Daily Reading List – June 9, 2026 (#801)

I like how much great content I’m seeing about practices, risks, and strategies for people using AI. It’s not “everything is awesome” or “everything is awful.” It’s about staying informed and focused on the desired outcomes.

[blog] In the Seams. If you’ve been around tech long enough, you’ve seen what Devin is calling out here.

[blog] Surviving the eviction: How to build interrupt-resilient AI workloads on GKE. All of this is outstanding advice to avoid any unexpected app behavior when running Kubernetes on ephemeral compute.

[article] Claude Code vs. Cursor vs. Codex vs. Antigravity — six months in. Even harder to predict the next six months! There convergence on some key concepts right now.

[blog] Modern Engineering Values. Fantastic piece. Really. Find details of a pragmatic way of working with LLMs, insights into the engineering values to hold on to, and the results you could see.

[blog] Paris Hilton is Android’s first icon in residence. I really, genuinely love this. You shouldn’t have to be an engineer to go from creative idea to implementation.

[article] 8 myths on software engineering and AI. These will sound familiar to many of you. But seeing them all together is useful.

[blog] Thoughts on starting new projects with LLM agents. I like the advice here: forcing the agents to do small CLs/commits, human in the loop the whole time, Go as the right language for agent-written projects.

[blog] Bringing the latest Gemini models to Apple developers. Easier access to models, along with Gemini in Xcode for help on coding tasks. That model access is courtesy of Firebase technologies.

[article] Companies Are Using AI for Efficiency. They Should Use It to Grow. Efficiency is a great reason to do automation or use AI. Full stop. But not the ONLY reason. Sounds like many execs are missing the top-line value of using AI to drive growth.

[blog] Antigravity Managed Agents Tutorial: Ship Production AI Agents. Simply fantastic writeup of this agent-as-a-service offering where you provide a prompt, and the service analyzes your objective, plans an approach, executes tools, produces results, and loops until complete.

[blog] The Intent Debt. Requirement docs, and such never really captured “intent.” Usually it was the result of discussion/agreement. Now with unattended AI loops, do we need better ways to share the thinking that goes into our decisions? Probably.

[article] Anthropic brings Mythos to the masses with Claude Fable 5, its most powerful generally available model ever. Your chance to try a Mythos-class model is here. Some safeguarded features aren’t opened up to the masses, just in case you wanted to try those. Anthropic’s blog, and one mentioning it’s available on Google Cloud.

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