>The Sandboxed Payload

You can't refuse to run code your state depends on. You also can't endorse it. The compilers sandboxed the art of killing — evaluated it without granting it the keys.

Confucius taught benevolence (仁). Mencius argued a truly benevolent ruler would have no enemies, governing by moral example with no need of the sword. At its most idealistic the Confucian tradition holds that virtue makes violence unnecessary.

The Siku Quanshu Zongmu Tiyao was compiled by men who passed the imperial examinations by mastering exactly this tradition. When Ji Yun and his team reached the military section (兵家類) in juan 99–100, they were confronting a body of literature that exists precisely because virtue alone does not keep the borders secure. The preamble is one of the most philosophically revealing passages in the 200-volume catalog, because the compilers have to explain — to themselves as much as anyone — why a library built on Confucian principles should preserve texts on deception, ambush, and the strategic uses of cruelty.

The compilers open with history. Military texts have existed since antiquity — the Rites of Zhou names a Minister of War, the Annals record constant warfare, even the Yijing contains hexagrams about armies. War is not an aberration in the classical tradition; it's part of the furniture. Then comes the distinction that lets them breathe: 其間孤虛、王相之說，雜以陰陽五行...故兵家恒與術數相出入，要非古兵法也 — military writing has always overlapped with the occult arts of Void and Fullness, Yin-Yang, omen-reading, "but these are not the ancient art of war."

That's the boundary. There's a legitimate military tradition — "the arts of raising and training troops, and the proper use of strategy and tactics" (生聚訓練之術，權謀運用之宜) — and an illegitimate one that mixes strategy with divination and cosmological numerology. Strategy runs in the main catalog. Divination is shipped to the occult-arts section. The dangerous payload executes, but inside a defined boundary.

The compilers' highest admiration went to texts that treated war in morally serious terms. Sun Tzu (孫子) received the top verdict: "the ancestor of a hundred generations of military writing" (武書為百代談兵之祖). What they admired was philosophical depth — Sun Tzu's insistence that the highest victory needs no fighting, 不戰而屈人之兵，善之善者也, "to subdue the enemy without battle is the supreme excellence." A proposition a Confucian can endorse: war is necessary but regrettable; the best general makes it unnecessary.

Wu Qi (吳子) posed a harder case. The man was morally appalling — he killed his wife to secure a command — and "his personal conduct is scarcely worth discussing" (其行事殊不足道). But his text emphasizes virtue over terrain ("it lies in virtue, not in strategic ground" — 在德不在險) and governing troops through ritual and righteousness. The verdict is the compilers at their most honest: a terrible person can produce a morally serious text, because ideas are not hostages to their author's character.

The text most congenial to their worldview was the Sima Fa (司馬法): "grounded in the Way, reliant on virtue, rooted in benevolence, founded in righteousness" (據道依德，本仁祖義) — Confucian moral vocabulary applied to how armies should be organized. The Han historian Ban Gu classified it not under military texts but under ritual (禮), which delighted the compilers. It preserves "one in a thousand" of the military regulations of the Three Dynasties — a time when war was governed by ritual rules, protocols for declaring hostilities and treating prisoners. Whether the Three Dynasties actually observed such rules is historically dubious; what matters is the compilers wanted it true. The Sima Fa gave them a text proving military affairs could be conducted within a moral framework.

If they praised moral seriousness, they condemned texts mixing strategy with supernatural claims. The cunmu (存目) is a catalogue of contempt — text after text dismissed for contaminating military thinking with divination, astrological timing, and spirit calculations. The late Ming drew special venom: "especially squalid and chaotic" (尤為猥雜).

The objection was specific. Confucianism holds that human affairs should be governed by human judgment — virtue, wisdom, experience. A general who consults the stars before deciding to attack is abdicating responsibility, substituting mechanical ritual for genuine strategic thought. In modern terms, what they objected to was the replacement of judgment with algorithm. The wooden men with firecrackers were laughable; the astrological timing systems were dangerous, because they encouraged generals to base life-and-death decisions on numerology rather than actual conditions.

Between praise and condemnation lay works that gave genuine difficulty. Wei Liaozi (尉繚子) is the most interesting. On one hand, principles any Confucian could endorse: "do not attack a city that has committed no offense; do not kill a person who has committed no crime" (兵不攻無過之城，不殺無罪之人) — standards the Warring States theorists "did not speak of." On the other, the Bingling chapter details desertion laws, collective punishment, and discipline enforced through terror. The compilers note this without flinching: Wei Liaozi is "not empty talk of benevolence and righteousness by those without practical experience" (亦非漫無經略，高談仁義者矣). A text that merely repeats Confucian platitudes is less valuable than one grappling with the ugly realities of discipline. Practical wisdom requires both moral seriousness and moral compromise.

The sole Ming general to earn full respect was Qi Jiguang (戚繼光), whose Lianbing Shiji (練兵實紀) and Jixiao Xinshu (紀效新書) drew admiring reviews. What they valued was a man who lived by his words. His famous passage on formation combat — "when a thousand men advance in formation, the brave cannot rush ahead, the timid cannot hang back; everyone simply pushes forward together" — proved he "truly understands the situation" (可謂深明形勢) and didn't deal in "stale phrases from strategy manuals." They noted, with approval, that he beheaded his own eldest son for looking back during battle: "he can be said to have lived up to his own words" (可謂不愧所言矣). The story is appalling by modern standards; for the compilers, it proved Qi Jiguang was genuine.

They never resolved the tension between Confucian benevolence and military necessity — no one has. But they found a working position: evaluate war texts the same way you evaluate any text — by the quality of the thinking and the honesty of the engagement with reality. The library was not about endorsement. It was about intelligence.