7 days into using Grok 4.5 as my daily driver in Cursor, I think we’ve crossed the “good enough” threshold for AI-assisted coding.
At this point, it can build 99% of what I want fairly competently and quickly. The bottleneck is no longer whether the model can write the code. It’s whether I can clearly define what I want to be built and recognise if it's good.
One caveat: I’ve mainly used it for game development and a handful of software projects, so your mileage may vary in other areas.