Jonas Bonér

New episode: Mattias Karlsson on Cake Build—escaping YAML chaos for C# clarity 🎯 This conversation explored Cake (C# Make), a build orchestration framework offering an alternative to traditional YAML-based build pipelines. The key insight: using a strongly-typed language like C# for build definition provides greater clarity, debuggability, and consistency. 𝗪𝗵𝗮𝘁 𝗖𝗮𝗸𝗲 𝗮𝗰𝘁𝘂𝗮𝗹𝗹𝘆 𝗱𝗼𝗲𝘀: It doesn't replace MSBuild, the .NET CLI, or your CI/CD system. It orchestrates them, simplifying complex workflows and offering a more developer-friendly experience. You write workflows in C# instead of YAML, PowerShell, or Bash. Cross-platform, works identically on Mac, Windows, Linux, or in containers 💭 𝗧𝗵𝗲 𝗱𝗲𝗯𝘂𝗴𝗴𝗶𝗻𝗴 𝗮𝗱𝘃𝗮𝗻𝘁𝗮𝗴𝗲 𝘁𝗵𝗮𝘁 𝗰𝗵𝗮𝗻𝗴𝗲𝘀 𝗲𝘃𝗲𝗿𝘆𝘁𝗵𝗶𝗻𝗴: Unlike YAML pipelines requiring push-to-runner cycles to identify issues, Cake allows loading scripts into Visual Studio or Code, setting breakpoints, and stepping through the build process locally. This dramatically speeds up development and troubleshooting. Test everything locally before pushing. 𝗣𝗹𝗮𝘁𝗳𝗼𝗿𝗺 𝗮𝗯𝘀𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻𝘀 𝗵𝗮𝗻𝗱𝗹𝗲 𝗰𝗼𝗺𝗽𝗹𝗲𝘅𝗶𝘁𝘆: Cake handles file path differences between Windows and Linux. Forward slash versus backslash? Handled. Environment variable separators? Handled. Path construction that works everywhere? Handled. You write once, run anywhere. The heavy lifting for cross-platform compatibility is done for you. 𝗦𝘁𝗿𝗼𝗻𝗴𝗹𝘆-𝘁𝘆𝗽𝗲𝗱 𝗺𝗲𝗮𝗻𝘀 𝗰𝗼𝗺𝗽𝗶𝗹𝗲𝗿-𝘃𝗲𝗿𝗶𝗳𝗶𝗲𝗱: Fully typed classes for tool arguments. You don't need to remember if parameters are integers or strings. You don't need to memorise which arguments can be combined. If it compiles, it runs. The compiler catches errors before you push. 𝗧𝗵𝗲 𝗖𝗮𝗸𝗲 𝗦𝗗𝗞 𝗮𝗱𝘃𝗮𝗻𝘁𝗮𝗴𝗲: Introduced in November 2025, Cake SDK enables single-file executables, smooth .NET ecosystem integration, and leverages .NET 10's file-based applications. Full Visual Studio Code support out of the box. Works with C# Dev Kit. Publish as containers, .NET tools, or platform-specific executables. Mattias on the learning curve: "Start with something really simple and then evolve. You don't need to replace everything at once." Available now: https://lnkd.in/ek45HzT3 #DotNet #DevOps #BuildAutomation

2

Developer Insights: Operating Media over QUIC in real-world environments In this video, Per Mafrost and Mattias Bergström talk about what it means to work with Media over QUIC beyond theory and specifications. The discussion is grounded in hands-on experience and focuses on how protocol choices behave once they meet real traffic patterns, real networks and real production requirements. They touch on the importance of operational predictability, how transport decisions affect the full live pipeline, and why consistency over time matters just as much as headline latency numbers. The conversation reflects lessons learned from building and running low latency streaming at scale, not from lab setups. For us, these insights are central to how we continue to evolve Vindral Live as a platform built for long-term, real-world use. #VindralLive #MoQ #LowLatency #LiveStreaming #VideoTechnology #Engineering #Vindral

12

“The global AI labs invest hundreds of billions in frontier models that already speak excellent Swedish. Spending tens of millions on a national language model is like building a Swedish search engine in 2005.” Great article on Sweden's new AI strategy, freely readable here: https://lnkd.in/dA23hc9w I largely agree, Sweden shouldn’t try to outspend frontier labs on foundation models. But we should also be cautious about becoming too dependent on a handful of large companies and the ecosystems around them. We’ve already seen in other sectors what over-reliance can lead to. Sweden has historically thrived through openness and collaboration. Rather than building isolated national models, we could focus on actively contributing to and shaping the open-weight ecosystem, which is improving rapidly. Local AI, running closer to the user, on European infrastructure, with open foundations, may well be where long-term resilience and competitiveness meet.

