Reserve Tree-of-Thoughts for high-stakes problems — it's 74% vs 4% but 10–50× the tokens

SurePrompts·medium signal

Tree-of-Thoughts solved 74% of Game-of-24 problems versus 4% for plain chain-of-thought, a dramatic gain on search-style problems with many viable branches. But it costs 10–50× more tokens, so the actionable rule is selectivity: deploy ToT only on genuinely high-stakes, branch-heavy tasks and keep cheaper CoT as the default.