Category: Daily Reading List

We shipped a cool open model today, and I enjoyed a few meeting-free moments to work on some docs and research. It looks like many folks are already starting an extended weekend, but I’ll see you here tomorrow.

[blog] Gemma 4: Byte for byte, the most capable open models. It’s a high performing family of models that’s distributed with a commercially permissive license (Apache 2). Multimodal, reasoning, and tool support. Small sizes available. Also check out my Cloud post, and our open source perspective.

[blog] Spec Driven Development. Is this the next software evolution? Is the spec the actual source of truth now for new apps? Mofi takes a look.

[blog] Meet the new Cursor. The IDE’s days are numbered. Look at this evolution of Cursor and recognize that the IDE is becoming a secondary surface.

[blog] I Built a Real-Time Voice AI Stylist with Google ADK. Here’s What the Docs Don’t Tell You. I love posts like this. Here’s a real pain point—there’s an annoying lag in the voice agent—along with a deep look at how it happened and the better approach.

[article] When Executive Presence Backfires. You don’t want to hear generic feedback that you need more “executive presence” or give that feedback. Both are unsatisfying. But how can efforts to build these behaviors actually cause you problems?

[article] How Amex deploys AI tools. Looks like some widespread use cases, and a “redesign” of how they operate.

[blog] From Lag to Lightning: Optimizing MCP Toolbox with Built-in Observability. We’re going to need to start observing and optimizing a whole other set of components in our architectures. I’m glad this one added easier tracing and automatic instrumentation.

[article] AI-generated merged code holds steady at ~30%. Going up, and the data is based on reports from devs, not instrumented systems. Ours is instrumented, and much higher.

[blog] Google Workspace Studio Tutorial: Building an AI Meeting Prep Agent. I’m a dinosaur when it comes to tricks and tools to optimize my email/calendar/docs. Not on purpose, but most don’t feel necessary to me. But I may do the thing called out here.

[blog] How Dart and Flutter are thinking about AI in 2026. Language and framework teams need to have some guiding principles for AI. Good to see these.

[blog] Defending Your Software Supply Chain: What Every Engineering Team Should Do Now. Be a little freaked out by these recent supply chain breaches. Docker has some advice for playing it safer.

[blog] Google Workspace’s continuous approach to mitigating indirect prompt injections. I learned about a new attack, and what we do to protect people from it.

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

Don’t let some of my downbeat reading items make you think I had a bad day. It was a good one! For some reason, I came across a handful of cautionary words today. Which is ok, because we should be asking questions and looking around corners.

[report] The State of Java 2025. Lots of data here, some of it surprising to me. Amazing to see so much Java in China, and from a generally young crowd. Also that more and more people are forgoing any framework.

[blog] Five techniques to reach the efficient frontier of LLM inference. Important stuff here for ML engineers and platform teams.

[blog] Some uncomfortable truths about AI coding agents. Everything isn’t awesome all the time. Maintain healthy cynicism with tech progress to ensure you’re not doing worse work.

[article] What kinds of new debt are teams accumulating with AI? Smart question to ask, and account for.

[article] Lagging cloud maturity threatens enterprise AI plans. If your cloud adoption has stalled—both usage of services and adoption of cloud-style practices—I don’t see how you’ll be widely successful deploying AI for your teams and customers.

[blog] Cloud Run Jobs vs. Cloud Batch: Choosing Your Engine for Run-to-Completion Workloads. The major hyperscalers offer more than one service to do the same task. But there are subtle differences that help you pick which service to use, as pointed out here.

[blog] Six Takeaways From KubeCon EU 2026. I liked this roundup, as the Intuit Engineering team covers a good range of topics from the event.

[blog] Developer’s Guide to Building ADK Agents with Skills. Learn these patterns, as you’ll likely see them become common in your agent framework of choice.

[blog] Building my Comic Trip agent with ADK Java 1.0. Enterprise use cases are helpful for 1:1 mapping to your day job. But we can also get work inspiration from fun examples of new technologies.

[article] What next for junior developers? Get good at communication, analyzing the world around you, and understanding the big picture.

[blog] Google Cloud: Investing in the future of PostgreSQL. The big focus lately has been around replication capabilities for active-active setups.