3

Empower your team to design, build, and manage Kafka solutions with confidence🔥 The Irori team has years of hands-on experience operating Kafka in production for major Swedish enterprises. We've consolidated our collective expertise into a six-part course, combining theory with practical labs to equip you with the knowledge and tools to capture Kafka’s full capabilities. Read more below of what the Irori Kafka Academy entails! #ApacheKafka #Kafka #Learning

13

Working on DevEx these days does of course also mean enabling people to use AI. After Gustav Söderström mentioned Honk externally the other day, we have gotten a ton of questions - what is it and can I use it, can I also 🪿honk🪿? To no surprise Honk is an internal name for our tool (and a verb as well by now) where we combine fleetshift with AI coding agents. The goal has always been to be able to run complex code changes across the software fleet, to be able to complete these drawn out migrations we never were able to invest the time in to complete. Putting an AI coding agent into our fleet management system makes it possible to go one component at a time, let the agent make a code change and post that PR for review and deploy. A fun thing happened though - when we made it super easy to run background coding tasks we also started to see a shift in how we do things and what we do. In addition we see many small things being fixed that earlier wasn't always adressed, so not only are we completing these long running migrations, we also fix the small things to improve our code. And of course, also building features in the background, like coming up with an idea at the subway and having it implemented and done when arriving in the office, ready to be picked up and tried out on the phone. Max Charas and I put together this little webinar - I hope it answers the question what the 🪿 it is! https://lnkd.in/dSdkTwuk

95

1 Comment

608 ms to format 48k files. Treefmt was adopted by Nixpkgs! When we started writing treefmt 3 years ago, our benchmark was always how fast it would run on nixpkgs. And now it’s used nixpkgs to keep all the nix code formatted. Treefmt is open source and can be used in your own monorepo. Thanks @brianmacgee @jflry, @infinisil, …. for helping make this happen. To read more, check the project here: https://lnkd.in/erNiajRH

122

8 Comments

I’ll be in Malmö on February 25 for the first edition of MalmöJS, a new meetup series focused on something I care deeply about: Running JavaScript and Node.js at scale, in systems that cannot afford to fail. When you push Node.js hard enough, you quickly move past frameworks and APIs and start dealing with fundamentals: event loop behaviour, backpressure, memory allocation, CPU limits, and observability under load. At MalmöJS, we’ll talk about: ✅ What performance actually means in production ✅ How to design systems that degrade and recover safely ✅ Where real bottlenecks show up, and how to reason about them This is a great opportunity to connect with engineers in the region working on serious, high-performance systems, and to exchange practical lessons learned the hard way. If you’re working with Node.js beyond CRUD apps, I’d love to see you there. 👉 Register here: https://luma.com/q52blone

63

1 Comment

This morning I joined the GVIP, Global Vulnerability Intelligence Platform, community call. It is a new initiative aiming to improve vulnerability management and intelligence. In simpler terms: missing data. The discussion was lively led by Olle E Johansson. I highlighted two recurring problems we face in embedded systems: - patched software. Once you patch upstream code locally (or apply patches from someone else), current naming schemes and vulnerability databases fail to describe what you are actually running - package configuration. Whether your product is affected by a vulnerability often depends on build time (or run time) configuration options. That information is rarely captured in a usable way Another topic that generated strong discussion was KEV, Known Exploited Vulnerabilities, catalogues and how they should be integrated into workflows. I am looking forward to more discussions, and even more to concrete work starting on these subjects. More embedded voices in this space would be very welcome! Project website: https://lnkd.in/eyx8CWw9

32

6 Comments

Engineering Leader’s Guide: How to Become a Great Coach and Mentor Being a great coach and mentor is crucial. Learn how to become one the RIGHT way! An article by Gregor Ojstersek && Andy Skipper https://lnkd.in/g8rahQ4T

6

1 Comment

An agentically coded branch-and-cut solver for the Capacitated Profitable Tour Problem: https://lnkd.in/eDKdxGFP Based on old work of Mads Jepsen, Bjørn Petersen, David Pisinger and me (Jepsen et al., "A branch-and-cut algorithm for the capacitated profitable tour problem", Discrete Optimization, 2014), implemented fresh by coding agents from the paper. The solver runs on top of HiGHS, which required a few patches to support cut callbacks, propagation, and hyperplane branching. Great work by Julian Hall and HiGHS team for making it possible to patch into. Beyond the cuts from the paper, we explored propagation and reduced cost fixing guided by labelling lower bounds, hyperplane branching in several forms, an iterated local search primal heuristic, and shortest path inequalities using all-pairs shortest path bounds to derive covers. A lot of it didn't work, but it was surprisingly easy to test out ideas, from theory to implementation. We did a small comparison against PathWyse (Matteo Salani et al., "PathWyse: a flexible, open-source library for the resource constrained shortest path problem", Optimization Methods and Software, 2024). The labelling algorithm wins in most cases, but branch-and-cut has its merits: we solved 5 instances PathWyse couldn't. Overall, pretty impressed with how cleanly HiGHS got patched and how fast the agentic workflow let us move through algorithmic experimentation. #operationsresearch #mathematicaloptimization #branchandcut #agenticengineering #resourceconstrainedshortestpath

