问道 2.0 · 原型 – RenaSci / 新科学 @ AI Era

方向文件Program program.md

问题意识Question

或粘贴你自己的or paste your own

禁区与约束Limits 评价标准Criteria

回合上限Turn budget 模型Model

自带 API Key（可选，只存本机浏览器，原型不发送）Bring your own key (optional; stays in this browser; not sent in the prototype)

玩法Mode

难度（人在回路程度）Difficulty (human in the loop)

AI 队伍Party Claude

每个队员的每次工作都会写进披露记录：模型、来源、耗时、费用。Every member's work is logged for disclosure: model, sources, time, cost.

披露记录Disclosure 自动auto

Claude

0进展Progress

0证据Evidence

0风险Risks

0回合Turn

或者直接下指令Or give orders directly

存档与分支Saves & branches 点任一存档即回档，之后的回合成为旧分支click a save to rewind; later turns become an old branch

工作台沉淀Bench

观众席Audience 👀 0

提议进"气氛层"；你点"采纳"才进入下一回合指令。投票只影响推荐顺序。Proposals stay in the atmosphere layer until you adopt them; votes only reorder suggestions.

通关条件Completion

作者亲笔导读（≥300 字，发布必要条件）Author's own guide (≥300 chars, required)

0 / 300

文衡预评Wenheng preview 启发式heuristic

六维：严密性、创新性、扎实度、启发性、表达力、产业化。真实版由文衡评估器出分。Six axes: rigor, novelty, solidity, inspiration, expression, industrial potential. The real score comes from the Wenheng evaluator.