Jarl Friis

Anthropic has taken many of the core concepts of Claude Flow and integrated them directly into its new Claude Agent SDK. In my Flow platform, I had to build orchestration around features like artifact tracking, retries, and context persistence because the SDK didn’t provide them. Now those same ideas are native. Tool governance, artifact durability, and context checkpoints are primitives in the SDK. For example, in Claude Flow I wrote custom routines to retry failed tool calls with exponential backoff. The SDK now includes built-in retry handling. I built artifact management so outputs could be stored, versioned, and reused. The SDK now provides artifact durability out of the box. I had to wire in checkpoints to let an agent recover from mid-flow errors. The SDK now exposes context checkpoints directly. This validates the Flow approach. Structured cycles weren’t a workaround, they were the correct abstraction. Developers can now spin up long-running agents without bolting on any orchestration. The friction is lower, the foundation is stronger. At the same time, Anthropic is absorbing practices we pioneered back into its stack. Not that they have given any acknowledgement. Subagents? That was literally our term. The good news is this makes agents easier for newcomers to build, but it also sets tighter boundaries on how Claude agents are expected to work.

182

33 Kommentarer

"Shift left” sounds good on paper. Faster feedback. Earlier testing. Cost savings. But here’s the uncomfortable truth: shifting left won’t save you if your culture stays stuck. You can throw tests into CI pipelines, automate early, and chant DevTestOps until the walls echo. But if your teams still treat testing as a phase, quality as someone else’s job, or feedback as a nuisance, all you’ve done is moved broken thinking upstream. I’ve seen companies install shiny new tooling with zero conversation about shared ownership. I’ve seen developers resist writing tests because “that’s not what we do.” I’ve seen testers struggle to add value early because they’re still expected to “verify the build” instead of shape the build. Shifting left without shifting culture is like moving the finish line but refusing to train differently. You don’t get better outcomes. You just get earlier dysfunction. If you’re serious about building quality in, here’s what needs to shift first: - The belief that quality is everyone’s job, not just the testers’. - The space for testers to be part of discovery, not just delivery. -The trust that cross-functional collaboration is not optional, it’s foundational. True shift left starts with mindset, not tooling. With conversations, not compliance. With people, not processes. If your culture still rewards speed over safety, delivery over discovery, and blame over learning, no amount of leftward motion will get you where you want to go. I help companies get this right not by selling buzzwords, but by rebuilding how teams think, act, and deliver together. If that’s the kind of change you’re ready for, I’d love to talk. #softwaretesting #softwareengineering #qualityleadership #teststrategy #brijeshsays

173

27 Kommentarer

The true measure of a software developer is being able to deliver value in the simplest possible way. It isn't knowing the latest languages, frameworks, or techniques; it's knowing when not to use them. It isn't delivering lots of sparkly new features; it's delivering the one feature that gives its users leverage they need. It isn't sitting behind a screen 16 hours a day; it's knowing who to talk to and building relationships. It isn't being right; it's being an inspiration.

221

10 Kommentarer

You'll find this code smell in many production codebases. I see this far too often: magic numbers and strings. The problem with magic numbers (or strings) is that they carry no meaning. The code is harder to reason about and more error-prone. My preferred solution is introducing a config value or enum. This is just one thing you can do to make your code cleaner. It's clean code Saturday, so here are 5 C# refactoring tips: https://lnkd.in/eZQY4CgH Do you dedicate time each week to refactoring your code? --- Sign up for the .NET Weekly with 74K+ other engineers, and get a free Clean Architecture template: https://lnkd.in/eJSR2bmm

419

19 Kommentarer

Flaky tests feel random. They’re not. The patterns are clear: * Timing issues (async workflows, race conditions) * Shared state (bad test isolation, order dependencies) * Environment drift (things that work locally but fail in CI) By definition, though, they’re not obvious to spot. A missing await, an unseeded random number, a fragile assertion. These problems tend to show up after the developer who wrote them has moved on to the next thing, reducing the likelihood that anyone will take time to fix them. AI is going to multiply these issues. It’s *also* going to help us catch them faster. The challenge for leaders: don’t accept flaky tests as inevitable. They’re just tech debt we’ve been too slow to pay down.

