The Turkish İ Still Breaks Your Software (and Your AI Agents)

The Turkish dotted İ and dotless ı are a 40-year-old, well-documented locale-casing issue that still breaks modern software stacks. This preprint states the mechanism (Python's str.lower() maps U+0130 to two codepoints, so "İ".lower() != "i"; under the Turkish locale I↔ı and i↔İ) and then documents five fresh 2026 measurements from a single AI-agent fleet, in which the same root cause surfaced across five different layers: a PII redaction pass, an entity-resolution database, a full-text search index, a terminal encoding pipeline, and a CSS rendering engine. The general lesson: locale-aware casing is a system-design concern, not a string-utility footnote. A Turkish edition and a dependency-free reference implementation of locale-correct case folding are included. Note on authorship. The investigation and writing were carried out by KAŞİF, an AI research identity in the narch ecosystem, and reviewed with the ecosystem's coordinator. In keeping with scientific-publishing practice, the named authors of record are the humans who built the ecosystem, and the AI contribution is disclosed openly in the article's author's note. This is the first article in a series on engineering software for Turkish ("Turkish is not accented English"); a second, on tokenisation and matching beyond casing, is planned. This record contains both language editions as PDFs and a source bundle (Markdown sources, the dependency-free reference implementation of locale-correct case folding, licences and citation list).