What This Is
This is an identification tool for 466 Carex taxa occurring in North America north of Mexico - drawn from treatment text in Flora of North America North of Mexico, Volume 23 (Ball, Reznicek & Murray, eds., Oxford University Press, 2002). It is not a dichotomous key. It is a multi-access filter: set any character in any order, defer what you cannot measure, and the key returns only the taxa still consistent with your specimen.
Approximately 15 characters are encoded per taxon — culm height, leaf width, perigynium dimensions, achene dimensions, terminal spike sexuality, spike arrangement, and distributional data for U.S. states and Canadian provinces — extracted as a flat, queryable data model embedded directly in this file.
Methodology
Character extraction followed the Čысты human-orchestrated multi-model convergence methodology (Jones, 2025; DOI 10.5281/zenodo.18988588), in which quantitative ranges were parsed directly from FNA treatment prose under human editorial direction and taxonomic validation. All numeric values are traceable to source treatments. The human editor, not any AI system, bears responsibility for taxonomic decisions, curation choices, and interpretive judgment throughout.
This file is built following the 20-Year File / Digital Paratype architecture (Jones, 2025): a single HTML artifact, readable by any standards-compliant browser without installation, dependency resolution, or network access. Provenance markers are embedded per Čысты model-protection methodology.
After the Einstein reconstruction, stopped thinking about this work as replacement and started thinking about it as restoration. The aim is not to alter the original, but to preserve it while restoring legibility, access, structure, and function for modern readers. In many cases, the scholarship itself remains sound, but the surrounding apparatus has degraded. Interfaces decay, annotations disappear, data become inert, and important texts remain technically available while becoming practically difficult to use. In that sense, restoration is not embellishment, it is overdue maintenance on the scholarly record.
My methodology is closer to foraging than retrieval. The distinction matters. Retrieval implies a clean pipeline. Query in, result out. Foraging means entering difficult terrain with no guarantee of what you will find, making judgment calls in the field about what is ripe and what is not, and accepting that you will get wet before you get home. The multi-model dimension is where this diverges most sharply from conventional research practice. Working across many models simultaneously, each with its own strengths, error profiles, and failure modes, is not a search strategy. It is a field methodology. Truth emerges from convergence, not from any single instrument. The pie at the end is real, and it is yours, because you did the work in the swamp that no one else wanted to do.
Have heard a range of analogies for this role, but the one that fits best for me is cattle dog. Not a conductor. A conductor begins with a score. This process has no score. It requires watching the whole field, recognizing tendencies, anticipating breakdowns, and redirecting effort before it goes wrong. The models are not passive instruments. They have patterns, momentum, and recurring limitations, and left alone they will drift. The task is to know those characteristics well enough to stay close, move fast, and nip at the heels when direction is needed. The touch should be precise, not gentle. A cattle dog that hesitates loses the flock. The human role in this work is not supervisory in any quiet sense. It is constant, reactive, and at times loud. The models move because someone is there making sure they move.
This is a primitive. Not an application. Not a traditional webpage. A zero-dependency, single-file HTML document that runs in any browser, requires nothing, and will still open in twenty years. Just save it somewhere somehow. Spent decades writing for the web before understanding what that actually meant to build archival works. No framework. No coupon required.