Are programmers losing ground to AI, or is human-centered design more important than ever? A cluster of posts lays out both sides: AI can let anyone write code, while others push for disciplined human–AI coding [1][2].
AI democracy in coding: AI can let non-developers generate code that looks functional, but big projects still demand architects, oversight, and quality gates. The idea rests on AI as a tool, not a replacement for professionals [1].
Discipline matters: human–AI coding: A call for a measured, rigorous approach to ensure accuracy and reliability in AI-assisted builds signals that architecture and governance aren’t going away [2].
LLMs speed up side projects: LLMs are the missing link to shipping side projects, turning ideas into runnable prototypes more quickly [3].
Tool-calling reality check: The practical world of tool calls is messy, with engineers tinkering models and settings to get reliable results.
• Qwen3-30b variants — reliability for tool calling is spotty; many users report frequent loops or failures, so results aren’t guaranteed [4]. • Roo — parsing tool-call responses and parameters matters; some teams push for unified tool-calling workflows [4]. • IBM Granite 4H-Small — described as very good with tool calls by practitioners [4].
Bottom line: automation accelerates delivery, but safe, scalable software still hinges on architecture, safeguards, and human judgment [1][2][3][4].
References
Ask HN: What Value Do Programmers Have If AI Allows Everyone to Write Code?
Discusses AI-enabled coding, programmer roles, quality and security, and human-AI collaboration maintaining value through problem solving and architecture design principles.
View sourceSerious Coding: The human–AI discipline for accurate and reliable development
Post discusses serious coding with AI, contrasting manual discipline vs vibe coding and how human–AI collaboration improves reliability
View sourceLLMs are the missing link to shipping side projects
Contends LLMs enable rapid shipping of side projects by pairing with tooling and automation, reducing traditional development barriers.
View sourceBest Model & Settings For Tool Calling
User questions tool-calling reliability of Qwen3-30b, praises IBM Granite 4H-Small for tool calls, and discusses prompts and settings in practice.
View source