先玩，再选。Play first. Then choose.
三个候选实验室，各留一个最小而完整的玩法。试玩不影响正式实验室的收藏。Three playable candidates. Try the mechanism; your existing lab collections stay independent.
返回五个正式实验室Back to the five labs生息实验室Life Lab
补一格，然后让它自己演化。自主性候选。Add a cell. Let it evolve. An autonomy candidate.
只需要知道这件事One thing to know
每格只看周围八格：邻居太少或太多会熄灭；三个邻居让空格亮起，两三个邻居让活格留下。所有格子同时更新。边界外为空。它展示内在局部规则形成整体行为，不等同于自主性的全部内涵。Each cell reads eight neighbours. Three create life; two or three keep it alive. All update together, with empty space outside the board. This demonstrates internally driven local behaviour, one limited aspect of autonomy.
灵机实验室Agency Lab
你给它规则，它自己作决定。自主性候选。Give it rules. Let it decide. An autonomy candidate.
带两枚种子回家Bring two seeds home
先按试跑。它会捡种子，却可能忘记补能。改一条规则，再看它自己应对。Try a run. It can gather seeds, but may forget to recharge. Change a rule and watch its own response.
左下方是充能巢穴；实心块不可穿过。地图无需操作。Home and charger at lower left. Solid blocks cannot be crossed. The map needs no direct control.
它怎样作决定？How does it decide?
补能规则按当前能量与回巢路程判断；绕行规则寻找真实可走路径；携种规则决定是否回家。你修改的是内部回应，运行时无需指挥每一步。改规则会重开当前试玩。Recharge compares energy with the route home. Detouring finds a real traversable path. Returning decides what to do with a seed. You change internal responses, not individual moves. Changing a rule restarts this trial.
观象实验室Perspective Lab
相同的局部，放回全景再判断。元整体候选。The same close-up. A different whole. A meta-wholeness candidate.
让叶片重新舒展Let the leaves open
三幅近景看上去相同。拉远寻找原因，再把一幅图景放进窗口：可以拖，也可以先选图、再点窗口。Three identical close-ups. Zoom out for context, then drag a picture into the window—or select one and tap the window.
为何要拉远？Why zoom out?
近景的萎蔫相同，整体却可能缺水、积水或缺光。雨云增加水分并稍遮光，日光增加照明并蒸发水分，开窗只增加照明。模型只用于图景推理，不是种植或医学指导。它借鉴画框/尺度操作，原理表达仍是候选。The close-up looks the same, but the whole may show drought, excess water or darkness. Rain adds water and shades slightly; sun adds light and removes water; opening the window adds light. A fictional visual puzzle, not horticultural or medical guidance.
候选 demo · 本页不保存成绩。原创简化模型，供比较玩法；最终原理表达待选定后细化。Candidate demos · No results are saved. Original simplified models for comparing interactions; principle framing remains provisional.