Category: Daily Reading List

Whew. A frantic day until about 3pm, and then I got a chance to work on an AI agent I’ve been messing with. I’m building it for use in a new blog post, but timeboxing how much more effort I put in. Hopefully posting next week.

[blog] 150 of the latest AI use cases from leading startups and digital natives. Get inspired by what startups are doing with AI right now.

[blog] 1,001 real-world gen AI use cases from the world’s leading organizations. We might have gone too far here. This started with an innocent 101 use cases, and now we’re drunk on stories and sharing 1,001 of them.

[blog] Embracing the parallel coding agent lifestyle. Is every developer a type of “manager” now? There’s a new workstyle that involves coordinating a series of agents doing various bits of work for you.

[blog] Vibe engineering. Also from Simon, this builds on the previous post. Now we’re talking broader engineering practices, not just using AI for a snippet of code.

[article] Control Codegen Spend. Unless you have an all-you-can-eat license (which seems rare), there are cost and consumption considerations with AI-assisted coding tools.

[blog] Gemini Computer Use: Giving LLMs Hands and Eyes. Our new computer use model definitely landed on my “try this out” list. Heiko gives it a run through here.

[blog] The Weak Point in MCP Nobody’s Talking About: API Versioning. Seems like something to consider, but I also don’t know how versioning might be different when we’re talking agents and tools.

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

Another day, another chance to learn something new. Today’s reading list had some useful data points, fresh ideas, and new products.

[blog] Introducing Gemini Enterprise. The AI platform era is here. It’s not just about a collection of random products. It’s about intentionally connecting people, systems, and knowledge bases so that we can get better work done. If you’re a Google shop, Microsoft shop, IBM shop or whatever, Gemini Enterprise is a major upgrade. More, from Sundar.

[blog] 4 ways Gemini Enterprise makes work easier for everyone. It’s useful to see actual examples of this platform in action.

[blog] How I Learned to Stop Worrying and Trust AI Coding Agents. Yes, we can learn some cool things from watching these agents at work.

[blog] Platform Shifts Redefine Apps. Important concepts here. What an “app” is changes with each tech platform evolution. Are you working with a modern definition?

[blog] Googler Michel Devoret awarded the Nobel Prize in Physics. It’s hard to have a big head at work when you’re surrounded by so many brilliant folks.

[guide] Choose a design pattern for your agentic AI system. This is fantastic, vendor-neutral architecture guidance for someone who wants to learn agent design patterns and when to pick each one.

[article] What the 2025 DORA Report means for your AI strategy. Extremely good takeaways in this post if you’re wondering how you’re supposed to actually land and scale AI at your company.

[blog] Predictions 2026: AI Moves From Hype To Hard Hat Work. Prediction season is upon us! See what various analysts and thought leaders are guessing.

[article] When You’re the Executive Everyone Relies On—and You’re Burning Out. Good advice here that resonates with me and how I’m feeling right now.

[blog] Give me AI slop over human sludge any day. He’s not wrong. Why are we automatically assuming AI-created stuff is worse or less useful than human created stuff?

[article] Survey: Engineers Want To Code, But Spend All Day on Tech Debt. Oof, these numbers are rough. Coding is rarely the bottleneck in your teams. It’s the 84% of the time that you’re not able to do what you love doing.

[blog] How to Build AI Security Agents with Gemini CLI. Good rationale here for when to build a tool versus when to leave the action to the agent to figure out.

[blog] LoRA Explained: Faster, More Efficient Fine-Tuning with Docker. If you kinda understood this approach to updating a model without a full training run, you’ll learn more from this post.

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

Flying home after a great day with friends at Comcast and talking about AI assisted engineering. Everyone is looking for the playbook for effectively landing AI in their org. Maybe I’ll write a book about it 🙂

[blog] Five Best Practices for Using AI Coding Assistants. A few months ago, our CEO asked me to run a few long-running engineering experiments with my team. Here’s what we learned from using the full spectrum of today’s AI coding tools to get work done.

[blog] Stitching with the new Jules API. I don’t think we’re just going to improve the existing toolchain. Instead, a new toolchain is emerging where you simply work differently.

[blog] The State of CI/CD in 2025: Key Insights from the Latest JetBrains Survey. What do folks use, are they using AI in CI/CD (answer: no), and why do folks use more than one tool? Answers here.

[blog] Now open for building: Introducing Gemini CLI extensions. This is a huge capability that brings all sorts of data and functionality into your agentic CLI.

[article] To scale agentic AI, Notion tore down its tech stack and started fresh. Bravo. Sometimes you need to reset, not refactor.