43

I’ve been thinking a lot about where multiplatform development is heading. With tools like Kotlin Multiplatform becoming truly practical, the expectation to share more logic across platforms is growing fast. Lately I started exploring Rust as a multiplatform option. Rust is usually aimed at C++ developers and systems programmers, but I keep coming back to this question: What if Rust were approachable for mobile developers too? One of the big reasons Rust interests me is its mix of performance and safety. You get native speed without garbage collection or ARC, strong guarantees around memory safety, and the ability to ship the same core module to iOS, Android, desktop, backend, or even WebAssembly without rewriting it per platform. That makes it another compelling option for shared business logic or core components across platforms. Some apps already use Rust as a strong core layer. I'm exploring to see how viable it is for mobile devs to try out, even if it's for a small module rather than replacing a full-blown business layer across multiple apps. I want to see how we can make it less intimidating and more accessible for everyday app developers. I’m exploring this and will share more as I go. If there is enough interest, I’ll go deeper. I'll share my findings on https://lnkd.in/emVy_hFv

95

15 Kommentarer

Tip: If you need to upgrade your code to a newer version of a package, point GitHub Copilot at the changelog and ask it to make the upgrade. You may also want to note "don't worry about supporting older versions of the package, just use the latest version" as some models like to add excessive exception handling and fallbacks. I'm using this approach to upgrade my Pydantic-AI samples today, and it's working pretty well, since there's a bunch of little fixes across multiple projects.

30

Waiting half an hour for your tests to finish? That’s not feedback - that’s punishment 🧪 You might have been there: Every new integration test adds another minute to your overall build time. Twenty minutes turn to thirty. CI times explode. Developers stop writing integration tests and become unhappy. Flaky tests appear. Bugs slip through. Releases slow down. But what if you could slash that time, without deleting a single test? With Spring Test, that's possible. It's a hidden gem called Spring TestContext Context Caching 🏆 So, how do you know how many contexts are spinning up? More importantly, how do you figure out why they aren’t being reused? That’s why I built Spring Test Profiler ⏱️ Meet Spring Test Profiler - the free, open-source x-ray for your test suite. Integration is as simple as this three-step process: 1. Add the dependency (Maven/Gradle snippets appear on screen). 2. Run your tests. 3. Open the report at: target/spring-test-profiler/report.html. Within the report, you’ll see: • How many contexts started, and why. • Which tests are sharing configs versus starting their own. • Exact diffs: profiles, properties, or annotations blocking cache reuse. The overall goal is to help identify cache misses and align your tests to reuse contexts for faster feedback 🚤 Spring Test Profiler: Profile. Optimize. Deliver. Try it yourself: https://lnkd.in/djXVJiNE Below is a demo of a section in the report to analyze and compare two contexts 🆚

49

5 Kommentarer

🔑 𝗠𝗮𝘀𝘁𝗲𝗿𝗶𝗻𝗴 𝗦𝗢𝗟𝗜𝗗 𝗣𝗿𝗶𝗻𝗰𝗶𝗽𝗹𝗲𝘀 𝗶𝗻 𝗦𝗼𝗳𝘁𝘄𝗮𝗿𝗲 𝗗𝗲𝘀𝗶𝗴𝗻 Writing clean, maintainable, and scalable code isn’t just about syntaxit’s about design. That’s where SOLID Principles come in. These five timeless guidelines help developers build robust applications that are easier to extend, test, and maintain. Let’s break them down with practical detail: 1️⃣ 𝗦𝗶𝗻𝗴𝗹𝗲 𝗥𝗲𝘀𝗽𝗼𝗻𝘀𝗶𝗯𝗶𝗹𝗶𝘁𝘆 𝗣𝗿𝗶𝗻𝗰𝗶𝗽𝗹𝗲 (𝗦𝗥𝗣) 👉 𝗗𝗲𝗳𝗶𝗻𝗶𝘁𝗶𝗼𝗻: A class should have only one responsibility or reason to change. 📌 𝗪𝗵𝘆 𝗶𝘁 𝗺𝗮𝘁𝘁𝗲𝗿𝘀: - Classes that do too much quickly become fragile and hard to - maintain. Changing one responsibility might accidentally break another. 🔑 𝗕𝗲𝗻𝗲𝗳𝗶𝘁: Focused classes are easier to understand, test, and reuse. 2️⃣ 𝗢𝗽𝗲𝗻-𝗖𝗹𝗼𝘀𝗲𝗱 𝗣𝗿𝗶𝗻𝗰𝗶𝗽𝗹𝗲 (𝗢𝗖𝗣) 👉 𝗗𝗲𝗳𝗶𝗻𝗶𝘁𝗶𝗼𝗻: Classes should be open for extension, closed for modification. 📌 𝗪𝗵𝘆 𝗶𝘁 𝗺𝗮𝘁𝘁𝗲𝗿𝘀: - Keeps existing code stable while allowing new features to be added. - Prevents bugs from creeping in when requirements change. 🔑 𝗕𝗲𝗻𝗲𝗳𝗶𝘁: Future-proof code that supports growth without breaking existing features. 3️⃣ 𝗟𝗶𝘀𝗸𝗼𝘃 𝗦𝘂𝗯𝘀𝘁𝗶𝘁𝘂𝘁𝗶𝗼𝗻 𝗣𝗿𝗶𝗻𝗰𝗶𝗽𝗹𝗲 (𝗟𝗦𝗣) 👉 𝗗𝗲𝗳𝗶𝗻𝗶𝘁𝗶𝗼𝗻: Subclasses should replace their base classes without breaking the application. 📌 𝗪𝗵𝘆 𝗶𝘁 𝗺𝗮𝘁𝘁𝗲𝗿𝘀: Violations lead to unexpected behavior and broken inheritance chains. Ensures your abstractions actually behave as promised. 🔑 𝗕𝗲𝗻𝗲𝗳𝗶𝘁: Reliable and consistent code with true polymorphism. 4️⃣ 𝗜𝗻𝘁𝗲𝗿𝗳𝗮𝗰𝗲 𝗦𝗲𝗴𝗿𝗲𝗴𝗮𝘁𝗶𝗼𝗻 𝗣𝗿𝗶𝗻𝗰𝗶𝗽𝗹𝗲 (𝗜𝗦𝗣) 👉 𝗗𝗲𝗳𝗶𝗻𝗶𝘁𝗶𝗼𝗻: No client should be forced to depend on methods it does not use. 📌 𝗪𝗵𝘆 𝗶𝘁 𝗺𝗮𝘁𝘁𝗲𝗿𝘀: - Large interfaces become “fat” and force implementing classes to provide dummy methods. - Leads to unnecessary coupling. 🔑 𝗕𝗲𝗻𝗲𝗳𝗶𝘁: Flexible contracts that encourage clean, modular design. 5️⃣ 𝗗𝗲𝗽𝗲𝗻𝗱𝗲𝗻𝗰𝘆 𝗜𝗻𝘃𝗲𝗿𝘀𝗶𝗼𝗻 𝗣𝗿𝗶𝗻𝗰𝗶𝗽𝗹𝗲 (𝗗𝗜𝗣) 👉 𝗗𝗲𝗳𝗶𝗻𝗶𝘁𝗶𝗼𝗻: Depend on abstractions, not concrete implementations. 📌 𝗪𝗵𝘆 𝗶𝘁 𝗺𝗮𝘁𝘁𝗲𝗿𝘀: - High-level modules shouldn’t be tied to low-level implementation details. - Makes code easier to test, mock, and extend. 🔑 𝗕𝗲𝗻𝗲𝗳𝗶𝘁: Enhances testability, reduces coupling, and increases flexibility. 💡 Next time you write a class, ask yourself: Am I following SOLID? 👉 Follow me Kanaiya Katarmal, and hit the 🔔 on my profile so you don’t miss upcoming .NET tips and deep dives. 𝗧𝗵𝗮𝗻𝗸𝘀 𝗳𝗼𝗿 𝗿𝗲𝗮𝗱𝗶𝗻𝗴 𝗹𝗲𝘁’𝘀 𝗰𝗼𝗻𝘁𝗶𝗻𝘂𝗲 𝗴𝗿𝗼𝘄𝗶𝗻𝗴 𝗮𝗻𝗱 𝗹𝗲𝗮𝗿𝗻𝗶𝗻𝗴 .𝗡𝗘𝗧 𝘁𝗼𝗴𝗲𝘁𝗵𝗲𝗿! 👉 𝗦𝘂𝗯𝘀𝗰𝗿𝗶𝗯𝗲 𝘁𝗼 𝗺𝘆 𝗻𝗲𝘄𝘀𝗹𝗲𝘁𝘁𝗲𝗿: https://lnkd.in/dp7tz_3V #SoftwareEngineering #SOLIDPrinciples #CleanCode #DesignPatterns #Programming #CodeQuality #Developers #SoftwareDesign #OOP

399

29 Kommentarer