As a software engineer, this something I've been seeing for a while.
In 5–10 years, we may not have enough truly senior developers, not because AI replaces them, but because it's taking away many of the tasks that helped juniors build judgment. Writing code, debugging and solving everyday problems were how many of us learned to think and make better decisions.
Many companies still measure engineers by throughput—how much code they write. That metric is becoming less useful when AI can generate code so quickly.
I think we should focus more on how engineers use AI to spot gaps in requirements, question assumptions, improve the customer experience, and make better technical decisions. Writing code is becoming easier. Good judgment is becoming more valuable.
You’re right Vasanth. But I don’t think this is just a software engineering issue. We’re likely to see the same challenge across many knowledge-based professions. The real question is how we redesign learning and development so expertise continues to grow, even as AI takes on more of the day-to-day work. Definitely an important issue to address in organizations.
Hi Clarice, informative read! Glad people like you are bringing awareness to this topic. In April 2026, I shared some research related to this under my framework, Somagraphic Learning™️
3 behavioral indicators track transfer Section 13.1 of the preprint, (Toprani 2026):
1️⃣ Does the query shift from generation-based to gap-based?
2️⃣ Does the user modify AI output against their prior map rather than accept it passively?
3️⃣ Do directional relationships in the pre-AI diagram get confirmed or corrected in the Refine stage?
⚖️ Without a prior structure, there is nothing to compare the AI output to. With structure, the user is already an evaluator, not a receiver.
Since this directly resonates with your post, I’d love to explore alignment.
Love it! I really hope both articles together serve to start the conversation about performance management in the AI era!
Thanks so much for the great exchange and your patience throughout. I’ve really enjoyed collaborating with you - it’s been a lot of fun! 🤩
Likewise this was really great I hadn’t looked at it from this lens prior and our work let me reframe things for my current job!
Great essay, Clarice! Very deeply thought and very sharp, pointed and practical suggestions. Love it!
Thank you, that really means a lot. ❤️ So glad it resonated with you. I really enjoyed writing this one.
As a software engineer, this something I've been seeing for a while.
In 5–10 years, we may not have enough truly senior developers, not because AI replaces them, but because it's taking away many of the tasks that helped juniors build judgment. Writing code, debugging and solving everyday problems were how many of us learned to think and make better decisions.
Many companies still measure engineers by throughput—how much code they write. That metric is becoming less useful when AI can generate code so quickly.
I think we should focus more on how engineers use AI to spot gaps in requirements, question assumptions, improve the customer experience, and make better technical decisions. Writing code is becoming easier. Good judgment is becoming more valuable.
You’re right Vasanth. But I don’t think this is just a software engineering issue. We’re likely to see the same challenge across many knowledge-based professions. The real question is how we redesign learning and development so expertise continues to grow, even as AI takes on more of the day-to-day work. Definitely an important issue to address in organizations.
Hi Clarice, informative read! Glad people like you are bringing awareness to this topic. In April 2026, I shared some research related to this under my framework, Somagraphic Learning™️
3 behavioral indicators track transfer Section 13.1 of the preprint, (Toprani 2026):
1️⃣ Does the query shift from generation-based to gap-based?
2️⃣ Does the user modify AI output against their prior map rather than accept it passively?
3️⃣ Do directional relationships in the pre-AI diagram get confirmed or corrected in the Refine stage?
⚖️ Without a prior structure, there is nothing to compare the AI output to. With structure, the user is already an evaluator, not a receiver.
Since this directly resonates with your post, I’d love to explore alignment.
More info here: https://soulfullearningwithai.carrd.co/