Category: Daily Reading List

Today was my last full day in Beijing before flying home tomorrow (Friday) evening. I was able to spend more time with industry analysts, partners, and customers today, and have enough feedback to last me the rest of the year!

[blog] Real-time in no time: Introducing BigQuery continuous queries for up-to-the-minute insights. To me, this is about simplification. Maybe I don’t need an extra component to analyze data streams? Now it’s baked into BigQuery.

[blog] Streaming BigQuery Data Into Confluent in Real Time: A Continuous Query Approach. Here’s a related demo post from the folks behind Apache Kafka. Connecting streaming solutions to in-warehouse analysis is cool.

[blog] How Indeed Replaced Its CI Platform with Gitlab CI. The engineering team at Indeed explains why they migrated from Jenkins to GitLab, and how they did it without automated migration tooling.

[paper] The State of FaaS: An analysis of public Functions-as-a-Service providers. Here’s a new paper that considers ten different function-as-a-service providers and looks at supported configs, regions, pricing, and more.

[blog] Announcing LangChain on Vertex AI for AlloyDB and Cloud SQL for PostgreSQL. This is good for those building RAG-style solutions or storing state for chat-based apps. It also highlights our Vertex AI Reasoning Engine which feels like LangChain-as-a-Service.

[blog] Deep dive into function calling in Gemini. This qualifies as a deep dive. Mete explains some sophisticated scenarios, and shows off a new capability to automatically call tools.

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

Another busy, but fascinating day in Beijing. I delivered my keynote, and spent most of the day learning from customers and local analysts.

[blog] Configuration editing is imperative. Brian wonders why there aren’t more tools to assist with the critical ability to edit system configurations.

[blog] DCPerf: An open source benchmark suite for hyperscale compute applications. Meta is open sourcing this benchmarking tool in the hopes of creating an industry standard.

[blog] Continuous Delivery on Google Cloud with Gitlab CI/CD and Cloud Deploy. I like this end-to-end overview of getting an app from source control to a compute service.

[article] When Your Boss Suddenly Reduces Your Scope. Not a great feeling! This article explores how to react in a smart way.

[article] Microsoft joins CrowdStrike in pushing IT outage recovery responsibility back to Delta. This has gotten spicy. There’s a shared responsibility to software and services that seems to be at the heart of this back-and-forth.

[blog] Mechanical Orchard Secures $50 Million to Safely Transition Large Organizations Off Risky Legacy Software. A lot of my friends from Pivotal are over at MO, and they’ve clearly landed on a modernization approach that resonates.

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

Today was a busy one as I had 10+ hours of education about the developer market here in China. So interesting!

[site] Minimum Viable CD. What qualifies as “continuous delivery”? This manifesto-ish site calls out nine criteria including use of continuous integration and rollback-on-demand.

[article] Introduction to Flutter — How Does It Compare to Tauri? Here’s a good intro to some of the aspects of a multi-platform app framework like Flutter.

[blog] Investigation of a Cross-regional Network Performance Issue. I enjoy posts like this! The Netflix crew had to figure out where the issue was coming from, and debugged a few layers of their stack.

[article] The Evolution of Embeddings. This was a very understandable write up about how embeddings work, and how they’ve evolved in the Transformers era.

[article] Analyzing the 2024 Stack Overflow Developer Survey: Productivity and DevEx (Part 1). Most developers aren’t happy. Why is that? This post looks at some data from the massive SO survey.

[blog] Top 7 Signs You’re Doing DevOps Wrong. However you define “DevOps”, there are some activities that might indicate you’re not doing it right.

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

I had a good first day in China after a long, but uneventful, flight. Tomorrow starts the work meetings and opportunity to learn about the local market. In the meantime, here’s an unusually-timed reading list!

[article] Are You a Micromanager or Too Hands-Off? This article offers up some red flags for managers, and how to course correct.

[article] Cognitive load drivers. I liked the categories listed here. Find fifteen drivers of cognitive load, and consider a list like this when finding (and fixing) friction in your organization.