[article] Kubernetes for agentic apps: A platform engineering perspective. While I’m extremely bullish on lightweight platforms for hosting agentic apps, there’s absolutely a place for Kubernetes in the mix. Abdel makes a good case.

[article] How to write nonfunctional requirements for AI agents. I haven’t thought much about this, but sure, there are going to be some adjustments to what requirements you gather for AI apps.

[blog] Expanding access to Opal, our no-code AI mini-app builder. When I tried this tool, it wasn’t what I expected. But it’s an interesting way to build a new style of app.

[blog] Not Another Workflow Builder. LangChain isn’t interested in adding to the pile of visual workflow builders. We’re back to arguing workflows versus agents again too!

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

Flew to Philadelphia today to do a keynote at a customer’s internal developer conference tomorrow. Should be fun, although I’m definitely getting a little burnt out by all the recent travel!

[blog] Introducing the Gemini 2.5 Computer Use model. What can you build with agents that can navigate user interfaces? Try out our new Computer Use model.

[article] From Autocomplete to Agents: Mapping the Design Space of AI Coding Assistants. Great analysis. These ten dimensions, and six personas, are a useful way to understand the landscape.

[blog] Why Google Cloud Run is the most underrated container platform – Part 1. I’d go as far to say that it’s one of the best (compute) services on the Internet.

[blog] MCP Development with Rust, Gemini CLI, and Google Cloud Run. Here’s one example why Cloud Run is great. You have virtually no limit on programming language or scenario.

[article] Beyond the Hype: Architecting Systems with Agentic AI. Long conversation, but some interesting points about some of the bigger lifecycle considerations for AI agents.

[blog] Databases on K8s — Really? (part 8). My colleague at Google has been sharing a series of thoughts about his journey to appreciating containers and Kubernetes a a viable host for databases.

[blog] Spec-driven development: Using Markdown as a programming language when building with AI. You can do spec-driven development in many ways. GitHub created a cool framework that they’re demonstrating here.

[blog] Connect Spark data pipelines to Gemini and other AI models with Dataproc ML library. Integrating apps or data pipelines with AI models is getting more and more seamless. This looks like a very handy way to bring Gemini models to Spark data.

[blog] Accelerate AI with Agents: Event Series for Developers in EMEA. This roadshow has been overbooked in every city it’s visited. We’re rolling across EMEA, and see if your hometown is in the mix.

[blog] Ask a Techspert: What is vibe coding? We’re all builders now. What we build, and how production-ready it is, depends on the builder and the circumstances.

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

Shorter reading list today, but some great depth. I liked the thinking and challenge that many of these pieces offered.

[blog] Introducing CodeMender: an AI agent for code security. This AI is both proactive and reactive in helping secure your code from vulnerabilities. This looks like the future to me.

[blog] The RAG Obituary: Killed by Agents, Buried by Context Windows. I’ve seen a few obituaries for RAG, but maybe the agentic workflows really do resolve some of the issues that made RAG necessary in the first place.

[article] Microsoft retires AutoGen and debuts Agent Framework to unify and govern enterprise AI agents. There are too many frameworks to choose from, so it’s a good thing when a new one explicitly replaces an old one.

[article] OpenAI launches AgentKit to help developers build and ship AI agents. I’m wrong. Apparently we need more frameworks and agent-building tools. OpenAI announced many interesting things today during their Dev Day. Also updated models.

[article] Top executives jump on AI upskilling. Great to see. The right type of executive upskilling with shrink the expectation-gap between leaders and their employees.

[blog] vercel vs cloudflare: two philosophies of building for developers. A good rivalry forces participants to keep upping their game. This one is fascinating. I like Guillermo and the Vercel team but either way, developers win.

[blog] More choice, more control: self-deploy proprietary models in your VPC with Vertex AI. Great. Better control and security when running these leading models in the cloud.

[article] How to Drive Digital Innovation Without Wasting Resources. Some of this goes against conventional wisdom, which is why I liked it.

[blog] Scaling Engineering Teams: Lessons from Google, Facebook, and Netflix. You can’t copy culture and practices from one team and assume they work on the next. But we can still learn from what others do.

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

Heading into what I hope is a lowkey weekend. I’m excited to not do a whole lot. May your weekend be as action-packed or boring as you’re hoping for.

[blog] In Praise of RSS and Controlled Feeds of Information. At least 80% of my reading list comes from the RSS feeds I follow. The rest is from more real-time sources, newsletters, or random links I come across.

