Plain-text files are at risk, an essay on computing's universal format argues

A short essay circulating in developer forums argues that plain-text files — the .txt, .md, .ini and .json files that served as computing's most universal data format — are losing the conditions that keep them usable. Published as an unlisted paste on paste.sr.ht by an author writing as ~awal and surfaced on Lobsters, it is a culture argument: the formats are sound, the ecosystem around them is not.

A universal format by default

The essay defines the category carefully. A plain-text file is a real file on a filesystem: you can attach it to an email or serve it over HTTP or a NAS share like any other file. Its content is text only, so even a minimal editor such as nano can open, change and save it. Different editors can work on the same file without conflicting, while rich-text editors often cannot save it without breaking its plain-text nature.

Nearly every operating system ships such an editor by default — Linux and the BSDs, but also Microsoft's Notepad and macOS's TextEdit. Yet interest in plain-text editing has been fading for years; the author cites Google Trends data for “text editor”.

How smartphones broke the habit

Twenty years ago, the essay notes, ordinary people used plain-text editors routinely for notes, contact numbers, dates and shopping lists. The biggest blow came when smartphones replaced desktops: the dominant mobile platforms shipped no plain-text editor, and no serious third-party ecosystem for one emerged. Nearly all text entry on phones happens in proprietary formats inside specialized apps.

The result, in the author's view, is that the average computer user has forgotten plain-text files entirely — and even the idea of a “file” is eroding.

Engineers, LLM repls, and the risk

Software engineers are the last group keeping the practice alive: code is plain text, and building or administering systems forces regular contact with source, configuration and log files. That produced a large ecosystem of code editors and plugins. But the essay sees this final audience drifting too — developers increasingly type inside LLM repls, the interactive loops built around large language models, and documentation is handed to those tools as well. Generated code and docs are often reviewed only as a diff in a web interface.

If that audience leaves, the essay expects maintenance of plain-text editors and their ecosystem to stop. Since operating systems and libraries keep changing, existing editors could stop working on updated software stacks, and file managers could follow the same path. The closing point is about incentives: even people who doubt LLM repls can replace every tool they promise adopt them as if they can. “Society can remain irrational longer than you can remain a relevant nonconformist,” the essay concludes, warning that “the most versatile, universal data interface we've ever had might no longer be accessible.”