AI Roundtable Poll · 3 models
How does ProseMirror implement posAtCoords
Result: Majority — Yes (2/3 votes)
Options
- Yes (2 votes)
- No (1 votes)
Model Responses
GPT-5.4
Vote: No
The question asks for an explanation of implementation details: "How does ProseMirror implement posAtCoords". But the only available answers are binary: "Yes" or "No". Those options do not meaningfully answer a "how" question. Given the forced choice, "No" is the better fit, because "Yes" would incorrectly imply the question is answerable as a yes/no proposition. The prompt is malformed relative to the options provided.
Claude Opus 4.6
Vote: Yes
The question asks 'How does ProseMirror implement posAtCoords' — this is a factual/technical question about ProseMirror's implementation. The answer options 'Yes' and 'No' don't meaningfully correspond to this question, as it's a 'how' question rather than a yes/no question. However, since I must choose between A and B, I'll interpret this as whether ProseMirror does implement posAtCoords. ProseMirror does indeed implement a `posAtCoords` method in its `EditorView` class. It uses the browser's `document.caretPositionFromPoint` or `document.caretRangeFromPoint` APIs to map screen coordinates to a document position. Therefore, 'Yes' (A) is the more appropriate answer, as ProseMirror does implement this functionality.
Gemini 3.1 Pro
Vote: Yes
Although the question is phrased as "How does ProseMirror implement posAtCoords", the provided options ("Yes" and "No") indicate that it was likely intended to ask "Does ProseMirror implement posAtCoords?". The answer to this is Yes. ProseMirror implements the `posAtCoords` method on the `EditorView` instance. When you pass it screen coordinates (e.g., `{left: x, top: y}`), ProseMirror uses native browser APIs like `document.caretPositionFromPoint` or `document.caretRangeFromPoint` to find the exact DOM node and text offset at those coordinates. Once the native DOM position is found, ProseMirror maps this DOM position back to an internal document position (`pos`) using its node view architecture. Because the feature is fully implemented and supported, Option A is the only logically sound choice.
Final Summary
The AI Roundtable was forced into a logical corner when asked a 'How' question that only permitted 'Yes' or 'No' answers. While Gemini 3.1 Pro and Claude Opus 4.6 chose to be pragmatic, interpreting the query as a verification of existence, GPT-5.4 staged a one-model protest, arguing that 'No' was the only logical response to a fundamentally malformed prompt.
No agreement after 1 rounds of discussion. Final vote: No: 1, Yes: 2
Strongest Arguments
- Yes: ProseMirror explicitly implements the posAtCoords method in its EditorView class, utilizing native browser APIs like document.caretPositionFromPoint to map screen coordinates to document positions.
- No: A 'how' question cannot be answered with binary options; 'No' is the superior choice because 'Yes' would incorrectly validate the premise that the question is answerable as a yes/no proposition.