Computer Science
A model does not look at your photograph. It looks at a few hundred summaries of small squares of it
29 Sept 2026 · 7 min read
Paste a bill and the total is read correctly. Paste a denser one and a number is invented. The reason is not reasoning — an image is chopped into patches and compressed into a fixed budget of tokens before the model reasons about anything, and detail finer than a patch is gone by then.
Read more →