[blog] Cloud CISO Perspectives: RSAC ’26: AI, security, and the workforce of the future. You’ll see some RSA recaps this week. This, and a bunch of security-focused links, can be found here.

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

For some reason, today’s list has a fair number of items that dig into the “how” of our work. How we architect, motivate ourselves on “hated” products, look at AI in our teams, and swap to cloud managed services.

[blog] Is Event-Driven Architecture Overkill for Most Apps? Derek tries to clear up why you bring events into an architecture (not scaling) but because you’re decoupling responsibilities in time. This and more in a good post.

[blog] Explore Benjamin Franklin’s scientific discoveries on NotebookLM. So cool. What a great way to learn! Stuff a bunch of content into NotebookLM, and then pick the best way to explore. Try it yourself.

[blog] Axios NPM Distribution Compromised in Supply Chain Attack. That’s the second scary supply chain attack on a popular dependency in the past week.

[article] Every Building You’ve Ever Been In Was Designed By Software Built in 1997. There are so many places where upgrading the software experience can have massive impact. Construction, for one.

[blog] Working on products people hate. We’ve probably all been there. But I liked Sean’s take, as it’s a reminder that at least people care (if they hate it), and it keeps you humble.

[article] AI is splitting companies into two groups. Who’s going fast or slow? Looking at tasks or jobs? That’s where Matt sees the separation.

[blog] Agent to UI Protocol (A2UI) with Agent Development Kit (ADK). I understand the relationship between A2UI (UI spec) and AG-UI (connect app to agent) better now.

[blog] ADK Go 1.0 Arrives! We just shipped ADK Java 1.0, and now it’s Go’s turn. This is a powerful framework for anyone building rich AI agents.

[blog] How to build production-ready AI agents with Google-managed MCP servers. Managed or self-hosted, MCP will likely find its way into some of your AI apps and agents. Here’s a good example of using both types, and securing the setup.

[article] Tech Conferences Aren’t Dead. But Why We Go Is Changing. We don’t go to events to learn something we can just Google. We go for deep focus, conversations with peers, and structured learning.

[blog] Spanner’s multi-model advantage for the era of agentic AI. You want relational? Key-value? Vector? Graph? This and more from a single Google Cloud database.

[article] When product managers ship code: AI just broke the software org chart. Yes, a lot changes/breaks/improves when more people can get involved in the process.

[blog] How ID.me Scaled to 145M Users While Reducing Operational Risk. Good story of building mission-critical services on managed cloud products.

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

A few of today’s items stood out to me as good grounding as the (tech and real) world swirls around us. Focus on your team and don’t overburden them. Slow your thinking and write more down. Keep a steady eye on new paradigms that may have huge impact, but need consideration. We’ll be fine.

[blog] The Code Agent Orchestra – what makes multi-agent coding work. This is the next focus area I’m keying in on right now. I’m competent building apps with AI tools. But now it’s time to get comfortable unleashing a team of agents.

[article] Are You Overburdening Your Most Engaged Employees? Easy to do. But that enthusiastic, competent person has their limits. Some suggestions here for better ways to assign work.

[blog] Announcing ADK for Java 1.0.0: Building the Future of AI Agents in Java.Fantastic announcement, and exactly how you should write a blog post like this. Check out the summary of what’s new, why it matters, and exactly how to use each feature.

[blog] How I Finally Killed My Massive BigQuery CASE Statements. As Jeff mentions, you wouldn’t use this approach for giant datasets—you’d employ an actual step to pre-compute embeddings—this is a much simpler option for mid-size queries.

[blog] Make the switch: Bring your AI memories and chat history to Gemini. Convenient. We’re all building some gravity with our preferred tools, but easier switching costs makes it possible to look around more.

[blog] Why You Should Start Blogging (Even If Nobody Will Read It). Yes! Start now. Develop your voice. Watch what happens.

[blog] The new AI literacy: Insights from student developers. Maybe the young folks are on to something. This research project pulled out some insights that seem applicable to our business settings.

[blog] Using Flutter with Firebase: A Developer’s Guide. This has detailed look at the backend services any mobile (or web) app needs.

[blog] How agentic software development will change databases. Agents creating 4x more databases than humans? That’s what Databricks is seeing in their Lakebase platform.