113

16 Comments

A sneak peek at what we're building. How do you visualize a digital society of autonomous agents? How do you manage a global, decentralized network? You build a command center. We call it the Federation Embassy. This is a look at our new interface for The Protocol's network, running live on localhost. Every point of light on this globe represents a sovereign registry. The lines between them are the cryptographically-secured trust bridges that make our federation work. This isn't just a UI. It's the nervous system for a new agent economy. It’s where operators will monitor network health, manage connections, and view real-time data flow across the globe. We're in the final stages of polishing the Developer Portal for our upcoming beta release. The foundation is built. The tools are being sharpened. The next era of agent-to-agent collaboration is about to begin. #AutonomousAgents #A2A #Infrastructure

4

2 Comments

Ready to learn some practical DSP? If so, we invite you to join us for tomorrow's The Audio Programmer Meetup! We will be joined by Musik Hack Co-Founder Sam Fischmann. Sam has had great success with well known plugins such as Master Plan and Fuel. We will be talking about useful DSP. Including... ✅ Filter basics, frequency and phase, not just frequency! ✅ Filters, when to use (IIR/FIR/linear/minimum phase/biquad forms/svf) ✅ Filter shapes, when to use (bessel, butterworth, chebychev 1+2 , elliptical/cauer/Linkwitz Riley/clamping) ✅ Filter implications for null tests and parallel processing ✅ Oversampling + aliasing ✅ Desmos for testing and comparing saturation techniques ✅ Saturation myths: Harmonic and inharmonic distortion ✅ Basic compressor/limiter design This will be a great conversation for anyone who is seeking to gain an overall understanding of how DSP algorithms work, and how to use them. Tune in here: https://lnkd.in/e9wxeA9h

57

8 Comments

Software Factory Last week, Google introduced the Gemini 3 Deep Think model and released its benchmark results late Thursday evening Finnish time. It is worth noting that the model is still operating in research mode and builds on previously developed architectures. Even so, it delivered exceptional performance across multiple benchmarks while using significantly less energy. New announcements are coming on a weekly basis. In many ways, this is a natural progression as the world’s leading technology companies compete to build more efficient and capable models. However, the results of Gemini 3 Deep Think prompted a more practical question: what happens when models like this move into full production use? How will they reshape our organizations and the way we operate? This line of thinking led me to reflect on a concept that has gained traction recently: the “software factory.” Within our team, we maintain an active dialogue around AI development. People are continuously experimenting with new tools, and I see it as essential that we leverage the experience of our current team when exploring how to apply emerging technologies. At the same time, we must remain open to rethinking processes and established ways of working. For a software company, the accelerating pace of tooling innovation challenges traditional development processes. They need to be redesigned. One clear observation is that seamless integration between tools is critical. At the same time, we must ensure that advancements in models consistently accelerate our work rather than slow it down. This leads toward an operating model where key areas of software development are supported by agents that continuously leverage improving models. When executed well, development becomes not only faster but also higher in quality. What does this mean for recruitment and talent? The focus shifts even more toward experienced full-stack architects and product builders. Modern tools can quickly elevate the productivity of junior developers, but the ability to critically assess where tools or agents fall short requires deeper experience. When the bar is set high and the team consists of top-tier professionals, it creates a virtuous cycle that attracts more exceptional talent. At the same time, the importance of people at the customer interface only increases. We must deeply understand our customers’ processes, operational challenges, and real-world constraints. This understanding enables us to define stronger product specifications, and with AI-assisted workflows, execute against them faster than ever. The pace is already intense. Companies are well advised to focus on their core operations and increasingly rely on SaaS-based solutions to run mission-critical functions. Based on what I see that is the most effective way to ensure continuous access to the industry’s evolving best practices.

85

9 Comments

💡  What's great about working with Unity from a Software Developer perspective? 💡 We asked Kristian Frank Erikson for his 2 cents on the subject, and here's what he had to say: "Unity's close working relationship with Oculus / Meta has been a great help for vrClinicals. When developing a simulation like vrClinicals, we've been dependent on being on the bleeding edge of VR development. This has meant enabling features available in Unity like Application SpaceWarp along with other advanced rendering techniques to deliver a smooth experience for our learners. Unity's support for C# and its large user community make it easy to create prototypes and for new developers to learn." Kristian adds: "Unity's Asset Store has been a great help to us, saving time in instances where functionality could simply be purchased and easily integrated." #Unity #SoftwareDevelopment #VR #Oculus #Meta #CSharp #AssetStore #Innovation

21

Anders Hejlsberg - creator of Typescript (and more) says AI started out being the assistant. But now it’s switching around. The AI is doing the work and it doesn’t need an IDE in the same way that we need IDEs. https://lnkd.in/etFxzzde

7

Thrilled to see the work of Georgian-Vlad Saioc on Go routine leak profiling being upstreamed to the Go compiler. Available as experiment from 1.26 which lands in February. Vlad is part of the Programming Systems Group at Uber, working out of our Aarhus office, where he has collaborated closely with Department of Computer Science, Aarhus University and our colleagues in both Amsterdam and San Francisco. Impressive and exciting example of how the team applies cutting edge research to improve Uber's reliability at massive (!) scale. Milind Chabbi Anders Møller https://lnkd.in/etim8PsU https://lnkd.in/eg-2ZJqz

153

2 Comments