HUMAN.md: Are we losing our skills in the SKILL.md era?
summary
Claim: industry adoption of SKILL.md trains agents to navigate code but produces human skill atrophy — juniors prompt stack traces instead of learning to debug, mid-levels lose blast radius intuition and seniors become full-time hallucination auditors. Concrete failures include a two-line MR that later causes a production NullPointerException, and problems that surface around 18 months after rollout. Proposed fixes are explicit rules: no AI-generated MR merges without a live rationale and scheduled synthetic friction like no-AI debugging days.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 3:15Skills.md and human atrophy
- 6:50Productive friction and Android ladder
- 10:03Stack Overflow outage and juniors
- 13:16Juniors gained and lost
- 16:35Mid-level atrophy and intuition loss
- 19:53Senior developers' changing role
- 23:05Technical hollowing consequences
- 27:00Trust AI-architected house?
- 31:39Diagnosis and treatment
- 37:18Conclusion
related talks
35mSkills.md and agents.md You Must Master in 2026!
15mThe AI Shift: Rethinking the Developer’s Role in a Machine-Driven World
27mSkills are the New SDKs
11mWhy (Senior) Engineers Struggle to Build AI Agents
41mSkill Degradation: An Empirical Analysis of 400+ AI‑Generated Security Fixes
17mManaging the Gap AI Brings: Keeping Developers Sharp and Engaged
01h 19mSkill Issue: How We Used AI to Make Agents Actually Good at Supabase
23mRethinking growing engineers in the age of AI