AI feedback can do more than help us move faster or reinforce ideas we already have. This week’s Weekly Challenge asks you to use AI differently: ask it to tell you what you are doing wrong. Instead of asking for another solution, use the tools you already use to identify a bad practice, security concern, weak assumption, or area for improvement.
This challenge grew out of our conversation with Salvatore Manzi about communication, feedback, and building a career beyond the code. One theme kept coming back: positive reinforcement feels good, but constructive criticism often drives real improvement. Whether you are building software, creating a product, communicating with a team, or developing your career, you need people—and sometimes tools—willing to tell you when something does not work.
Why Negative Feedback Can Be More Valuable
We naturally like hearing that an idea is good. It is much harder to hear that something does not make sense, will not work, or needs to change.
The problem is that constant positive feedback does not give us much information. Michael connected this idea to building products and businesses. You can have what you believe is the perfect application, but that belief does not matter if you have not talked to the people who might actually use it. Your reason for building something and your customer’s reason for buying it may be completely different.
It is not enough to ask someone whether your idea sounds good. You eventually need a stronger question: Would you actually buy it?
The “no” answers matter. They can tell you that you are talking to the wrong customer, solving the wrong problem, presenting the solution incorrectly, or simply building something people do not want. That feedback can hurt, but it also gives you information you can use.
Think About AI Feedback Like Testing Software
Rob connected negative feedback to something developers already understand: testing.
Imagine running one hundred unit tests and having every test pass immediately. That sounds great, but an experienced developer may start wondering whether the tests are actually testing enough.
Now imagine that fifty tests fail. You suddenly have information. You know where something is broken, and you have specific problems you can investigate. After fixing those problems and watching the tests pass, you have much more confidence that you made meaningful progress.
AI feedback can work the same way. “That’s great” gives you very little to work with. “This part works, this part doesn’t, and here’s why” gives you something actionable.
The goal is not to surround yourself with negativity. It is to create feedback loops that reveal problems early enough for you to do something about them.
AI Feedback Can Become Too Agreeable
There is an interesting complication when AI becomes part of that feedback loop.
Many of us use AI because it is easy to brainstorm with. We can throw out an idea, explore it, refine it, and quickly move forward. However, if every conversation becomes reinforcement for what we already think, we lose some of the value.
During this week’s discussion, we talked about deliberately pushing AI away from the “sunshine and rainbows” response. Instead of asking it to help you execute your next idea, ask it to examine how you are currently working and identify weaknesses.
Useful AI feedback should not simply tell you that you have a great idea. Sometimes the most valuable response is the one that tells you where your idea, process, or implementation falls short.
That brings us to this week’s challenge.
This Week’s AI Feedback Challenge
Choose one area where you regularly work with AI.
Do not ask AI to solve another problem for you. Instead, ask it to review how you have been working and identify something you could improve.
Your question can be as simple as:
- Based on the work we’ve been doing, what am I doing that is not a best practice?
- For developers, you might narrow the question:
- Based on the code and approaches we’ve been discussing, where am I following weak development or security practices?
You could apply the same exercise to communication, project management, business planning, writing, testing, or another area where you regularly use AI. The idea is intentionally simple: ask where you can become better and then use the answer as a starting point.
Keep Your AI Feedback Focused
There is one important guardrail: do not ask AI to critique everything at once.
Michael pointed out how quickly this exercise can become another AI rabbit hole. Ask about security and suddenly you have twenty categories to investigate. Follow one suggestion, and that produces another five. Before long, you have forgotten the original problem you were trying to improve.
Instead, choose one or two areas. If AI returns a long list, do not immediately ask it to expand every item. Save the list and select one thing you can realistically work on.
The process should look something like this:
- Pick one area of your work.
- Ask AI what you are doing poorly or where you are missing a best practice.
- Ask for a short, prioritized list.
- Choose one item from that list.
- Verify that the criticism makes sense.
- Make one improvement.
- Return to the list later and repeat the process.
This turns AI feedback into a tool for reflection rather than another source of endless tasks.
Do Not Let AI Automatically Fix Everything
There is another important distinction between asking for feedback and permitting AI to act.
As AI tools become increasingly capable of modifying files, changing code, interacting with applications, and performing multi-step tasks, a question can sometimes turn into an action faster than you intended.
Michael recommended configuring your tools so they ask for approval before making changes when possible. If you ask, “What am I doing wrong?” you want an assessment. You do not necessarily want the tool immediately rewriting your code or changing your environment.
For this challenge, keep those stages separate:
- Review what AI identifies.
- Decide whether the recommendation makes sense.
- Act only after you understand the change.
That separation keeps you involved in the thinking process instead of allowing AI to turn every recommendation into an automatic change.
Remember That AI Feedback Still Needs Evaluation
There is also an important limitation to this exercise: AI can be wrong.
Treat AI feedback the same way you should treat feedback from a colleague, customer, code review, or automated test. Examine the evidence. Decide whether the recommendation applies to your situation. Test the change when appropriate.
The purpose of this challenge is not to make AI the final authority on how you work. It is to use AI to help expose assumptions and habits that you may no longer notice.
That is especially useful when you have worked on something for a long time. Familiarity makes it easy to accept an approach simply because it has always been your approach. A fresh critique can make you look again.
Better Communication Requires the Same Skill
This exercise also connects directly to our conversation with Salvatore Manzi.
Developers and engineers often move quickly from hearing a problem to solving it. That works well in some technical situations, but people do not always communicate that way. Sometimes someone needs to know that you understood the problem before they are ready to discuss the solution.
As your career moves beyond purely technical conversations, that distinction becomes increasingly important. You begin communicating with customers, business leaders, managers, sales teams, and other people who may process information differently.
Learning to bridge that gap can make you much more valuable than simply being the person who knows the technical answer. Constructive feedback works the same way. You have to listen to it before racing toward a solution.
Your AI Feedback Goal for This Week
Your challenge is intentionally small: ask AI what you are doing wrong in one area of your work.
Do not ask it to tell you how great your idea is. Do not ask it to immediately fix everything. Ask it to identify weaknesses, bad practices, or assumptions worth reconsidering.
Then pick one recommendation and investigate it.
Maybe you discover a security practice that needs improvement. Maybe your code could be simpler. Maybe your emails have started sounding like AI. Maybe your project plan contains an assumption you never validated. Maybe the tool identifies something you disagree with completely.
Any of those outcomes can be useful if they make you stop and examine your work.
We spend a lot of time asking AI to give us answers. This week, change the conversation.
Ask it to challenge you instead.
Stay Connected: Join the Developreneur Community
👉 Subscribe to Building Better Developers for more conversations on momentum, leadership, and growth. Whether you’re a seasoned developer or just starting, there’s always room to learn and grow together. Contact us at [email protected] with your questions, feedback, or suggestions for future episodes. Together, let’s continue exploring the exciting world of software development.