#Announcements

• AutoMapper and MediatR Licensing Update | Jimmy Bogard - An update around licensing for popular .NET libraries MediatR and Automapper, which indicate dual license is the likely option. Meaning companies and individuals would be required to pay for commercial use. Pricing has not yet been determined.
• OpenAI Said to Be In Talks to Buy Windsurf for About $3 Billion - Bloomberg - Rumors is that OpenAI will acquire WindsurfAI IDE for $3 billion USD in the next week. This is massive, as Windsurfai IDE has become well known as the leader in AI powered IDE, and Microsoft has punished Cursor to the degree that the market is becoming smaller. With Microsoft being on the board of OpenAI, it seems like a move to get a strong grip on the programming AI market.
• Kaggle and the Wikimedia Foundation are partnering on open data. - Wikimedia, the enterprise APIs to access Wikipedia data and by the organisation managing Wikipedia, partners up with Google’s Kaggle company to host their data set. Kaggle is the training platform, challenges and large source of big datasets for machine learning. I’m sure this partnership will help both platforms for the best.
• JetBrains announces a free tier for its AI tools - SD Times & JetBrains debuts free AI tier and Junie coding agent in IDEs - JetBrain’s AI agent, Junie, is now GA and offers a free tier, Pro ($10/user/month) and Ultimate ($20/user/month) license. Note if you have JetBrain’s All Products Pack subscription, it’s include ($28.90/user/month). The non-agent AI Assistant is already available and allows you to use any of the big 3 AI models.
• GitHub Copilot Highlights in Visual Studio 17.14 Preview 3 (Available Now) - Visual Studio Blog - GitHub Copilot comes to Visual Studio, yet the implementation is a bit behind from Visual Studio Code.
• Docker Desktop for Mac: QEMU Virtualization Option to be Deprecated in 90 Days | Docker - Apple silicon Docker QEMU virtualization will be deprecated soon, in favour of using Apple Virtualization Framework.

#.NET

• Who can cancel Carmen Sandiego? - Ayende @ Rahien - A great tip for C# .NET, you can register a delegate for when a cancellation token gets cancelled.

#Architecture

• Microservices to Monolith - CodeOpinion - The classic discussion of why microservices are painful, but is your problem actually you abstractions and coupled design around your business capabilities.

#Technical Articles

• The case of the UI thread that hung in a kernel call - The Old New Thing - Raymond Chen back at it with another fun painful C++ issue, going on to explain how a thread should not be able to suspend another thread which it doesn’t own, or else you’ll potentially get into a deadlock state.

#AI

• What the heck is MCP and why is everyone talking about it? - The GitHub Blog - Another great explanation of Anthropic’s open sourced Model Context Protocol (MCP), and how it’s a open, and now common, standard for AI to be able to pull information from data sources such as APIs, files, network/web, databases, etc. MCP allows you to create a client which has context of and can execute, then call your MCP server over JSON RPC, the AI will figure out what arguments it needs to pass into the client and the server will be able to call some custom code like access file content and return it back to the model for it to use in the response. Or even update file content like many editors have implemented.
• Build MCP Remote Servers with Azure Functions - .NET Blog & Franky’s Notes: Making AI smarter with an MCP server that manages short URLs - A nice wrapper around the above described model context protocol, and even using Aspire to run your MCP server .NET project.
• Move Beyond Chatbots, Plus 5 Other Lessons for AI Developers - The New Stack - One thing to pull out of this article is evals, to do automated tests against your LLM. They’re not binary, so testing isn’t simple, but could you prompt an LLM and get a similar sample response back 9/10 times?
• Google used AI to suspend over 39M ad accounts suspected of fraud | TechCrunch - Google seem to be using AI to check Google Ads for deepfake or scamming advertisement uploads. They’ve improved this counter measure detection, resulting in 39 million ad accounts being suspended from being able to advertise on Google.
• Can LLMs earn $1M from real freelance coding work? & Research Paper - A group of researchers picked out 100 software freelance job tasks from Upwork which in total were paying around 1 million dollars to complete, the team then ran different LLMs like Claude and OpenAI to see how many the models could finish. The code was submitted to another reviewer model, which then picked out the best solution. Finally some real software engineers built tests for the submitted code to pass to consider the task to be considered complete. The results were interesting, GPT-4o complete $303,535 worth of tasks, o1 Model: $380,350 and Claude 3.5 Sonnet: $405,325.
• OpenAI debuts Codex CLI, an open source coding tool for terminals | TechCrunch - OpenAI Codex CLI now allows you to use Open AI in your terminal. Note that Claude already had this tool.
• Symbiotic Security launches AI tool for detecting and fixing vulnerabilities in code - SD Times - A new AI agent tool for detecting and fixing security vulnerabilities from your IDE, a pretty handy product.