[blog] Unlocking the Power of LLMs: Gemini Function Calling Simplified with Vertex AI SDK. If an SDK isn’t adding convenience functions, there’s not much point in using one! This Python SDK really does make it so much easier to send supporting “tools” to the Gemini model.

[blog] This one important fact about current AI explains almost everything. Gary says that current approaches to machine learning are bad at handling outliers, and this is being ignored by the loudest AI champions.

[blog] How I Use “AI”. An excellent reminder that LLMs are offering a lot of tangible value right now, in their current form.

[article] 7 Techniques for Database Performance & Scaling. If you’ve been around databases for a long time, you might not find anything new here. But it’s a good refresher of foundational practices for optimizing a database.

[article] Google’s AI comeback: New Gemini models dethrone OpenAI in shocking upset. Those early missteps were educational, and also part of the journey itself. I’m glad to see us continuing to ship good tech, and attract great people.

[blog] Bigtable transforms the developer experience with SQL support. We see an unbelievable number of API calls each month for Bigtable, and now that you can use SQL to query, I expect that number to go up again.

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

It was a productive Friday, and tomorrow I’m off to Beijing for a week. I’ll be speaking at Google I/O Connect China and getting immersed in the needs of local developers. Should be something. I’ll keep up the daily reading list, but on a very different posting schedule!

[article] How does Google describe software quality? This looks at a recent paper of ours and explores the four types of quality we measure and look to improve.

[blog] The Product Model and Org Design. It helps if the shape of your org complements your org’s priorities. But do you have to reorg to do things like adopting the product model? This post says “no.”

[blog] Detailed Guide of How To Set up MLflow on GCP in a Secure Way. MLflow is a popular MLOps platform and this post shows you how to run this open source software on Google Cloud Run.

[blog] Tech Leaders Target Scale And Enhanced Productivity With 2025 Budgets. Analyst firm Forrester is sharing new guidance for those starting 2025 budget planning. More cloud and AI, fewer single-function apps. Here’s another Forrester post targeting executives.

[blog] TPU transformation: A look back at 10 years of our AI-specialized chips. Our TPUs have been around a while, but they’re getting a moment in the sun right now. Major companies are using them, and Google has an advantage by being able to train and serve on GPUs or TPUs.

[blog] Introducing workflows beta: a new way to create complex AI applications with LlamaIndex. It’s not surprising that we’re seeing higher order tools coming for AI app builders. Not everything requires direct API calls or code-driven orchestration.

[blog] “Death of a Salesforce”: Why AI Will Transform the Next Generation of Sales Tech. Hot take from a16z! They see big changes coming to sales workflows and core systems of record.

[blog] Our Next Phase of Growth. The talented crew at Character.AI are partnering with Google Cloud more closely, and a handful of their great folks are joining Google DeepMind.

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

Today’s reading list offers up items about the promise of AI, the skepticism of AI, and the implementation of AI. Plus other stuff too.

[blog] Different Shades of PLG: Free-Trial or Freemium? Does a forever-free product with feature limitations beat out a full-featured free trial? Sometimes, not always.

[blog] Launching a landmark partnership to support AI startups with Y Combinator. It seems both helpful and wise to assist startup founders in their tech journeys. Lots of these folks come to Google Cloud by default, but we’re making it even more attractive.

[article] How to Ask for Help Without Making Yourself Look Bad. This looks like good advice. I often don’t “start strong” and will focus on that.

[article] How to Implement a GenAI Agent using Autogen or LangGraph. Lak takes a couple of agent frameworks for a spin and shows how to use them with different LLMs.

[article] This Week in AI: Companies are growing skeptical of AI’s ROI. Be skeptical. Don’t take a vendor’s word for it. Run your own evaluations and studies, and see if AI can fix what needs fixing.

[blog] Can’t stop, won’t stop: More innovations from the Google Cloud database portfolio. We announced some solid updates at Cloud Next Tokyo today. Get SQL support in our Bigtable database, use the new graph database features in Cloud Spanner, and get a new tier of perf and uptime for managed SQL Server. News here and here.

