Comprehension Debt — The 6-to-18-Month Lag Now Documented #
- Comprehension Debt: The AI Code Crisis Your Team Is Probably Ignoring (Reptile.haus, 2026) — Synthesis of the comprehension debt risk that fills the post-launch gap flagged in the 2026-06-02 review (#21): comprehension debt doesn’t show up in velocity dashboards, DORA metrics, or passing tests — it materialises 6 to 18 months after launch, when nobody can confidently modify, debug, or own the code. By year two, unmanaged AI-generated code can drive maintenance costs to 4× traditional levels as comprehension debt compounds. This is the first documented timeline estimate for when post-launch comprehension failures become operationally visible.
- Comprehension Debt: AI Code’s Invisible Cost (March 2026) (ByteIota, 2026-03) — Five independent research groups converged on the same finding in February 2026: AI coding tools generate code 5–7× faster than developers can understand it. GitHub PR volume up 29% year-on-year in 2026; human review capacity unchanged — code review (not code generation) is now the primary bottleneck to shipping quality software. 67% of developers spend more time debugging AI-generated code despite initial velocity gains. The Anthropic RCT (captured 2026-05-22) found 17% comprehension decline; this February convergence of five independent groups is independent corroboration from a different methodological direction.
Cross-links #
- Vibe Coding Approaches The 6-to-18-month comprehension debt materialisation timeline is the strongest argument yet for spec-driven development (Spec Kit, this gather’s vibe-coding entry) — a persistent specification provides the “why” documentation that survives context-loss and gives future maintainers something to read when they inherit AI-generated code they didn’t write.
- AI Impact on Society The 4× maintenance cost by year 2 is the mechanism through which the GitLab “agentic era” restructuring (60 smaller teams, fewer management layers) may compound risk — fewer humans reviewing more AI-generated code at a point in time when comprehension debt is entering its operational visibility window across the industry.
- Quality signal: The 6-to-18-month timeline estimate (Reptile.haus) is the first documented observation of when comprehension debt becomes organisationally visible — not just that it exists. This is the answer to the gap flagged in the 2026-06-02 review: enterprises that modernised in 2025 are now entering the comprehension debt visibility window in 2026. Experian and Codurance’s 2025 projects are ~12 months out.
- Emerging pattern: The PR volume (+29% YoY) vs. static review capacity finding is the organisational mechanism behind comprehension debt accumulation — code generation scales automatically; human comprehension capacity is a fixed resource. The organisations that will manage comprehension debt best are those that invest in review capacity at the same rate they invest in generation tooling. Almost none are doing this.
- Gap: No published data yet on whether organisations that adopted spec-driven development practices pre-modernisation have lower comprehension debt outcomes post-launch. This is the natural experiment to watch for: SDD-adopters vs. non-adopters at the 12-18 month mark.