Category: Daily Reading List

Walked in the door at midnight last night after flying home from a business trip, and that 5am alarm was rough! But today was a productive catch-up day, and I hope many of you are enjoying the tail end of Summer.

[blog] Structured logging in Spring Boot 3.4. A well-instrumented system is better than one that isn’t. This post shows how Java apps using Spring Boot can have nicely structured logs with little effort.

[blog] Gemma explained: What’s new in Gemma 2. The architecture for this latest open model version differs from its predecessor. This post explains the differences and the impact they had.

[article] Rackspace Goes All In – Again – On OpenStack. I can’t say I’ve heard much about either lately. But private cloud fans may like seeing a fully managed OpenStack offering form Rackspace.

[blog] Maximize your LLM serving throughput for GPUs on GKE — a practical guide. This is a good set of questions and answers for those standing up infrastructure to host models.

[article] Google debuts free ‘Prompt Gallery’ in AI Studio, supercharging developer tools. At this point, it seems we’re offering three of the most important open tools for those learning and using AI: YouTube, Colab, and AI Studio. This article highlights awesome improvements to the latter.

[blog] Achieving Gold Medal🥇-Level Data Science Communication with Gemini and Vertex AI. Fun and insightful demo from one of our expert folks.

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

I had an enjoyable day here in Kansas City, meeting with four different customers. It was very helpful to get reminders of “real world” challenges and use cases. Flying home now, and back in the San Diego saddle tomorrow.

[article] Research: What Happens When Influencers Turn Off Comments. The comment section. You’ve never find a more wretched hive of scum and villainy. But keep it on!

[blog] Platform engineering problems: can ops actually do product management? My friend Coté has a good long post about platform engineering and the need for product thinking.

[blog] Cloud Run GPU: Make your LLMs serverless. I’m starting to think that this is a big deal. Easier access to LLM hosting and experimentation will only expand the realm of use cases.

[blog] Serverless Sucks? But why all this serverless stuff? Just use servers and make life easier. Not so fast, says Derek.

[blog] Postmortem: Latency spikes in Apache Hbase on kubernetes. When you stack platforms on top of platforms, troubleshooting becomes more of an exploration.

[blog] Building Quality Software: 4 Engineering-Centric Techniques. Here’s some useful guidance from the Software Engineering Institute. Good instrumentation is a wise callout.

[blog] gRPC Communication Between Go and Python. If all your app components are written in the same language, kudos. That makes life easier. But for most everyone else, we need to think about interop and how to get these pieces talking. Here’s a good example.

[blog] Introducing delayed destruction for Secret Manager, a new way to protect your secrets. Use a secret story, I don’t care which. But if you use ours, there are new ways to get notified if a secret got deleted.

[blog] I’m Tired of Fixing Customers’ AI Generated Code. I don’t envy consultants who are going to have to come fix apps written entirely by well-meaning non-programmers.

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

I made it to Kansas, and I’ve got a bunch of fun demos lined up for customers tomorrow. How do you feel about doing (or watching) live demos? I know there’s a risk, but it feels more genuine than playing a recording.

[blog] Run your AI inference applications on Cloud Run with NVIDIA GPUs. This isn’t “classic” serverless. Cloud Run is the best modern runtime, and now you can run LLMs and GPU-powered apps here. Also, Cloud Run functions is now a thing!

[blog] 10 Google AI Tools for GenAI Developers You Need to Know. A couple of these were new to me, believe it or not. There’s some goodness here for those creating prompts, apps, or just experimenting with AI.

[article] How to Set Transformation Targets That Actually Drive Change. We’re in a planning cycle now at work, so this article feels timely.

[blog] Using the Kubernetes Resource Model to provision Cloud infrastructure. I’m still a big fan of this approach over “just” the Terraform approach to provisioning infrastructure. But, it hasn’t yet become mainstream.

[blog] Transformers in music recommendation. Neat stuff on how to build a more contextual and useful recommendation engine for YouTube Music.

[article] C-suite enthusiasm over generative AI wanes, putting pressure on quick wins. Entirely predictable, and frankly, welcome. Let’s get past the extreme hype phase and start solving genuine use cases.

