This blog post discusses testing the capabilities of two AI models, Codex and Claude, in solving a complex text layout issue specific to Japanese vertical writing. It highlights the performance disparity between the two models, with Codex underperforming significantly compared to Fable's solution.