Daily Reading List – April 3, 2026 (#756)

I’m ready for the weekend and looking forward to downtime. Maybe some tech experiments, but mainly investing my time in a great Easter with the family.

[blog] The New Bottleneck That Will Chill Your AI Vibes. Unless you’re a single-person startup, you’ll want some sort of foundational platform that makes it easier (possible?) to go fast.

[article] Directing a Swarm of Agents for Fun and Profit. I like this collection of observations from Adrian. Those of us who have been in tech for a while can appreciate how he’s looking at the AI world.

[article] Backstage is dead. Do we need different things from an Internal Developer Platform that what Backstage offered? Yes. It’s a piece, but a platform matters.

[blog] Supporting Google Account username change in your app. Yes, US users can now change their usernames without losing anything. Is your app ready?

[blog] developer relations after the cheat code machine. I like this. Does instruction (and DevRel) move away from “hello world” and “how to use the thing” towards more apprenticeship-style guidance of how to use the AI tools well? Seems like it.

[blog] Creating a Wikipedia MCP Server in Java in a Few Prompts with Skills. Just build things. Like your own MCP server for Wikipedia in minutes.

[blog] Welcome Gemma 4: Frontier multimodal intelligence on device. Hugging Face knows how to write a launch blog. This one for the new Gemma 4 is outstanding.

[blog] Gemma 4: The new standard for local agentic intelligence on Android. The edge and mobile impact of Gemma 4 is huge. But the license change is getting the most positive chatter.

[blog] We Trained 47 Models and Lost the Best One. Then We Found Vertex AI Experiments. Can you reproduce your model tuning? Looks like many can’t. This is an educational piece on progressing towards smarter experiments.

[article] Axios hack exposes AI-coding’s dependency problem. AI-generated code uses a lot of open source references. Are they up to date? Exposing vulnerabilities? Without the right checks in place, you face real risks.

[blog] Run real-time and async inference on the same infrastructure with GKE Inference Gateway. Two key patterns, one set of infrastructure. That’s convenient, and might simplify your architecture.

Want to get this update sent to you every day? Subscribe to my RSS feed or subscribe via email below: