Ján Jaroš

Devs, we come bearing only good news  🎉 Just launched: monday dev × Cursor integration! We'll let Andrew Milich, Head of Engineering at Cursor, explain why this changes everything: "With the monday dev Cursor extension, developers can manage their tasks directly in Cursor without switching tools. All the surrounding context, from code history to team activity, flows into the editor, making AI-generated code more accurate and useful. Developers stay in control, while teams move forward with less friction."  👉 THIS is the integration developers have been waiting for.

283

38 komentářů

In modern software development, we don't just guess if our code works. We write unit tests, run integration tests, and build CI/CD pipelines. We replaced manual guesswork with rigorous, automated validation. So why are many of us still in the "guesswork" phase with LLM prompts? The common workflow is a manual loop : tweak a prompt, test it, eyeball the result, and tweak it again. This is artisanal, slow, and doesn't scale. A prompt that works today might break tomorrow with a slight model update. It’s not an engineering discipline. The paradigm shift we need is Systematic Prompt Optimization. This is the move from "prompt art" to "prompt science." It’s about treating a prompt not as a magic incantation, but as a key component of a system that can be algorithmically tested, measured, and improved. The framework for this is surprisingly simple and powerful: 1./ Hypothesis (Your Base Prompt): Your initial, best-guess prompt. 2./ Ground Truth (An Evaluation Dataset): A set of inputs and ideal outputs that define success for your use case. 3./ Objective Function (An Evaluator): A measurable score for success (e.g., accuracy, semantic similarity, factuality). 4./ Optimizer: An algorithm that intelligently searches the vast space of possible prompt variations to find the one that maximizes your objective function. This approach is a repeatable, data-driven process. It allows you to prove why one prompt is better than another and ensures your system is robust. I've been exploring frameworks that enable this, and Comet's Opik is a fascinating, concrete example of this principle in action. It provides the optimizer and structure to automate this entire loop. Check here: https://lnkd.in/dZEfCW6S By adopting this mindset, we're not just writing better prompts. We're building more reliable, maintainable, and predictable AI systems. What steps is your team taking to bring more engineering discipline to your work with LLMs? #llm #ai #optimization #agents

194

93 komentářů

We’re excited to share that the official PHP MCP SDK is now in the works! 💫 The PHP Foundation has partnered with Anthropic’s MCP team and Symfony to build a framework-agnostic, production-ready SDK the entire PHP ecosystem can rely on. https://lnkd.in/ejN9K2s9 - Maintained by the Symfony team. - Joined by Kyrian Obikwelu, who brings his experience with MCP libraries. - Open for contributions from the community. ⭐️ Star the repository: https://lnkd.in/etEPQU5b

1 525

43 komentářů

There are many posts about Python 3.14, but for pi-themed versioning, there’s something better. The Knuth versioning system, a unique approach to version numbering introduced for his TeX and Metafont systems. It is designed to reflect Knuth's philosophy of approaching perfection in these software systems, emphasizing stability and minimal changes over time. The version numbers are designed to converge to a specific mathematical constant : For TeX, the version numbers approach π, for Metafont, the version numbers approach e. The design of both systems was frozen after version 3.0 for TeX and 2.0 for Metafont, with subsequent updates containing only bug fixes. Each new version adds another decimal place of precision to the current version number. For example, TeX's version history includes numbers like: 3.0, 3.1, 3.14, 3.141, 3.1415, etc. Similarly, Metafont versions approach e with updates such as: 2.0, 2.7, 2.71, 2.718, etc. As of 2021, the version numbers are 3.141592653 for TeX and 2.71828182 for Metafont. Knuth has stated that the "absolutely final change (to be made after my death)" will be to set the version numbers to π for TeX and e for Metafont, at which point all remaining bugs will be considered features.

83

12 komentářů

Java 25 comes with great features and enhancements, including making the language more succinct and beginner-friendly. This is a fully valid Java program. No imports. No method modifiers. No unneeded arguments. And you can run it directly via the "java" CLI. No extra compilation steps. #Java Java

483

17 komentářů

Need to reverse a Rust binary? Not that easy the first time. Many concepts and features of Rust result in assembly specifics. The Rust compiler does wonders, I'm always amazed by the intelligence of such tools... So, well, if you want to reverse a Rust binary, have a look at my new article: 📜 https://lnkd.in/eZ-eACci I also recommend those: - https://lnkd.in/e5t2cavv - https://lnkd.in/eFiyfWwK 👉 https://lnkd.in/e5UyRmNb #rust #reverse #assembly #radare2

105

2 komentáře

Inlining is a great #LabVIEW feature to improve code execution performance, but some side effects exist that can cause bugs in your apps. I learned it the hard way a couple of months ago on one of the projects I've been working on for years 😬 Because those side effects occur in very particular cases, it can be challenging to find the bugs they are causing. Inlining principle is to remove the overhead of calling the VI by removing it and placing its code directly in the calling VI during the compilation. It's smart, but if the inlined code relies on the Call Chain node, there are chances you may have a problem. In the picture below, I simplified the issue I had when I inlined one of the VI in my code. You can see that the output of the exact same VI changes just because one of them is inlined. If it can save you debugging time in the future, that will make my day. I can't think of other issues related to the inlining feature, but if you experienced other one, you are more than welcome to share your experience in the comments.

52

15 komentářů

Whoa - the past 24 hours were hot: 2000+ files refactored, and that doesn't include moves. The change? Basically reorganizing structure to simplify imports and thus improve performance. What could easily have taken weeks without assisted coding - was doable in a bit more than 1 calendar day thanks to AI doing busywork. But it was a great way to see how easily coding assistants get things wrong: - it tried to brute-force ALL paths to be identical - without realizing that I am using subfolders. - it didn't get that backend and API are different modules. - it used upperquotes in the middle of a filename - and forgot encapsulation! - it tried mocking the productive code when the tests failed. - it tried deleting tests when they failed. - it tried insisting that a few failing tests are OK. - it had trouble finding the interpreter, then tried to install it - despite it showing proof of work. About 50 times. - it continuously tried creating "simple test" files all over the place, then forgetting to clean them up, opening security loopholes. Yes - AI coding is a boon. If you have a good test suite, you can achieve a LOT in a very short time. And the constraints? 1. You need *good* tests. Not just "some tests." 2. You must know better than AI what works and what doesn't. 3. You must constantly pay attention, as AI can quickly do the wrong thing. 4. You need a reliable build infrastructure. 5. You need a reliable CI/CD pipeline. And it should be *fast.* If you have all that, AI can boost your productivity. Without any one of these 5 points, you're playing Russian Roulette with your codebase. #continuousdelivery #coding

51

29 komentářů

🎉 Announcing GitLab 18.4 with new features including GitLab Duo Model Selection (GA) and the introduction of the GitLab Duo AI Catalog! Key highlights: 🤖 GitLab Duo Model Selection is now GA! Organizations get greater control over AI models powering development workflows 📖 Introducing GitLab Duo AI Catalog - a central library where teams can create, save, and share custom-built agents across their organization. 🧠 GitLab Knowledge Graph provides rich code intelligence across codebases for better context and planning ⚙️ End user model selection lets developers choose preferred models for Agentic Chat 🔐 CI/CD job tokens can now authenticate Git push requests 🛡️ GitLab Duo Agent Platform context exclusion protects sensitive information ☁️ Expanded AWS region support for GitLab Dedicated Plus: Pipeline simulation against different branches, enhanced job artifact controls, OpenTofu modules support, and Duo Code Review GA on Self-Hosted! Thank you to our community for 136 contributions! 🧡 https://lnkd.in/gM2fHYW2

184

3 komentáře