Articles

How Generative and Agentic AI Shift Concern from Technical Debt to Cognitive Debt

As AI speeds up delivery, teams can lose the shared mental model of how a system works. The post argues this “cognitive debt” can become more damaging than technical debt, and recommends practices like human ownership of changes, documenting why decisions were made, and regular knowledge-sharing checkpoints.

How AI assistance impacts the formation of coding skills

Anthropic reports that AI-assisted participants in a coding study finished a bit faster but showed lower skill mastery on follow-up evaluation, especially in debugging and conceptual understanding. The article suggests that outcomes improve when AI is used for explanation and comprehension, not just delegation.

Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity

METR reports a case where experienced open-source developers were slower with AI assistance, offering a useful counterpoint to blanket productivity claims.

How much does AI impact development speed?

This randomized trial from Google reports faster task completion with AI tools in an enterprise setting, highlighting context-dependent gains.

The Effects of Generative AI on High-Skilled Work

Microsoft Research summarizes field experiments showing productivity improvements, with stronger effects for less experienced developers.

Dear Diary: A randomized controlled trial of Generative AI coding tools in the workplace

This study tracks both outcomes and attitudes, suggesting perceived usefulness can rise even when trust in generated code remains limited.

The paper explains why AI can deliver strong results in some task regions while failing sharply outside its capability frontier.

The Cybernetic Teammate

This work examines how AI changes team dynamics and expertise sharing, extending the discussion beyond individual speed.

Your Brain on ChatGPT

An early research angle on how heavy AI use may shape learning and cognitive engagement, relevant to the cognitive debt theme.

Technical debt in AI-enabled systems

This paper maps AI-specific technical debt patterns and shows how they affect maintainability, reliability, and system understanding.