There’s an enormous difference between great engineers using AI and randomly trying to one-shot everything.

This is often overlooked. Many people think using AI equals zero coding and things start breaking down for anything slightly more complex. But if you use AI in the right way while keeping yourself in the loop — then it makes so much difference.

Coral AI is the most powerful AI for documents. See the difference yourself:

One-shoting is the new 🤞while deploying some untested code

Wait sure he go on to say exactly he selects context? I want to know!

That is the reason why fundamentals are so so important and essential.

Do you have an example of a prd?

Hearing from legit engineers is such a hack. The perspective and just the way they structure sentences shows you there are levels to this

It's true. Agents today are awful at selecting the right context. I suspect at least some of this is because the AI IDEs want to save on input tokens but I guess that's why many are switching to Claude Code and Cline.

Grok 3 is interesting because it often creates code that anticipates feature sets that don't exist in the codebase yet, but should exist by logic. This is a negative where it assumes a little too much based on documentation and has no agency beyond a PR to effect change, but it's a net positive if you control the codebase because it can often construct the missing feature for you to incorporate and then move forward.