Brad Frost visits his daughter's after-school coding club to share his passion for web development. He demos browser DevTools tricks on Minecraft's website, shows his daughter's vibe-coded game, and builds a live interactive webpage using JSON data collected from the kids in real time. The post reflects on the joy of sharing the magic of coding with children, free from professional jargon and business concerns.

Pele is a lightweight open source library that renders Mermaid diagrams to SVG, themed with CSS variables. It has zero dependencies and is 6 times smaller and up to 500 times faster than Mermaid.js. Gzip bundle size is 230 kB versus 1.6 MB for Mermaid.js, and render time is 0.05 to 0.3 ms versus 6 to 82 ms. It supports every diagram type built into Mermaid 12.1.0 and is developed for Obsidian.

This is a fan-made StarCraft game. It's free to play on web and mobile. You command Terran and Protoss forces to hold the bridge and see how deep the swarm goes. The first run is a warm-up before Endless. A Wild West season is a 3-day board, then the real season begins.

Google DeepMind announced Genie 4, a new world model for real-time interactive simulation. It is natively multimodal, generating audio, ambient sound, and interactive dialogue alongside the simulated world. Gemini Omni anchors visual realism while Gemini 3.8 Flash acts as a guiding agent, observing frames and adapting plans in real-time. DeepMind is also experimenting with Gemini 4 Argon and Genie 4 alongside Gemini Robotics. The model is rolling out as a research preview to trusted testers first, ahead of availability to Google AI Ultra subscribers in Project Genie and early access Gemini API partners. All generated video and audio carries SynthID watermarking, and DeepMind notes visual drift and object permanence issues still occur in longer sessions

A developer revived wasp-os, the MicroPython-based OS for the PineTime smartwatch, by porting it to mainline, up-to-date MicroPython instead of its old 2019 fork. The work required three pull requests merged upstream into MicroPython, targeted for version 1.30. Build tooling now uses uv, Bluetooth flashing uses bleak and dropped from over eight minutes to about three and a half, and free RAM at boot rose from 7.1 KB to 12.4 KB. New additions include a Dutch watch face, an 8-bit style clock, and a 'Guido' app showing a blocky portrait of Guido van Rossum with Python and MicroPython logos. The code is available on a GitHub fork branch.

Java 26 arrives with 10 JEPs and relatively few new features compared to Java 25. Key highlights include: JEP 500 which warns when final fields are mutated via deep reflection (preparing to block this in a future release), JEP 504 which removes the long-deprecated Applet API, JEP 516 which improves AOT object caching to work with any GC, JEP 517 which adds HTTP/3 support to the JDK HTTP Client, and JEP 522 which improves G1 GC throughput by reducing thread synchronization. Notably, UUIDv7 support is added via UUID.ofEpochMillis(). No preview features graduate to standard in this release, with several features like Structured Concurrency, Vector API, and Primitive Types in Patterns remaining in preview. Miscellaneous API additions include Process.AutoCloseable support, case-folding String methods, and new BigInteger/Duration/Instant utilities.

Databricks and the Linux Foundation have evolved Delta Sharing into OpenSharing, an open protocol that extends zero-copy data sharing to agent skills, AI models, and unstructured data. The announcement came on June 10, 2026, after Delta Sharing reached more than 28,000 data recipients, with about a third of shares crossing platforms. OpenSharing adds a Share-Schema-Asset hierarchy covering Tables, Volumes, Agent Skills, and Models, plus support for Apache Iceberg REST Catalog clients and on-premises storage through partners like MinIO. Existing Delta Sharing clients stay compatible, and the project now sits under the Linux Foundation AI & Data Foundation at sandbox stage.

The Model Context Protocol (MCP) is an open standard connecting AI apps to external tools and data. A server declares what it can do, the host app shows that to the model, and when the model wants something the app calls the server and returns the result. The model never talks to a server directly. MCP runs as JSON-RPC over stdin/stdout or a single HTTP endpoint, with servers able to talk back. A worked example traces 11 messages across 5 steps: the app lists tools, adds 'create_issue' to context, the user asks to file a bug, the model requests the tool, and the server returns 'created #482'. The spec is dated 2026-07-28 and is stateless with no handshake but multiple round-trips.

A developer built an interactive system design reference site that lets you view the same architecture through AWS, Google Cloud, Azure, or open-source tooling. It covers 36 systems, 36 components, and 27 concepts, with auto-laid-out diagrams you can step through hop by hop. Each system page includes requirements, APIs, capacity estimates, data models, bottlenecks, and scaling notes. The site is built with Next.js static export deployed on Vercel, React Flow plus ELK for diagrams, Tailwind CSS, and Zod for schema validation, and it's open source on GitHub.

A curated list of 15 small apps (Linktree, TinyPNG, Bitly, Day One, Splitwise, Typeform, Calendly, Toggl Track, Sortly, Content Snare, Trello, Notion, Slack, Statuspage, YNAB clones) that can be vibe coded with an AI agent instead of paying for subscriptions. Each links to a build guide from the author's Vibecode It Yourself catalog of 109 project ideas, with scoped prompts, realistic time estimates for a basic version versus a full product, and explicit non-goals to keep the first build small. The guides support stacks like AHA, Next.js, Laravel, or Ruby on Rails, and the piece recommends starting with simpler self-use tools (Linktree, image compressor) before attempting ambitious ones like a Slack clone.