coding

AI coding tools often invent non-existent software package names, posing risks to the software supply chain. This “hallucination” allows malicious actors to create fake packages, capitalizing on typos or inaccuracies from developers. Many developers rush installations without verifying these packages, leading to potential malware exposure. Experts advise double-checking package names before installation, and organizations should adopt better practices to mitigate this risk.

https://www.theregister.com/2025/04/12/ai_code_suggestions_sabotage_supply_chain/

Google Code Assist now features Gemini 2.5 in its free tier, improving its capability significantly compared to last year's version. The individual, standard, and enterprise variants cater to different user needs. New agent capabilities include generating software from specifications, migrating code, implementing features from GitHub, code reviews, generating tests, testing AI models, and creating documentation. Despite the advancements, challenges remain regarding the practical implementation and the reliability of features across different tiers. Overall, Gemini Code Assist has potential as a powerful programming tool.

https://www.zdnet.com/article/why-google-code-assist-may-finally-be-the-programming-power-tool-you-need/

Software exec Birgitta Böckeler highlights the need to balance AI coding skeptics and enthusiasts at QCon conference. She asserts AI coding, while evolving to improve productivity, presents challenges and risks, emphasizing responsible usage. Caution is urged against over-reliance on AI as it may exacerbate code quality issues. Teams should foster both experimentation and skepticism, valuing contributions from all members regardless of their stance on AI.

https://devclass.com/2025/04/09/encourage-the-ai-coding-skeptics-curb-the-enthusiasts-says-software-exec-at-dev-talk/

Stop using AI code editors due to loss of competence and over-reliance; manual engagement recommended. Started using AI in 2022, felt fast but grew dependent. Dropped AI in late 2024 to regain skills. Maintaining in-depth knowledge is crucial, as AI fails in complex contexts. Use AI wisely, but don't rely fully on it—stay engaged and curious to avoid becoming obsolete in programming. Enjoy coding for its own sake and keep practicing basics to remain capable.

https://lucianonooijen.com/blog/why-i-stopped-using-ai-code-editors/

Cognition has launched a low-cost pay-as-you-go plan for its viral AI coding tool, Devin, previously priced at $500/month. The new plan starts at $20 for about 9 computing credits, with usage potentially increasing costs. Although Devin 2.0 claims improved functionality, its effectiveness in complex coding remains questionable, with earlier evaluations showing it completed only 3 of 20 tasks successfully.

https://techcrunch.com/2025/04/03/devin-the-viral-coding-ai-agent-gets-a-new-pay-as-you-go-plan/

Zencoder introduces ‘Coffee Mode,' allowing developers to use AI for unit tests and coding tasks while taking breaks. The AI, integrated into popular IDEs like Visual Studio Code and JetBrains, claims superior performance on industry benchmarks by analyzing entire codebases. Its unique “Repo Grokking” technology offers contextual insights, mitigating common AI coding tool issues. Zencoder positions itself against competitors like GitHub Copilot while enabling developers to maintain familiar workflows. The company aims for production-ready AI code with security checks, projecting transformative changes in software development by 2025.

https://venturebeat.com/ai/zencoders-coffee-mode-is-the-future-of-coding-hit-a-button-and-let-ai-write-your-unit-tests/

Vibe Coding Overview:
Vibe Coding emphasizes the “vibe” or essence of software rather than focusing solely on the code itself. It leverages AI to create software by using natural language inputs to guide the development process, allowing AI to iteratively refine and enhance code behavior based on human conceptual directions. This shift raises new challenges for businesses, such as reliance on code that may not be fully understood by humans, the acceleration of prototype development, and implications for software testing, security, and team structure. As AI-generated code becomes more prevalent, it's vital for businesses to adapt and stay competitive in this evolving landscape.

https://www.forbes.com/sites/nishatalagala/2025/03/30/what-is-vibe-coding-and-why-should-you-care/

Breaking up with vibe coding: once enjoyable, now inefficient. Over-reliance on AI leads to time sink, high costs, lack of understanding. AI aids mockups, but structured planning needed for complex tasks. Exploring balanced alternatives like Gemini Code Assist and Open WebUI for better control and reduced expenses. Creativity valuable, but unsustainable without structure, especially under deadlines.

https://www.lucasaguiar.xyz/posts/vibe-coding-pitfalls/

Vibe coding uses AI (like ChatGPT) to generate complete software code for non-programmers. Coined by Andrej Karpathy, it enables users to create apps without programming knowledge, but results can be buggy and unreliable. While some claim it may replace developers, experts assert it won't, as software engineering involves more than just writing code.

https://www.newscientist.com/article/2473993-what-is-vibe-coding-should-you-be-doing-it-and-does-it-matter/