[guide] Use generative AI for utilization management. This reference architecture is targeted at health insurance companies, but it offers a blueprint for anyone doing automation of document processing.

[article] Angular’s Efforts To Be Easier and Developer Friendly. A JavaScript framework focused on performance, developer experience, and AI smarts? I like that.

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

I’ve got a short trip up to Kansas City tomorrow to chat with some customers and do some live demos. I like working at a place that requires me to constantly update my demos to show off new functionality!

[blog] Gemma explained: An overview of Gemma model family architectures. This is an informative deep dive into the architecture of these open models and helped me better understand how we put them together.

[paper] An Introduction to Reinforcement Learning: Fundamental Concepts and Practical Applications. I didn’t understand all of it, but I got the gist. Skim this paper to learn about this branch of AI that’s different from supervised or unsupervised learning based on datasets.

[docs] Best practices for securing SSH access to VM instances. This topic is the opposite of flashy AI stuff, but there are probably 10,000 people SSH-ing into a server for every one person building an AI app today!

[blog] Developer Content That Ranks in 2024. What type of content attracts the most developers? Adam takes a look.

[blog] Use intelligent suggestions for table conversion in Google Sheets. Little feature, big value. Anything that saves me time formatting data is a winner in my book.

[article] How close is AI to replacing product managers? Generative AI is pretty darn good at some of the core tasks of a PM.

[blog] C4 VMs now GA: Unmatched performance and control for your enterprise workloads. If you don’t care about price performance, keep scrolling. Otherwise, check out this great new VM type.

[blog] Does Open Source Work as a Long-Term Business Model? It doesn’t look like Yugabyte is going to copy CockroachDB’s license change. This post looks at the rise of PostgreSQL and the risk of positioning against it.

[article] Platform Engineering: The Next Step in Operations. Long read, but a good one about some of the fundamentals of platform engineering.

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

I had a fun weekend with visiting family, and now I’m back to regular work. Today’s reading list has a lot of AI-related content, but coming at it from a variety of angles.

[blog] Is GenAI Falling Short? Not In The Contact Center. I like these two examples. Both automated call summarization and support for “infrequently asked questions” are valid use cases for AI today.

[blog] Is It Time to Version Observability? (Signs Point to Yes). Another banger from Charity who distinguishes Observability 1.0 from Observability 2.0 and why the future is “wide, structured log events.”

[blog] PyTorch is dead. Long live JAX. Edgy title! But also a very well-reasoned argument for why DeepMind’s JAX framework may be a better long term choice.

[blog] Build a text-based adventure game with Gemma 2. Ok, this is probably NOT a use case you’ll use AI for all the time 🙂 But still cool.

[blog] Gemma for Streaming ML with Dataflow. This, however, is a strong use case for AI. Add it to data processing pipelines for actions like sentiment analysis.

[paper] Imagen 3. This looks like a huge jump in performance for text-to-image generation. Read this new research paper for information about this Google model, and some human ratings of generated images.

[blog] Postgres is all you need, even for vectors. Maybe you don’t need all those specialized services when a few workhorse products can solve a myriad of use cases?

[blog] Your infrastructure resources, your way, with new GKE custom compute class API. It took me a moment to “get” this feature, but I think I have it. By defining choices of infrastructure for your Kubernetes cluster, you can scale or auto-migrate into the desired “class.” Sweet.

[article] Onboarding to a ‘legacy’ codebase with the help of AI. These tools aren’t perfect, but they can often point you in the right direction.

[blog] Improving Design Reviews at Google. We’re constantly studying data about how we work at Google, and experimenting with ways to make it better. This post summarizes a paper about our efforts to improve the design review process.

[blog] Deploy Meta Llama 3.1 405B on Google Cloud Vertex AI. The folks at Hugging Face explain how to run Meta’s latest model on Google Cloud’s flagship AI service.

[blog] Why Google Kubernetes Engine (GKE) Leads the Pack: A Comprehensive Comparison of Managed Kubernetes Services. Here’s a solid comparison of cloud-based Kubernetes services, with GKE looking like a strong choice.

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

It was a fun Friday with more vacation time, and a morning reading list. Enjoy your weekend!

[blog] The Atlassian method: The power of developer joy. Whether you’re serving internal or external devs, how focused are you on “joy” and ensuring developers like using what you deliver?

