What “AI mode” are you in when you build?

by

Hey everyone 👋

I’ve been continuing to explore the question:

How do we use AI without outsourcing our thinking?

One thing I’ve started noticing in my own workflow is that not all AI usage feels the same.

Sometimes AI helps me think more clearly.

Sometimes it helps me move faster.

But sometimes, if I’m honest, I use it because I want to avoid friction. Friction being active thinking in this context.

That pushed me to think less about “using AI” in general, and more about AI modes:

Supportive modes

  • explaining unfamiliar code

  • exploring tradeoffs

  • critiquing a plan

  • testing assumptions

Mixed modes

  • generating boilerplate

  • suggesting refactors

  • drafting documentation

Risky modes

  • blindly accepting generated solutions

  • delegating architecture too early

  • debugging through AI without understanding the root cause

Over time, these habits shape how we think as builders.

I started grouping these patterns into 4 cognitive archetypes:

AI Architect
AI expands your thinking without replacing ownership.

AI Balancer
Mostly healthy usage, but mixed-mode creep needs monitoring.

Autopilot Builder
Efficiency may be masking weaker comprehension.

AI Passenger
AI may be driving too much of the reasoning path.

These archetypes are not meant to be scientific labels. They are just a way to build awareness.

I also built a small AI Thinking Balance Tracker around this idea to help track AI modes, dependency drift, and whether AI is acting more like leverage or replacement over time.

I’m sharing it for free and iterating based on feedback.

Curious to hear from other builders:

What AI mode are you in most often?

And if you had to pick one, which archetype best describes your current workflow?

Happy to share the tracker if anyone wants to try it. Just comment and I will share the link with you. 🙌

106 views

Add a comment

Replies

Best

The dangerous part is when speed starts feeling more rewarding than understanding.

 completely agree, productivity gains in the short term can be misleading if understanding doesn't follow. I built a tracker that measures drift in AI usage and recommends action to stay aligned. Feel free to check it out if you are interested, you can get it for free here:

i like this perspective because most AI discussions focus only on speed. For me, the bigger question is whether AI is improving my reasoning or quietly replacing parts of it over time

 I am glad this resonates with you. I created a tracker to detect drift in reasoning through the concept of AI modes and other variables. If you are interested, you can try it out for free, would love feedback on it:

I like the idea of tracking dependency drift over time instead of treating AI usage as simply good or bad.

 I am glad this approach resonates with you. Feel free to try out my free tracker:

I’d probably place myself somewhere between AI Architect and AI Balancer. For me, AI works best when it sharpens my thinking instead of replacing it.

 that looks like a solid place to be. I couldn't agree more with your approach here. Here is my tracker if you want to try it out for free, curious to know your thoughts!

Sometimes AI removes friction. Other times it removes learning. The difference matters.

 absolutely. Feel free to try out my free tracker to detect drift in AI usage over time:

If I have to choose between the modes that you described, I'd go with the mixed one, mostly because I'm sometimes working as an outsourcer as well, so I need to clearly understand what the code does.

 that makes a lot of sense. Using AI tools to increase learning and understanding of an unfamiliar codebase is a good use case. Here is my tracker if you want to try it out for free, curious to know your thoughts!

I think my best mode is “AI as leverage, not replacement.” I’m comfortable letting it move fast on implementation, refactors, and exploration, but I still want human judgment around architecture, risky changes, and anything that changes the product’s core behavior.