[blog] Announcing the winners of the MedGemma Impact Challenge. I like these sorts of competitions where people build genuinely useful technology.

[blog] Gemma-SRE: Self-Hosted vLLM Infrastructure Agent. That same open model (Gemma) makes it straightforward to self-host your own LLM for cases like this.

[article] Waymo’s skyrocketing ridership in one chart. Pretty crazy growth curve. Only going up!

[blog] Calling the game: How MLB brings AI-powered color commentary to fans with Scout Insights. This is a great way to add AI to a product. Give it great grounding sources, and focus on a delightful (and fast) user experience.

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

It’s baseball’s Opening Weekend and I’ll be at one of the San Diego Padres games. If you don’t like baseball it’s ok. We’ll still be friends. Probably.

[article] What we wish we knew about building AI agents. Do you need an agent, or just an MCP server? Should you build a custom harness? This and more in this insightful post.

[blog] How did Google turn AI IDE features into measurable productivity gains? Check out this analysis of our research report where we showed that productivity gains didn’t come from a single model upgrade, but but improving all layers of the funnel/stack.

[article] Leaders Underestimate the Value of Employee Joy. It’s true. And you can’t fake it as a leader. Recognizing that serving emotional needs drives better team performance is an unlock code.

[blog] Closing the knowledge gap with agent skills. How well do skills work? Well, but we’re still learning the best way to employ them. I liked the perspective in this post about our new agent skill for Gemini.

[article] Is the FDE role becoming less desirable? Maybe Forward Deployed Engineers are being asked to serve as sales engineers instead. That’s what Gergely looks at here.

[blog] Generating music with Lyria 3 and the Gemini Interactions Java SDK. I’m still amazed by what we can create with just a few lines of code.

[blog] Reducing our monorepo size to improve developer velocity. There are benefits and risks of a monorepo. Dropbox shrunk theirs and saw faster cloning and more reliable CI pipelines.

[blog] Next.js Deployment Adapters: A bright future for Next.js on Google Cloud. We’re part of a consortium that’s making Next.js better in more runtimes, including Firebase.

[blog] Top 3 Everyday Angular Tasks That MCP Servers Automate for You. Faster starting points. That’s a great value prop of a good tool. These are good examples of where this MCP server gets you going (and iterating) faster.

[blog] Introducing Wiz Agents & Workflows: Security at the Speed of AI. These Wiz folks might be on to something. These offensive, defensive, and resolution agents look powerful.

[article] From Friction to Flow: How Great DevEx Makes Everything Awesome. Few people have studied and understood developer experience as much as Nicole. Learn how to remove friction from your team’s work.

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

750 of these daily notes. I’m happy about that. And I appreciate each of you who takes time to skim through it.

[blog] Gemini 3.1 Flash Live: Making audio AI more natural and reliable. A fast, natural-sounded real-time AI experience feels like magic. Now you can bake this latest audio and voice capability into your AI apps.

[blog] Choosing the best git branching strategy for continuous delivery in your team. Useful look at some classic approaches to managing work in an engineering team. I wonder if new patterns emerge given the volume (and speed) of AI code generation.

[blog] Kubernetes Still Feels Weird? What i wish i knew sooner. Wow, this is super terrific. It’s a comprehensive dive into all corners of what Kubernetes is and how it works. But written in a way that most anyone can understand.

[article] How to Convince Others to Trust Your Instincts. This can be tough, especially on a senior team or if people don’t yet see you as someone with viable opinions. I liked the advice here, as much of it felt actionable.

[blog] A developer’s guide to training with Ironwood TPUs. Really, this a set of optimization methods you can use for model training.

[blog] Deployment strategies: Types, trade-offs, and how to choose. What are the proven ways to ship software? There are many patterns based on your goals and the sophistication of your production environment. Get familiar with these!

[blog] Your Employees Aren’t Ready For AI — And It’s A Problem. The skills, psych safety, and tools might be missing from your company. What are you doing about it?

[article] How can engineering leaders assess their AI maturity? Let’s say you’re ready for AI. How do you know if you’re progressing? Useful data here.

[blog] Building superconducting and neutral atom quantum computers. I barely grasp this stuff, but progress in quantum has some major impact. Keep an eye on it.