[article] The Product Model in Outsourcing. Fascinating piece from Marty and Josh that offers specific advice for those that want more product-oriented thinking from their outsourced/service partners. It requires major changes on both sides!

[blog] Reliable Generative AI Multi-Agent Solutions with Gemini, RAG and Grounding in MongoDB Atlas. Big demonstration of an agent-based app that grabs trusted data from a MongoDB database instance.

[blog] Serverless Is Trending Again In Modern Application Development. Great post from Devin that looks at why “serverless” (redefined) is making a comeback. I’m seeing some of this too.

[article] MIT researchers release a repository of AI risks. This is useful for researchers, customers, and even vendors. I’m sending this to my docs team to see if we should improve our documentation.

[blog] Bash to Terraform with Gemini 1.5 Flash. All those scripts you have laying around might be turning into tech debt. Maybe converting to something like Terraform is more maintainable? Either way, this is an interesting look at how to do that.

[article] Walmart used AI to crunch 850M product data points and improve CX. The financial benefits aren’t spelled out here, but I’m assuming this newfound insight is adding value.

[blog] Meet the nine new web delivery partner integrations coming via Service Extensions. Augment the request traffic through our Load Balancers? Cloud customers can use these partner solutions as part of our Service Extensions.

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

Took vacation today to show my visiting nephews around San Diego. Most of the day was at the Zoo, which meant all of my reading happened before 8am!

[article] Analyzing the 2024 Stack Overflow Developer Survey: How Engineers use AI (Part 2). I’m glad we keep learning more about the real usage of these AI-powered dev tools. And, what devs need to start using them more.

[article] Google Angular Lead Sees Convergence in JavaScript Frameworks. Lots of interesting perspectives here about web framework similarity, and convergence.

[blog] Kubernetes v1.31: Elli. The latest Kubernetes release is out, and find a platform that helps you stay up to date.

[article] Are Shared Service Platforms Too Restrictive? Consistency comes at a cost. Can you build a shared platform that meets everyone’s needs? Not easy.

[article] IRS has loads of legacy IT, still has no firm plans to replace it. Yikes. Like many big orgs, there are lots of legacy systems here that need to be retired, but aren’t.

[article] Cockroach Labs shakes up its licensing to force bigger companies to pay. It’ll be fascinating to see if all these licensing shenanigans pay off in the end. If they do, congrats!

[blog] The top AI courses for a summer of learning with Google Cloud. It’s never to late to “catch up” on emerging tech ideas. Well maybe it can be too late, but we’re not there yet with AI. Run through this coursework to up-level your skills.

[article] Experiences from Doing DORA Surveys Internally in Software Companies. Using the DORA metrics to focus your team on what matters most for software delivery is a good deal. But don’t ONLY use DORA metrics for team measurement.

[docs] Short query optimized mode. This is awesome. We added this feature to BigQuery to execute short queries even faster (without job setup). I tried it, and it’s blazing fast.

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

Today’s reading list has a lot of tech deep-dives and how-to articles, which I liked. Get some lessons learned by your peers.

[blog] Developing a Spring Boot Project Using Google Project IDX. I’ve been messing around with IDX lately—it’s a dev environment consisting of a loaded IDE and inner loop tools—and this post shows how to build a Java app in it.

[blog] GenAI Experiments: Monitoring and Debugging Kubernetes Cluster Health. Can generative AI help on-call engineers manage alerts, debug faster, and fix unhealthy clusters? Intuit hopes so.

[article] How to Manage Feedback Like An Olympic Athlete. We all LOVE getting feedback, right? Right? This post is about handling and using feedback.

[blog] I just want mTLS on Kubernetes. You want encrypted traffic within your cluster, but without adding a bunch of overhead. John looks at options.

[blog] Let LLM suggest Instagram hashtags for your pictures. I’m not a big hashtag guy. Even if you aren’t either, you might like Guillaume’s post about using AI to figure out the right image tags.

[blog] Iceberg vs Hudi — Benchmarking TableFormats. Data lake structures are not my strength. But it was good to learn a bit in this post from Flipkart.

[blog] Deep-dive into Cloud DNS Routing Policies. When anything goes wrong, blame DNS. However, it’s still good to actually understand how it’s working.

