Before the Word for It
The most accurate description of any experience is the one made before there's a word for what it is. We associate articulation with accuracy. The instinct runs wrong. Language isn't neutral; vocabulary is already a theory.
When the Trinity photographers recorded the first nuclear detonation, they were doing a job. Set the frame, trip the shutter. The images are honest because the conceptual vocabulary that would have organized the shot didn't exist yet. Five years later, a photographer at a nuclear test knows what they're photographing, and that knowledge shapes what they frame: the cloud and the scale. The earlier record is more accurate about the thing; the later one is more accurate about what the thing means. Those are different accuracies.
The mechanism runs in both directions. A framework lets you see more. A trained observer catches things a naive one misses. But it also organizes what you see into the slots the framework already has. You record the exemplary case, the thing that confirms. The anomalies go unrecorded because you have no slot to file them in. The selection happens at the moment of noticing, not at the moment of writing.
Medical symptom diaries make the case clearly. Pre-diagnosis, a patient records what they actually notice: light sensitivity on Tuesdays, the specific quality of fatigue after stairs. Post-diagnosis, those same observations get reorganized into vocabulary. Light sensitivity becomes "photophobia." Useful for treatment. The particular shape of the experience disappears into the category.
First encounters work the same way. The kid who learned to program on a 1981 home computer didn't know he was acquiring "computational thinking." He knew he'd typed something and his name filled the screen in a loop. Something happened, it worked, it was his. Decades on, the description comes through the frame. Whatever he wrote or said at the time is the only access to what it actually was.
Code comments read the same way if you know how to look. A comment written when a function was first drafted describes what the author understood they were building at that moment. Three refactors later the function has changed; the comment hasn't. It looks wrong. It isn't wrong. It's a timestamp on an understanding, the only accurate record of how the code worked before the current architecture's conceptual frame reorganized everything. Most people update it or delete it.
The arrival of a framework doesn't end honest observation. You can work against your categories, notice what doesn't fit, hold the anomaly long enough to get it down. But the window for pre-framework record-making is narrow. It closes when the vocabulary arrives, and from inside the framework you can't fully recover what it covered over. You can describe the category. You can't reconstruct the thing before it became one.
The record made before the word existed is the only one not already written in the word's shadow.