In a May 29, 2026 essay, a former historian turned backend engineer at Hagerty Insurance argues that software

A former historian who began a tech career at 32 and now works as a backend engineer at Hagerty Insurance published an essay on May 29, 2026 warning that AI threatens an informal oral pipeline that keeps software maintainable. The author says teams routinely rely on people’s recollections — stories, warnings and tacit memory about why code was written — to navigate complex systems, and that losing that human layer would make future maintenance harder.

The essay catalogs where that oral knowledge lives: legacy codebases with few authoritative design documents, early specifications that were rewritten before implementation, scattered wiki pages mixing long-solved and unresolved issues, and terse in‑line comments that merely warn “don’t change this.” Even straightforward technical questions — such as why a stored procedure was chosen — are often answered by tracking down a long‑tenured developer rather than consulting durable documentation.

The author ties these operational habits to industry culture. Documentation is commonly deprioritized because engineers find coding more engaging, and because Agile’s core value—“working software over comprehensive documentation”—helped normalize underdocumentation as a virtue in contrast to Waterfall‑era bureaucratic specs. That ideological context, the essay argues, explains why teams tolerate gaps in recorded rationale.

Staff turnover amplifies the risk. Employers commonly shed engineers every few years, the author notes, unlike historical oral cultures where storytellers remained stable; on some projects they became the only original contributor left. That personnel churn shifts maintainers’ central question from “what does this code do?” to the harder “why was it written this way?”—the kind of provenance that oral exchange traditionally supplied.

AI tools that automate code reading and generate summaries may reproduce surface behavior — the “what”—but are ill‑suited to transmit provenance, nuanced trade‑offs or the cautionary tacit knowledge held by living practitioners, the essay warns. For maintainers, this gap means assistants could reduce incentives to record decision rationale while simultaneously making it harder to reconstruct institutional memory after people depart.

The author does not dismiss oral tradition outright — historical societies successfully passed complex knowledge by word of mouth — but stresses the mismatch between rapid industry turnover and oral transmission. The practical prescription is clear: engineering teams should preserve records that explain trade‑offs and motivations, not only code or bug tickets, so both human successors and AI tools can access the “why” behind engineering choices.