Sovereign Al is a hot topic, so I built a RAG agent that lets you compare OpenAl's GPT-4o Mini with models you can host yourself in a simulated recruitment agency scenario. Try the demo and reach out if you need to build Al or data engineering solutions!.

Good summary of how LLM providers approach training data.

Interesting paper from Nvidia on why a selection of small language models may be better than large generalist models for agentic systems. Also provides a road map for extracting small model subsystems from agents relying on large models.

Great to see mass-market software migrating to memory-safe languages.

Interesting learnings on building AI agents for the enterprise market.

RT @dbreunig: As your context bloats, you hit different failure modes. These failures hit agents hardest because they operate in exactly t….

AI can drive massive developer demand in Europe once governments figure out that its now feasable to build and replace foreign-controlled software. Past attempts to adopt Linux stalled, but AI-development means it's a lot cheaper to address gaps and build homegrown solutions.

Android needs an app permission for audio. Some apps interrupt audio playback of other apps even when the app itself isn't playing any sounds. Bet they would behave better if users could revoke their audio permissions.

Good overview of the state of the streaming data ecosystem. Reach out if you need help building data pipelines with Kafka, Flink, or Google Dataflow! . (or maybe one of the newer contenders?).

Good summary of some of the problems apps that use LLMs need to manage.

Seeing Claude Code search the web to work on a task and wondering how sustainable that is as AI chats increasingly replace web searches.

Good article on why you may want MCP as a layer over your APIs for use by AI assistants. My tool SpecToMCP helps you get started by generating MCP servers from API specs

RT @_avichawla: Finally! A RAG over code solution that actually works (open-source). Naive chunking used in RAG isn't suited for code. Th….

Big fan of Cursor, but VS Code isn't the best IDE and there seems to be better code AIs out there. Going to give Intellij IDE with Claude Code a shot to see if better disparate components can compete with Cursor's integration advantages.

At the Nordics kickoff for the Agentics Foundation to track what's new with AI agents

Started the same cross-platform desktop app with a UI that integrates a native lib in 3 different GUI frameworks and can't say I'm convinced by the dev experience, perf, future support, and maintainability of any of them. Might dig up JavaFX to see what it looks like these days.

RT @BdKozlovski: With a (one-time) budget of ~$42,000, you can configure a 3-broker Kafka cluster with HW capable of 56 GB/s of sequential….

AI being good at UI dev but bad at environment setup might lead to a new kind of cross-platform development. Frameworks like React Native require more environment setup, so it makes sense with UI specs that AI can build using native UI kits with Rust libs for shared functionality.

Languages that prevent or force handling of null pointers in the type system do provide meaningful stability improvements.

RT @simonw: Anyone building "agentic" systems on top of LLMs needs to take this principle into account every time they design or implement….