[blog] A collaborative approach to image generation. Not quite getting the image you’re after? The Google Research team looks at a new reinforcement learning agent that helps refine images by learning user preferences.

[article] What’s AI’s impact on software delivery performance? (2025 DORA Report). Analysis of our recent report is trickling out, and I thought this piece bubbled up some key findings.

[blog] This Week in Gemini CLI (vol. 5). Very good updates this week. I really like the potential of a managed todo list of agentic tasks.

[article] When Managing Your Team Becomes Too Much. There are times that I feel this. I manage a big (awesome) group and it can be overwhelming. Good advice here.

[blog] AI Won’t Replace Design Engineers, But It Will Change How We Work. Actively define how (or if, or where) AI complements your work before someone else does.

[blog] The History of Core Web Vitals. I knew basically none of this. I remember AMP and such, but didn’t realize all this other work happened. Impressive!

[article] Top CIO conferences to attend in 2026. Not a bad list of events to consider if you’re looking to make smarter strategic choices.

[article] Who stops wasteful cloud spending? “Waste” is inevitable, right? It’s difficult or impossible to be fully utilized. But aim for better! That said, how much on-prem spending is wasted? I’d argue it’s more 🙂

[article] Generative AI drives cloud spend blitz. You’re going to spend more in the cloud, because it offers the best way to take advantage of AI tech.

[article] The EM’s guide to AI adoption (without your engineers hating it). Here’s a good playbook for incorporating AI into your engineering team in a way that sticks.

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

I’m grateful that the most persistent stress in my life comes from following my favorite sports team. There are worse things in life. But one of these years, I really would like my hometown team to, just one time, raise the trophy.

[blog] Gemini 2.5 Flash Image now ready for production with new aspect ratios. The best image model in the world is now GA. These new aspect ratios are terrific.

[blog] Building on the bananas momentum of generative media models on Google Cloud. What stood out to me here was a bulleted list at the bottom that clearly explained when to use which model. We’re getting better at doing that.

[blog] Inside Husky’s query engine: Real-time access to 100 trillion events. That’s a lot of events. The Datadog team explains their event store and how its query engine works.

[blog] Gemini CLI: Discover, configure, debug, and document a Google Cloud feature (Part 2). These AI tools can help everyone learn new things. One of our technical writers, Moi, explains how she used the Gemini CLI to go deeper on a product area she’s writing about.

[blog] Two strategies to succeed when AI seems to be eroding jobs around you. This is relevant to the above post. Tom’s a tech writer here at Google, and has great perspective on what skills need further investment.

[blog] Real AI Agents and Real Work. We’re learning more about human-agent interactions. Agents are capable, and still require our judgement to decide what’s worth doing with them.

[blog] Examples are the best documentation. Great post. People who deliver projects and products need to consider all the examples the user needs access to.

[article] OutSystems Launches a Low-Code Workbench for Building Enterprise AI Agents. Of course low-code vendors will get into the vibe and agent game. They’re irrelevant if they don’t!

[article] Salesforce launches enterprise vibe coding product, Agentforce Vibes. See previous point. This seems natural for Salesforce, as they’ve got piles of data for you to ground on.

[blog] Introducing Microsoft Agent Framework: The Open-Source Engine for Agentic AI Apps. Do we need more agent frameworks? Sure, why not. This one targets .NET devs.

[blog] Designing agentic loops. What does a responsible YOLO mode look like when using your coding agent of choice? Simon’s got a helpful post on designing good agentic loops.

[blog] Meet Jules Tools: A Command Line Companion for Google’s Async Coding Agent. This is cool. Control and interact with this remote agent via the CLI.

[blog] Angular support for generating apps in Google AI Studio is now available. Excellent! It’s now simple to vibe code web apps that use Angular and then deploy them.

[blog] Build Your Foundation First: The Hard Truth About Successful AI Deployments. No skipping ahead. There are prerequisites required to get good AI results.

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

Had a day-trip to Sunnyvale to present to a room full of Google Cloud partners. Wasn’t feeling 100%, but still seemed to go ok. I referenced something from today’s reading list, which proves to me that I should keep doing the work to uncover these pieces 🙂

[blog] Effective context engineering for AI agents. Exceptionally good writeup that helps us better understand what context engineering is all about, why it matters, and how to approach it.

[blog] Cloud CISO Perspectives: Boards should be ‘bilingual’ in AI, security to gain advantage. I think a lot of non-engineering teams are thinking about how to become more fluent in AI so that they can make better decisions.

[article] Boards lack AI savvy as adoption advances: PwC. I can see why the above advice is so important. Most board members are flying blind.