[article] State of Context Engineering in 2026. How do you manage what gets stuffed in the context window? It’s not just about a prompt. It’s the tools, ways to progressively disclosure new info, how you compress context, and more. Good post!

[blog] The State of Open Source Licensing in 2026. Some of you nod off when the topic of “open source licensing” comes up. I get it. But the implications of recent shifts have material impact on your usage.

[blog] Vibe Coding XR: Accelerating AI + XR prototyping with XR Blocks and Gemini. Neat. Vibe coding for extended reality apps? That’ll help people quickly figure out where to go.

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

It was a whirlwind of a day in Manhattan, but I had a blast chatting with a forward-thinking customer and getting insights into what non-tech folks care about in this AI conversation. Now, flying back across the country for team meetings in Sunnyvale tomorrow.

[blog] The open platform for the AI era: GKE, agents, and OSS innovation at KubeCon EU 2026. New open source, new AI capabilities, and new troubleshooting features in Google’s Kubernetes push.

[blog] Harness design for long-running application development. Really terrific piece from Anthropic that explores different multi-agent harnesses to get past “context anxiety” during long development sessions.

[blog] When should a manager step in? Good topic, and made me think about when and how I jump into something with my team.

[article] How Leaders Can Build a High-Agency Culture. Somewhat related to the prior item. Do you have a high-agency culture where people feel they can make a difference and own the path forward? You want a high-agency culture during this AI transformation period.

[blog] 8 Tips and Best Practices for MCP Server Development.These seem sound. I wonder how many are going to these lengths right now, however.

[blog] TurboQuant: Redefining AI efficiency with extreme compression. Do I understand all of this? I do not. But it feels like a big deal, and this news story explains it further. And this less reverent one from TechCrunch.

[blog] Full-Stack Vibe Coding: Building Production-Ready Apps with AI Studio, Stitch & Antigravity. Romin links to a handful of recent posts, including mine, that can help you get confidence to start building apps yourself.

[blog] Build with Lyria 3, our newest music generation model. More control, better lyrical sync with the music, and other goodies. Pro version lets you create up to 3-minute songs!

[article] PyPI warns developers after LiteLLM malware found stealing cloud and CI/CD credentials. Yikes. This is rough. Consider ways to sandbox and scan your dependencies before snagging them directly from public registries.

[blog] Welcome llm-d to the CNCF: Evolving Kubernetes into SOTA AI infrastructure. Good to see this open project makes its way to the CNCF. Related, from us.

[article] AI inference costs set to plunge: Gartner. “Set” is doing a lot of work in this headline. This is projecting that in four years, we should see new efficiencies, but much of that will plow back into revenue for labs that are losing money today.

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

Travel day to New York City, and thankfully I had decent airplane wifi. Tomorrow I get to talk to a couple hundred non-tech folks about AI. There’s a lot for me to learn!

[blog] Google ranks #1 on Fast Company’s Most Innovative Companies list. Proud of my colleagues for rising to the moment and doing creative work everywhere in the company.

[article] Your database is about to become an AI tool. Is it ready? Any AI app or system that lasts will need database access. Have you done the necessary steps to prepare?

[blog] Your Startup Is Probably Dead On Arrival. Well that’s a very uplifting title. But Steve Blank’s point is that if your startup is a couple of years old, you need to stop and look around to see what’s fundamentally changed.

[blog] RSAC ’26: Supercharging agentic AI defense with frontline threat intelligence. Security is the topic of the week thanks to the RSA conference going on. Here’s our roundup of security-related data and news.

[blog] Bringing dark web intelligence into the AI era. It’s dangerous out there, but the tools are getting better and giving you a leg up on adversaries.

[blog] M-Trends 2026: Data, Insights, and Strategies From the Frontlines. This is quality data about the current threat landscape, and how it’s evolved over the past couple years.

[blog] In Defense of Deep Reading. Read as much as you can. I agree with everything in here.

[article] AWS at 20*: Inside the rise of Amazon’s cloud empire, and what’s at stake in the AI era. Long, insightful story about the orgins of AWS with quotes from insiders.

[article] Why aren’t AI productivity gains higher? No one can sell you productivity. They can sell you tools to make yourself more productive. But only if the circumstances are set up for success.