[blog] Introducing Spanner Graph: Graph databases reimagined. I thought this was an informative deep dive into this new service. For those using graph databases, this might be a chance to consolidate.

[blog] Beyond temperature: Tuning LLM output with top-k and top-p. You could skip past this, but Karl does a great job writing an engaging post that helps us understand how to tweak the creativity of an LLM’s output.

[blog] New strides in making AI accessible for every enterprise. Wow. Now get support for 100+ languages in Gemini 1.5 Flash and Pro, along with a model uptime SLA, and dramatically lower cost.

[article] Report: High Risks to Software Supply Chains are Commonplace. Too many apps, too many alerts, and too little analysis.

[blog] Retrieval Augmented Generation (RAG) with Cloud SQL for MySQL. Very solid tutorial from Julia here, and I like the reuse of “boring” technology like MySQL.

[blog] 3 new Chrome AI features for even more helpful browsing. I don’t do a lot of fancy things with my browser and likely use a tiny subset of what it’s capable of. This upcoming feature that lets me ask question of my browsing history? I like that.

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

Half of today’s list was about AI, which is probably par for the course right now. Even if you’re a skeptic of the applicability or staying power of generative AI, keep paying attention to the developments!

[blog] Sentiment analysis with few-shot prompting. Are you familiar with the ideas of no-shot prompting and few-shot prompting? This is about priming the LLM with good examples to follow. Here’s a post with a practical example.

[blog] Faster continuous integration builds at Canva. Giant post that will be interesting to folks that want to think their CI pipelines and are struggling with slow build times.

[blog] Smaller, Safer, More Transparent: Advancing Responsible AI with Gemma. We just shipped a new 2B open model, and launched a new tool that gives you insight into the model’s inner workings. Looks cool. Some news here.

[blog] Gemma Scope: helping the safety community shed light on the inner workings of language models. Here is that set of tooling. This is great for researchers or those simply wanting to explore. Also check out the tech paper.

[article] Is It Time to Pivot Your Strategy? Thinking of making a strategic shift? Why? This post looks at some questions to ask yourself.

[blog] Vertex AI Search: Leverage the Power of Google Search and Gemini for Your Information Needs. This is a powerful, underrated service. Holt explains some details of building your personalized search engine.

[blog] AI And The End Of Software Development As We Know It. And I feel fine? Diego has been on this topic for years and has good perspectives on where the industry is going.

[blog] Apache Flink® on Kubernetes. Learn how Airbnb runs this popular data stream processing platform atop Kubernetes.

[blog] Announcing IAM group authentication in Cloud SQL. The fewer sets of unique credentials, the better. Now you can authorize database users with IAM roles. And use AI assistance to observe and troubleshoot the database.

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

Today was the opposite of yesterday. Only a handful of quick meetings, and a lot of time to get other work done. I’m so used to days full of meetings that ones like today feel weird. You?

[blog] How Google uses AI to reduce stop-and-go traffic on your route — and fight fuel emissions. Smarter stoplights? That’s some AI that we can all rally around.

[blog] More enjoyable Code Reviews with Gemini. Cool post that explores a good portion of the code review process, and how to use AI and automation to make it better.

[blog] Java 21 Virtual Threads – Dude, Where’s My Lock? This post from the Netflix engineering team goes deep into how they diagnosed a deadlock issue with Java virtual threads.

[paper] Apple Intelligence Foundation Language Models. To my untrained eye, this looks like a fairly transparent paper from Apple into how they built and evaluated their foundation model.

[repo] Kratos. There’s no shortage of web frameworks for most every programming language. Go seems to be getting more and more of them. This one is popular, and looks fairly rich.

[article] Gen AI Increases Workloads and Decreases Productivity, Upwork Study Finds. Spicy! This (apparently controversial) survey shows a disconnect between executives and employees on productivity benefits, and workers aren’t getting help after having tools shoved at them.

[blog] Anatomy of a great principle. Brian looks at guiding principles and what makes a good one. With bonus example included from his world of enterprise architecture.

[blog] Google Cloud Private Marketplace, now GA, helps control costs and maintain governance. Get more control over what your folks can install in the cloud.

