I’m noticing different "thinking modes" in how I use AI
Hey everyone 👋
Lately I’ve been noticing something in my workflow:
Sometimes I jump into prompting AI before I fully understand the problem itself.
Or I use AI to reduce friction instead of slowing down and thinking more deeply.
And because the output often sounds convincing, verification can become weaker too without noticing.
That’s what pushed me to start experimenting with different "thinking modes" before using AI:
Explore
Challenge
Audit
Reflect
I dont use this as a rigid framework but more as a reminder that different tasks require different kinds of thinking.
I’ve started tracking these modes over time to better understand how I’m actually using AI.
I’m starting to think many AI mistakes are not prompting problems.
They’re thinking problems.
Curious if anyone else has noticed similar patterns in their workflow when using AI tools?
What habits or systems are helping you think more clearly while using AI tools?
Replies
I’ve noticed the same thing. Sometimes I use AI to escape uncertainty instead of sitting with the problem long enough to understand it first.
Jots
@perceval_drake it's good you have awareness of that, I am adding no AI usage blockers in my day when I feel that happens too frequently
I’ve noticed the same thing. The biggest mistakes usually happen when I ask AI to move before I’ve defined what a good answer should look like. I try to separate the flow into: understand the problem first, let AI explore options, then switch back into audit mode before trusting anything.
Jots
@alpertayfurr that is a great flow. Manual touchpoints both at the start and end to control for deviations.
@avz Exactly. I think those touchpoints are what stop AI from drifting in a confident but wrong direction. The middle can be fast and messy, but the start and end need real human judgment.
Mostly switching between exploring/validating my own ideas and delegating boring/boilerplate code generation to the LLM, then reviewing manually and making sure that it makes sense. So I agree - in the end, the responsibility is mine, therefore it wouldn't be proper to blame the LLM for something that I missed
Jots
@sk_uxpin good point, the responsibility line is important here
I find it useful to let AI ask me questions first, especially when I’m still not fully clear on what I want to build or why.
It helps me slow down, clarify the problem, and avoid jumping straight into execution too early.
Jots
@not_icso agree, for planning and ideating it is useful to get inspired and weigh options
The mode framing is useful because it gives you a reason to pause before the first prompt.
One habit that helps me is naming the review job before generating: “I’m exploring,” “I’m trying to preserve a point of view,” “I’m checking claims,” or “I’m cleaning wording only.” The risky moment is when those collapse into one generic “make this better” request — AI can improve the surface while changing the judgment underneath.
For writing especially, I’d add a small before/after audit: what did the AI clarify, what did it invent, and what did it smooth out that was actually important?