As a maxi cursor and replit user, i can share my opinion: - 6500 prompts to generate ~80 modules (python, js) - ~28000 lines of vibe coding And i respectfully think you got this all wrong. AI is good at syntax, but... that's it. Definitely, dont spend much time on syntax issue as AI capacity to resolve the issue is ~100%. BUT (and it is a massive but): AI sucks at: - Writing complex efficient readable and maintainable codes across a whole codebase. - Thinking in systems, - Choosing wisely (assessing the tradeoffs of x and y) => You literally have no idea about the amount of redundancy ai coding creates... like it is just insane. AI agents are to engineers what calculators were to engineers, too: - Calculators didn’t replace engineers, they removed the burden of manual computation, freeing engineers to focus on design, modeling, and problem-solving. - AI agents aren’t replacing engineers, they’re taking over the mechanical parts of coding, letting engineers concentrate on system architecture, integration, verification, and human-centered thinking. => This is a productivity boost that you guys present as a "this job is over". Everyone can code now, but not everyone is an engineer. Yes, AI will continue to improve — but before it can confidently manage an entire codebase, there will likely be a shift away from LLMs toward new architectures for AGI, as Yann LeCun suggests.