#Tech News

• Cybersecurity World On Edge As CVE Program Prepares To Go Dark & CVE Program rescued at the last minute after concerns over losing its government funding - SD Times - In a rollercoaster of events, the US CVE (Common Vulnerability and Exposures) program’s U.S. Department of Homeland Security funding nearly expired causing panic for the cyber security community. However, on the day of expiry, the contract was renewed, however, it highlights a potential single point of failure as there is only 1 funder of the CVE program and a lot of reliance on it’s database and open practices, should we be hedging ourselves on relying more on EU CVE program or have many organisations funding the program?
• Zoom is back up after outages this afternoon & Downtime April 16th: App and website not loading - The Spotify Community - Spotify and Zoom go down for a prolonged period. For Zoom it appeared to be a block on a specific service from GoDaddy, causing their zoom.us domain to go down.
• Future Chips Will Be Hotter Than Ever - IEEE Spectrum - A research’s article explaining that microchips are struggling with heat being emitted from more transistors and semiconductors being crammed in together, causing electrical waste and other issues. Researchers are trying to tackle the thermal energy problem from the semiconductor level and try dissipate the energy faster, but we’re likely to see new chips pushing for more power and therefore becoming hotter without any breakthrough in this space.
• Startup funding hit records in Q1. But the outlook for 2025 is still awful. | TechCrunch - We’re likely to see a bit of a collapse of the US venture cap investing due to US economic instability and tariffs. But also, AI funding is soaking a lot of the possible capital. Notably Claude and OpenAI took a majority of last fisical funding.

#Infrastructure

• Did we just make platform engineering much easier by shipping a cloud IDP? – Richard Seroter’s Architecture Musings - Google release an Cloud Internal Developer Platform (IDP) as a service, there’s a lot of interesting AI implementations in this service, showing a very interesting way forward for IDPs. Safe to say Google Cloud is really ahead of the Kubernetes hosting standards.
• Optimizing Our E2E Pipeline - Engineering at Slack - Slack has been using a mono-repo for their frontend which had a larger build & test pipeline time. The team looked at how they can split out the build and tests to optimize build and deploy times with some big wins.
• Conversing with GKE Clusters: Google’s Agent Development Kit in Action | by Sayed Imran | Google Cloud - Community | Apr, 2025 | Medium - An example of how to create an AI agent to interact with your Google cloud kubernetes (GKE) cluster and getting it to control your cluster with natural language, for example scale up my workload. Truely the new world of how engineers will be managing their platforms.
• Azure Files: More performance, more control, more value for your file data  | Microsoft Azure Blog - Huge improvements to Azure Storage Files, with a focus on performance.
• Express Brokers for Amazon Managed Streaming for Apache Kafka | Amazon Web Services
• Stop Using Docker and Local Kubernetes for Dev Environments! (feat. Okteto) - Okteto is a product for developing and testing your app in a non-prod kubernetes cluster, isolated to its own namespace. Although the product looks good, comparing it to Azure Kubernetes Bridge, there seems to be not too many game changing features.
• Speeding up Terraform caching with OverlayFS - Hacking a solution for concurrent Terraform init’s to allow you to run an init per provider concurrently. Neat little win.
• GitOps and IaC at Scale – AWS, ArgoCD, Terragrunt, and OpenTofu – Part 1 & GitHub - gitops-bridge-dev/gitops-bridge - GitOps bridge is a project showcasing patterns and best practices for bootstraping your initial kube cluster and then managing everything through GitOps via ArgoCD or FluxCD, by examples and diagrams.
• Kelsey Hightower on Nix vs. Docker: Is There a Different Way? - The New Stack - A discussion with Kelsey Hightower, on how Nix might not take over from Docker but potentially could work symbiotically to create a more secure base layer. Nix is a package manager and ensuring package dependencies are met.
• kubectl debug | Kubernetes - A kubernetes 1.30 feature which can spin up a sidecar ephemeral pod, which can attach to another pod, e.g. a debugger service, for debugging purposes. I wasn’t aware a feature like this existed! So often I’ll shell connect to another pod just to see if I can curl it, this approach seems way better.
• Overview of Telemetry for Kubernetes Clusters - A list of popular tools for monitoring your kube clusters.

#JavaScript Ecosystem

• Bun v1.2.9 | Bun Blog - Bun 1.2.9 is released and has a built in redis client, built in Zig, it’s experimental and Zig is also not very stable. Other than that, there’s a lot of nice little features and performance improvements.
• npm Status - Issues publishing and installing packages - NPM accidentally wiped out everyone access tokens 🫢
• Fastify + React is 7x Faster than Next.js & Why is Next.js so slow?? - YouTube - A breakdown of what’s making React SSR feel slow and what can be done to optimize it.
• Node Weekly Issue 574: April 15, 2025 - Weekly javascript treat.