained one row. So they built a custom kernel that prefetched selected expert weights into the TPU's fast VMEM and processed these GEMV-like operations directly. This resulted in a 3.6x faster MoE block at concurrency one on a TPU configuration. This does not mean grouped GEMM is inefficient. Instead, it implies that the best execution strategy depends on the expert-batch shape: - Several rows per expert → grouped GEMM can work well - One row per expert → hardware-specific weight streaming may work better The architecture stays the same, and concurrency and routing overlap change how the work is assigned to each expert. To dive deeper, I wrote a full article covering this and the other engineering problems behind MoE inference, including memory, cross-GPU dispatch, expert placement, load imbalance, quantization, and offloading. Read it below.

Avi Chawla @_avichawla

5 days ago

x.com/i/article/2100…

6 36 270 100K 394

50 System Design Patterns Every Engineer Should Know buff.ly/RbLwBps

keyboard_arrow_left Previous keyboard_arrow_right Next

@freeCodeCamp freecodecamp.org/news/hands-on-…

Playwright MCP or Playwright CLI: pick either, your agent gets all of this 🎭 🧰 WebMCP: pages hand the agent their own tools 🔎 Find: grep the page instead of reading all of it 🪶 Leaner snapshots, fewer tokens on every step 🔴 Do it by hand once, get Playwright code back 🗣️ Codegen in TypeScript, Python, Java and C# 🔌 Your real Chrome, logins included 🎬 60 fps video with an animated cursor and chapters 📱 Mobile and device emulation 🌗 Light/dark flipped mid-session 🌐 Mock any request, or pull the plug and go offline 📸 Hi-res screenshots 🩹 Crashed tab or dropped connection? It'll recover npx @playwright/mcp@latest github.com/microsoft/play… npm i -g @playwright/cli@latest github.com/microsoft/play…

If you are using SDKMAN .sdkmanrc file (you should) then you can reference .sdkmanrc file in GitHub Actions instead of repeating JDK version at multiple places. #java #sdkman #tip

@barranquillajug 😂😅 cuando estaba de practicas, que épocas.

Wow, no esperaba saber de Struts en estas epocas. Como primer framework web que use, lo ojeare a ver que diferente tiene ahora :D

The ASF - Apache Software Foundation @TheASF

5 days ago

Apache Struts 6.11.0 and 7.3.0 are now available for download: buff.ly/9qJRZa8 Struts is an elegant, extensible framework for creating enterprise-ready Java web applications.

0 5 29 10K 8