[blog] Why Large Organizations Struggle With Disruption, and What to Do About It. Meaty post from Steve with a lot of context and guidance for how disruption plays out.

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

Happy Monday. It was a busy day here, with 30 free minutes between 830am and 530pm. Most of my reading list time was early in the morning. Enjoy!

[article] Is Your Team Playing It Too Safe? We can say “incentivize big bets and don’t punish failure”, but people need to see that in action to believe it!

[blog] Analyzing video, audio and PDF files with Gemini and LangChain4j. Excellent post, not just because it uses tech I like. Guillaume used relatable examples and gave me confidence I could follow along.

[article] “Authentic” is dead. And so is “is dead.” I feel attacked by this post, which also be a phrase I can’t use anymore. I really liked Jason’s take on overused and nebulous terms that should be replaced by something with meaning.

[paper] Retrieval Augmented Generation or Long-Context LLMs? A Comprehensive Study and Hybrid Approach. Long context is a powerful tool in your toolbox. The model predictions mirror RAG in many cases, and this research paper proposes a better-together approach.

[article] The Guide to Going Multi-Product: 11 Tactics for Tackling Your Next Bet. If you’re looking at expanding your product portfolio (internally or externally), this article offers some good advice.

[article] SysAdmins split on AI’s impact, but still want training. It doesn’t feel shocking that IT admins will be slower to adopt AI. There are some compelling use cases, but I understand the hesitation.

[blog] Portable Training Data Generation for Supervised Fine-tuning: A Reverse RAG approach! Training set generation using an LLM? Be skeptical, but read along to see if this has merit.

[article] Why You Feel Underappreciated at Work. This was a good read, and encouraged a proactive approach versus just feeling sorry for ourselves.

[blog] DAGify: Accelerate Your Journey from Control-M to Apache Airflow. There’s so much wildly popular tech out there that I’ve never heard of. It’s humbling. Control-M is heavily used workload automation software. If you want to move to an open Airflow platform instead, this new open tool makes it easier.

[blog] Go 1.23: Interactive release notes. How easy are you making it for folks to try out the new thing that you released? Anton builds these “interactive release notes” for Go and I love it.

[blog] Cloud Run Idle Instance Conundrum. If you need “serverless” instances to stay online longer, you can sometimes be out of luck. but with Cloud Run, you can keep CPU always allocated.

[article] Thinking Like an Architect. Good architects are a huge value to a team; bad ones cause generational impact on the org. Gregor has a good piece on how good architects amplify smart people in the org and aim to make smart contextual decisions.

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

I had a quiet day today with just one meeting, and a chance to get some writing done. I also had a chance to play more in depth with the IDX developer workspace for building apps. Neat stuff. More to learn!

[article] How Spotify maintains team autonomy at scale. Many folks refer to Spotify when discussing/debating tech team structures. This piece goes into depth on how they enable team autonomy.

[blog] Whisper Goes Wall Street: Serving Speech-to-Text with Ray Serve and Cloud Run — Part II. Here’s a creative look at serving up a fine-tuned translation model on a serverless runtime.

[blog] Meet Chrono, our scalable, consistent, metadata caching solution. I can imagine that Dropbox needs a strong caching solution. This post digs deeper into their internal system.

[blog] The Balancing Act of Chasing Gold. Steve looks at the “tightrope” effort of performing at the highest levels. Applies to business and Olympic athletes.

[blog] AI achieves silver-medal standard solving International Mathematical Olympiad problems. Speaking of Olympiads, our AI system is delivering some fairly elite mathematical performance.

[article] Using Persuasion to Spur Lasting Organizational Change. When the “godfather of persuasion” talks, I listen. Or I’ve been conditioned to listen. I’m not sure. Either way, good article on creating an environment to land durable changes.

[blog] Package and deploy your machine learning models to Google Cloud with Cog. Want an easier way to work with ML models? This post explores an OSS tool for packaging models into container images.

[article] Google and Formula E break world record for most participants in generative AI hackathon. Wild! There are literally record keepers for everything. I’m happy to see so many folks coming together to learn new tech.

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