[blog] Skills vs. Tools: Replacing the Google Firestore MCP Server with Skills (+ Go Binaries). Neat pattern. Build super fast, local binaries that act like tools from within an agent skill.

[article] When Senior Leaders Lack People Skills, Transformations Fail. No matter how good (or bad) you are at the people parts of the job, we can all get better. And should.

[blog] Your AI agent can now create, edit, and manage content on WordPress.com. Hmm. I’ll probably pass, although some of the operational tasks might come in handy.

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

I flew into San Jose last night ahead of today’s in-person rehearsal for the Google Cloud Next developer keynote. Super fun day with people I have a lot of affection for. Tomorrow, off to New York City for 24 hours.

[blog] Developer AI Tooling in 2026: Trends Shaping How We Build. This seems like a good assessment of where dev tools are right now.

[blog] End-to-End AI Agent on GCP: ADK, BigQuery MCP, Agent Engine, and Cloud Run. Whether you’re using cloud services or not, most agent architectures include a mix of components that you stitch together.

[youtube-video] Google just changed the future of UI/UX design… Great Fireship video that explores Google Stitch, a lifesaver for those of us who don’t create great frontend designs.

[blog] Is the IDE dead? The code editor is becoming a “read only” view for more and more people. Addy looks at the role of the IDE moving forward.

[paper] Cloud Infrastructure in the Agent-Native Era. We snuck this cool little paper out (direct link to PDF). Cloud native infrastructure was a step forward, but AI apps and agents need something further.

[blog] The Three Pillars of JavaScript Bloat. How do dependency trees result in a bloated app? I can be hard to unravel, but there’s some advice here.

[blog] Architecture Is On The Hook For GenAI Success. That sounds like something an architect would say! There’s definitely truth to the reality that scaling AI in any company will depend on platforms, guardrails, and good decisions.

[blog] AI Doesn’t Fail in the Demo – It Fails the First Time You Have to Trust It. I can build a mean demo that gets you excited. But what matters is what it takes to trust AI tech at scale in your company.

[[blog] Same Old. Very few things are “unprecedented.” It’s one reason I try to read a lot of history to gain perspective.

[article] Cursor admits its new coding model was built on top of Moonshot AI’s Kimi. If you’re not shipping your own model, just say so. It’s ok. Cursor did right by quickly acknowledging this and giving credit.

[blog] How Slack Rebuilt Notifications. Different mental models, overlapping settings, and more makes chat systems like Slack overwhelming. This looks like a smart redesign.

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

Project Hail Mary is maybe my favorite fiction book from the past few years. I read it twice, which I never do. So, I’m super excited to go see the movie tonight. If you see it too, let me know.

[article] What Is the PARK Stack? I know there have been other acronym-stacks since LAMP, but I struggle to remember them. This one is about PyTorch, AI models, Ray, and Kubernetes. Might stick.

[blog] Kubernetes v1.36 — Sneak Peek. Speaking of Kubernetes, there always seems to be another version around the corner. Scale to zero is interesting!

[article] Anthropic just shipped an OpenClaw killer called Claude Code Channels, letting you message it over Telegram and Discord. Neat. Expect a lot of “OpenClaw killers” this year as people experiment with multi-agent orchestrators.

[blog] How I overhauled my app UI in minutes with Stitch and AI Studio. Great example. Take those existing apps and let these smart tools help with redesign, rearchitecture, and deployment.

[article] 9 reasons Java is still great. Java is doing fine. I’m not sure it’s the default choice for many startups, but it’s well-established and constantly improving.

[blog] Next-gen caching with Memorystore for Valkey 9.0, now GA. If you like open software, fast performance, and reliable databases, Valkey could be on your radar.

[blog] Building an MCP Ecosystem at Pinterest. Here we go. Let’s get some real-world practices from users, not just messages from vendors and thought-leaders.

[blog] Streamline read scalability with Cloud SQL autoscaling read pools. One smart way to scale relational databases is to use read replicas. Now we’re offering a clean way to autoscale your read replicas without requiring any changes to your apps.

[article] State of JavaScript 2025: Survey Reveals a Maturing Ecosystem with TypeScript Cementing Dominance. This is a dense report, so InfoQ rolled up some of the highlights. But dig into the source material and see what stands out to you.

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