[blog] 2023 Open Source Contributions: A Year in Review. We build with open source software, and we build a lot of open source software ourselves.

[blog] DoorDash Empowers Engineers with Kafka Self-Serve. Yes, a good platform team can even serve up complex middleware like Kafka to self-service interactions.

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

Jetlag is over, and today was a productive workday. I’ve got family coming into tomorrow, and taking some time off, but I’ll likely keep the reading list rolling!

[blog] Made by Google 2024. Lots of fun hardware announcements today. The AI features of the Pixel are super cool, and the new camera features are rad.

[article] Inside the First Year as a CEO. Whether you’re becoming a CEO or supporting a new one, it’s useful to recognize the playbook. First up, building the right leadership team. Then, looking at strategy and culture. Good article!

[paper] Microservice Vulnerability Analysis: A Literature Review with Empirical Insights. Here’s a new paper that explores the wide attack dimensions of a microservices architecture. It looks at infrastructure (like gateways and runtimes) along with authentication and data access.

[blog] Create a powerful Kubernetes security duo with Custom Org Policy and Policy Controller. Make it easier to do the right thing. Shift down. Here’s an example of establishing platform controls that make a Kubernetes cluster safe by default.

[blog] Generating, transforming, and patching Kubernetes configuration with Kustomize. Speaking of Kubernetes, I liked this post from Brian that explores the need for a Kubernetes configuration tool. He then explains how kustomize came to be.

[paper] Discovering Datasets on the Web Scale: Challenges and Recommendations for Google Dataset Search. Here’s a new paper from Google Research that looks at a search engine for datasets.

[article] How developers want to use AI tools. This article looks at research from Microsoft about how their devs look at AI tooling. It shows that they, like most, want help with routine tasks, not super advanced activities.

[blog] How to Build Low-Latency Global Applications with Google Apigee and YugabyteDB. There are benefits and tradeoffs to a running an app that serves a global audience. This post looks at an architecture that delivers performance and availability across multiple regions.

[article] 5 Lessons For Building a Platform as a Product. Colin offers some real-world advice for platform builders and those trying to get the most out of their stack.

[blog] DeepEval and Vertex AI. Can you unit test the output of an LLM? Yes you can, as Mete shows off here.

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

I’m back from China after a fun, but tiring week. And my flight home was cancelled (after five hours of sitting on the plane) Friday night, before getting rescheduled for Saturday night. But, I got a story out of it, and made some friends along the way.

[blog] Gemini 1.5 Flash price drop. Simon noticed our big price change. Get one of the best models at the best price, and for the most multimodal use cases. A lot to like! Related post from our product folks.

[paper] The AI-Native Software Development Lifecycle: A Theoretical and Practical New Methodology. This proposes a future state where code generation is free and instant. Most devs spend time verifying, and AI impacts ALL stages of the SDLC.

[blog] Level up your Kubernetes security with the CIS GKE Benchmarks. This is cool. Get these 3rd party security recommendations applied to your Kubernetes cluster in Google Cloud.

[blog] How we migrated onto K8s in less than 12 months. From the tech team at Figma. They seem happy with the results.

[blog] Step-by-Step Guide to Integrating Spring Boot with OpenTelemetry and GCP. Good post for those looking to add more instrumentation to their Java apps.

[blog] Individual efficiency vs administrative efficiency. So great. Jason does an excellent job looking at the tradeoffs of what’s good for the person, versus what’s good for the team. And, giving advice for how to approach it.

[blog] Farewell to overprovisioning: How to unlock cost-effective elasticity with Spanner. Not every workload is spiky or requires extreme elasticity. But when you do, it’s cool to be able to access such flexible cloud services.

[blog] Docker Best Practices: Understanding the Differences Between ADD and COPY Instructions in Dockerfiles. I mostly see and use “copy” versus “add” in my Dockerfiles, but it’s good to know the real difference.

[blog] Gemini Nano running locally in your browser. An LLM built into the browser is fascinating, and now possible.

[blog] Introducing Approximate Nearest Neighbor (ANN) search to Spanner. This might not be needed (yet), but as your datasets grow, you’ll want the cost and latency relief from this related approach.

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