Jiquan Ngiam’s Post

#AI is transforming #programming by boosting #code efficiency, accuracy, and security. AI #Tools provide real-time code suggestions, helping #developers work faster and reduce errors. AI-driven #debugging tools can detect bugs, inefficiencies, and security flaws, while automated #testing frameworks identify edge cases, making code more robust. AI also aids in #coderefactoring, suggesting structural improvements for better readability and performance. Additionally, AI-powered code #searchengines streamline finding relevant snippets, saving time. Overall, AI optimizes workflows, minimizes repetitive tasks, and enhances the quality of code, becoming an essential asset for modern developers.

Why You Shouldn't Blindly Trust AI for Coding In our tech-driven world, AI has made significant strides in many areas, including code generation. However, it's crucial to understand that relying on AI for coding can lead to mediocre and sometimes problematic results. Here's why: Lack of Context: AI often misses the nuances and specific requirements of your project, resulting in generic or suboptimal code that doesn't fully address your needs. Limited Creativity: AI lacks the creativity and problem-solving abilities of a skilled human developer. It can’t innovate or think outside the box to find the most efficient or elegant solutions. Quality Control Issues: AI-generated code might pass basic syntax checks but often fails to meet higher standards of code quality, security, and maintainability. Incomplete Testing: While AI can assist in some aspects of testing, it doesn’t replace the comprehensive and critical testing done by experienced developers. Human Touch is Irreplaceable: The expertise, intuition, and collaborative skills of human developers are essential for producing high-quality, scalable, and secure code. AI is a powerful tool, but it should complement, not replace, human expertise. Let's leverage AI to automate repetitive tasks and assist our coding efforts while relying on skilled developers to drive innovation and ensure excellence. #AI #Coding #SoftwareDevelopment #QualityControl #Tech

AI is not just a tool but a transformative force that can reshape the developer role, empowering teams to harness their creativity and drive progress in software development. Embracing AI thoughtfully positions organizations to thrive in an increasingly competitive landscape. #AI #Softwaredevelopment

GenAI for code generation can be quite useful as long as the team understands the pitfalls and antipatterns it brings along! To mention a few: > Lack of team collaboration (Paired programming offers good collaboration and ability to build software together), lack of software equity - ownership of the codebase and understanding the complex flows and paths specific set of programs can take, business understanding - have you or (GenAI) built the right thing?, GenAI can amplify the good, bad or even the ugly - coding assistants are as good as the data or the code patterns we feed them, they will repeat and reproduce the good, bad, ugly to the superlative and can make things super complex!, Last not least, ALL generated code must be verified and reviewed - there is no alternative to the good-old code reviews and due diligence. Bottom-line when using GenAI, human intelligence must be applied - its the responsible thing to do! I would like to rephrase the ever-so-famous AI adage with: "GenAI doesn't warrant skipping human diligence"

The ignorant scream that developers will soon be replaced by AI. But that's because they don't understand development. A good developer codes only 20% of the time. The rest is spent on architecture, meetings, thinking, infrastucture and figuring out how to turn the problem into code. Development is not about typing on the keyboard. It's about figuring out what the customer needs and turning that into code. AI is not a replacer, it's an enabler. Sure, in the future there will be less need for engineers, but that's because one engineer will be much more efficient with the help of AI. You will get more done faster. I am a heavy user of various AI tools and I've never felt so productive in years! What used to take days literally takes hours today. So my tip is to lean into the whole AI trend, master it and then raise your hourly rate. Because AI will soon make you 10X more efficient

Unit tests will soon become obsolete AI Models Will Suggest Specs and Verify Implementations Advanced AI models will soon start generating specifications for our code and verifying that the implementations meet these specs. Instead of writing unit tests ourselves, the AI will understand what we want to achieve and ensure the code does exactly that. Specification Languages Are Evolving Specification languages are stepping up to capture concepts that were hard to define before. They'll provide a clearer way to express what our code should do, making it easier for both humans and AI to work together seamlessly. Compilers Will Verify Through Specs Compilers won't just translate code anymore—they'll also verify it against the specifications. This means they'll catch issues early on, ensuring that the code aligns perfectly with what's needed. External APIs Will Provide AI-Friendly Specs (not just for humans) APIs from external services will start offering specifications designed for AI understanding. This will make integrating with other systems smoother, as the AI will know exactly how to implement and interact with them. --- In a nutshell, the tedious task of writing unit tests is set to be replaced by smarter AI-driven processes. We'll be able to focus more on creativity and building great software, while AI handles the verification and testing behind the scenes.

🚀 Are We Ready to Fully Integrate AI into Our Development Environments? 🤖 AI is reshaping how we code, but we’re now facing a critical question: Should we give AI full access to our development workflow? Tools like Codium are pushing boundaries with full contextual awareness, enabling AI to: ✅ Analyze entire codebases for smarter suggestions and refactoring. ✅ Detect and debug issues across multiple files in real time. ✅ Run commands intelligently, streamlining repetitive tasks. ✅ Pick up where you left off, understanding your workflow like never before. This is a game-changer for productivity, but it also raises important concerns: ⚠️ Security risks – How do we ensure sensitive code stays protected? ⚠️ Over-reliance – Will developers lose critical debugging skills? ⚠️ Ethical considerations – Can AI-driven coding remain aligned with best practices? The potential is undeniable, but so are the risks. 💡 👉 Are we ready to trust AI with full access to our codebases? Or should we set limits before taking the leap?

I want to clear something up about AI and the idea of it replacing developers. Can AI replace junior to mid-level developers by 2025? Maybe—but I’m still skeptical. Why? Because while AI can write code, that doesn’t mean it can build applications. Writing code is just one part of engineering. Understanding edge cases, systems, and how to connect everything to make it work seamlessly is a whole different skill set—and AI co-pilots are still far behind in this area. Code generation is one thing, but building reliable, production-ready applications is a completely different challenge. AI will likely struggle with that for a while.

Yes please! ❤️ “Even at a human-level there is so many problems in software delivery - I'm not sure how there is an assumption that AI will come in and magically solve everything.” This is so true, and definitely a significant part of the problem. If you shift further left, you start to notice humans and soft skills in need of help. I’m finding communication, individual needs not being met, business and engineering team misalignments, and much more are suffering. Even before you get to the work that needs to be done, humans are being left behind. Processes of any kind are not going to fix this. Quality and QA rely on all of this even before automation or AI come into the picture. Quality in the final product starts with the people building it.

I find it fascinating the number of companies chasing to build AI coding agents. But also generally the hype around it. Even at a human-level there is so many problems in software delivery - I'm not sure how there is an assumption that AI will come in and magically solve everything. This reminds me of the same "silver bullet" mindset we've seen in the industry with Automation. The fact is that there is a substantial problem with requirements in software development. Developers spend a huge amount of time in clarification meetings and making assumptions just to get to a point where they can start to code. If we want to do anything intelligent with AI or beyond - we have to start by sorting the input into software delivery - REQUIREMENTS. I believe visual modelling is the solution to this - but I'm also open that there are other approaches - the bottom line is the current way requirements are formulated and specified is really not good enough. There will be "interesting" results from these AI coding companies - but they will soon hit a ceiling of what can be achieved in the real-world with garbage requirements being the input. Rant over. #QualityEngineering #RequirementsEngineering #AICoding #DeveloperProductivity