Two Modes, One Codebase: Durable vs. Disposable Code in the Age of Cheap Generation

Durable and disposable code aren't two points on a quality scale — they're two different kinds of software with different cost models and different reasons to exist. AI codegen has collapsed the cost of the disposable kind and shifted the ratio between them, which makes the boundary between the two the real engineering skill: naming a piece's half-life up front, designing the seam so promotion or disposal is cheap, and refusing to let throwaway glue silently harden into a load-bearing production dependency. This deep dive starts from the four blog posts that framed the debate and checks them against the 2026 research, security data, and regulatory calendar.

In this episode

  • The frame, from four blogs. Charity Majors' bifurcation thesis ("disposable software is a skill set; durable code is a profession"), Atomic Object's lifecycle take, a fintech "two tracks" model, and the contrarian "fast, cheap, good — choose three."
  • The cost model, now measured. Maintenance — not authorship — is the cost. DORA's negative correlation between AI adoption and delivery stability, the 30–41% technical-debt rise, and the velocity gains that vanish at ~two months.
  • The seam. The discipline the anchor posts gestured at but never drew: Feathers/Bland/Fowler seams and anti-corruption layers as the structural mechanism for cheap disposable-to-durable promotion, paired with observability that flags accidental durability.
  • The drift, with a security bill. Veracode's ~55% security pass rate (flat since 2023), Georgetown's independent corroboration, the agentic self-improvement loop that raised critical vulnerabilities 37.6%, and "slopsquatting" — the hallucination-born supply-chain attack.
  • Two tracks, now legally enforced. The EU AI Act's August 2026 high-risk deadline, DORA (the resilience act), the May 2026 US executive action, and why fintech compliance guidance says human review can't be automated away.
  • The discipline nobody has tooling for. No standard for code-expiration metadata; why expiry without renewal governance is its own fragility vector; and Marimo vs. Jupyter as the two-modes tension inside data science.
  • The contrarian beat. The case that AI dissolves the dichotomy entirely — the real 10–100× cost collapse — and why it quietly depends on the same seams and feedback loops Majors demands.

Sources & References

Primary / originating sources (operator-provided — ground zero)

Research & critique

Security & supply chain

Governance, regulation & compliance

Architecture, lifecycle & tooling

The cost-collapse counterpoint

Have questions about this episode? Reach out.