[paper] On the Use of Agentic Coding: An Empirical Study of Pull Requests on GitHub. I haven’t seen a study like this one. It explores pull requests created by Claude Code, and how agent-generated PRs differ in size, acceptance rate, and such.

[blog] 100X Faster: How We Supercharged Netflix Maestro’s Workflow Engine. There’s probably no action item for you after reading this post, but it’s ok to just enjoy learning about what other people built.

[blog] Responsible Agents: A Phased Approach on Google Cloud. Excellent post. Here’s a well-rounded look at making responsible decisions during design, development, evaluation, deployment, and operations of agent systems.

[blog] Securing MCP Servers with Spring AI. Learn how to build MCP servers, but pay special attention to how you secure them too.

[blog] Forecasts and data insights come to BigQuery’s MCP and ADK tools. Or use turnkey MCP servers that already have great tools available in a secure fashion.

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

Spent some free time today building an AI agent with persistent memory. Everything’s not working yet, and it reminds me how annoying out-of-date LLMs can be. Hard to get accurate advice on new technologies!

[blog] The common sense unit of work. Many things—how you plan, what you report, your approach to AI—depends on what you consider a unit of work.

[blog] The Agency Spectrum: An AI Risk Management Framework. This makes sense to me. What levers of governance do you have at your disposal when building an agent system?

[blog] Your employees are already AI-augmented. Your enterprise isn’t. Here’s why that matters. I’ve seen a few pieces on this, and observed this in real life. People are using whatever tools help them. That doesn’t always align with corporate initiatives or buying patterns.

[article] Legacy tech woes fuel IT upgrades as CIOs grapple with costs. Somewhat related to the previous entry. Most enterprise leaders are navigating dense portfolios, while employees figure out how to get today’s work done.

[article] Flow State to Free Fall: An AI Coding Cautionary Tale. Great analogy, and many of us who work with these tools understand the pain of climbing too fast with AI.

[blog] Everyone Wants To Grow — Few Want To Do The Work. You won’t advance—or see your team advance—without intentional investment. Many folks talk a good game, but not many put in the work.

[blog] What is “good taste” in software engineering? Great post. It takes time to develop “taste” and by the time you have the experience, you might be too stubborn about “best practices” to adapt!

[article] How do developers collaborate with AI agents when solving real-world software engineering problems? It’s probably confirmation bias, but this data affirms what I’ve been seeing and believing.

[blog] Building a Next-Generation Key-Value Store at Airbnb. You should buy versus build in about 96% of scenarios here, but it’s still cool to see what Airbnb built.

[article] The YouTube Tip of the Google Spear. Fascinating writeup of what’s still an under-valued service: YouTube.

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

I’m mostly adjusted back to Pacific Time. I had a great weekend with visiting family and was ready to get going today.

[blog] Creative Java AI agents with ADK and Nano Banana. Cool post. How do you build an agent that can process image results and then use those in further agentic steps?

[blog] Give your AI eyes: Introducing Chrome DevTools MCP. This will be a top-3 MCP server for a lot of web devs. Check out this post for all the ways you can bring Chrome actions into your agent or IDE.

[blog] Introducing Claude Sonnet 4.5. Impressive update from Anthropic, and some big statements. They’ve also updated their coding tools. You can use their latest model on all the clouds, including Google Cloud and Vertex AI.

[blog] The AI coding trap. I’m learning a lot from people who are sharing their experience using AI within real engineering teams.

[blog] Development Productivity, Not Developer Productivity. Measure and improve the whole system (or at least, the actual bottlenecks) if you want to see tangible improvements.

[youtube-video] What is GenKit Go? This framework (for TypeScript, Go, and Python) is underrated, but also fairly popular among those who want a clean way to build AI apps.

[blog] What are popular AI coding benchmarks actually measuring? Educational post for those who want to understand what’s good (and bad) about the current crop of coding benchmarks.

[blog] Flight Recorder in Go 1.25. I’m very impressed by the thoughtfulness of this built-in solution for grabbing trace data when something goes wrong in my Go app.

[blog] 90%. Here’s another terrific experience report. AI tools are writing 90% of Armin’s code on recent projects, but that’s because he’s providing a strong guiding hand.

[blog] Documentation. Another short title, but another strong punch. Lee offers up then traits of incredible developer documentation, and it reflects many of the things my team is chasing now.

[article] Why the Frontend Should Embrace Platform Engineering. I haven’t seen this perspective before, and it’s good to see a focus on serving all types of devs with our platforms.

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