The 80% problem: comprehension debt replaces the last-mile gap¶
Insight: The "70% problem" has evolved to 80%+ completion by agents, but the nature of the problem shifted from last-mile code completion to comprehension debt. AI errors evolved from syntax bugs to conceptual failures: assumption propagation (building on faulty premises), abstraction bloat (1,000 lines where 100 suffice), dead code accumulation, and sycophantic agreement. Jeremy Twei coined "comprehension debt" — the gap between code generated and code understood by its nominal owner.
Detail: Addy Osmani documents the shift from 80% manual coding to 80% agent coding (citing Karpathy and Boris Cherny of Claude Code). Armin Ronacher's poll of 5,000 developers shows 44% now write less than 10% of code manually. Key failure modes: agents don't manage confusion, don't seek clarification, don't surface tradeoffs. The "verification bottleneck" data shows only 48% of developers consistently check AI code before committing, while 38% find reviewing AI logic harder than human-written code. Abel Seyoum argues builder literacy must exceed AI capability — long-running agents change the work from requesting code to delegating execution inside systems you're responsible for.
Sources
Related: existing entry "The 70% problem" in external/claude-code.md — SUPERSEDES