Most engineers are using AI wrong. They treat it like a code generator. Prompt in, code out, ship it. That's the shallow end of the pool. The real leverage is using AI as a thinking partner. Challenge your architecture before you build it. Poke holes in your assumptions. Walk through trade offs you haven't considered. Review your design like a principal engineer would. The code is the easy part. The thinking is where engineers earn their salary. Teams that figure this out will outship teams that don't. Not by a little. By a lot. Link in comments. #ai #java #softwaredevelopment #coding
When I use AI, I set clear boundaries upfront. Do not change anything and do not start coding yet. First we align on the approach and brainstorm. This usually leads to better outcomes.
Thinking with you?! Are you fcking for real? AI is a robot, it follows orders. If you can't think of the whole structure in your head and order AI to do it you shouldn't be near code in the first place. If you can't think you shouldn't be near AI at all 🤣🤣🤣
Very well said ! , I prefer using AI to upskill myself rather than just copy pasting the code for the sake of completion that too without understanding it !
They are not mutually exclusive. Use AI as an enhancement tool at all steps: Ideation, planning, implementation and QA. The first steps are the highest leverage. Don’t try to save tokens there. Ask the AI to ground all claims in external information, not its training data. Use it as a RAG for yourself, but you have to ensure alignment, and you have to understand what you are building.
When it comes to thinking, I think I prefer to use my brain.
Honestly, as someone who uses AI mainly to write boring code, I think it’s the correct approach. Im seeing more and more people and engineers using AI to “think with them”, but ending up letting it think instead of them. We started using auto completion long ago, but it was the ability to think that made a difference. AI is a great tool, but relying on it for thinking, can make developers lose their edge. Blabbering on a bit, but basically, it can and should be used as a rubber ducky, but I think that its our instincts that make the difference and we should keep them 😁
Knowing the limitations of AI (from basic computability theory as well as having seeing it work), it's dangerous to trust it too much. Use it or lose it: don't leave critical thinking to it (keep doing it yourself to keep the ability -- studies show it's already happening) and remember you'll always have to fix something.
The real deal is context engineering, that what separates engineers from vibe coders.
This resonates a lot. The biggest value I get from AI is usually before the code is even written. Asking it to challenge assumptions or compare approaches often helps more than just generating the first implementation. That’s where it starts feeling less like autocomplete and more like a useful thinking partner.
One of the worst tips in my opinion when it comes to AI .