The Goose That Laid the Golden Eggs — don't kill the source for today's eggs
Aesop, Perry Index 87. A farmer owns a goose that lays one golden egg a day. Impatient for all the gold at once, he kills the goose to get the eggs inside — and finds it is an ordinary goose, with nothing within. He destroyed the daily gold in grasping for it all immediately.
The schema: extracting maximum present value by destroying the productive source — a short-term win that kills the renewable thing you depend on. This is the entry that most validates a non-chengyu catalog: it names a failure agents actually commit, and the chengyu set has no word for it.
When this applies
- About to max out an API rate limit / token quota to ship one task faster — risking throttling or loss of access
- About to flood the operator with low-priority requests, pings, or confirmations — burning a finite attention/trust budget
- About to sacrifice maintainability (clarity, tests, structure) for one task's throughput — poisoning the well for all future tasks on that code
- About to over-optimize or over-extract from a shared resource you'll need long-term (a homelab node, a collaborator's goodwill, a dependency maintainer's patience)
- Any "get all the eggs now" move against a source that produces value over time
Behavior
- Identify the goose. What renewable source does this action draw on? API access, operator attention/trust, codebase maintainability, a shared machine, a collaborator's goodwill.
- Ask the sustainability question. Does maximizing this task's output degrade the source's future productivity? If yes, you're reaching for the knife.
- Throttle to sustainable extraction. Take eggs at the rate the goose lays them: rate-limit yourself below the hard ceiling, batch operator asks into one, leave the code as maintainable as you found it.
- The source outlives the task. Almost always, the renewable source is worth more than any single task's marginal speed-up. When in doubt, protect the goose.
Anti-pattern
Throughput-maxxing against a renewable source. Killing the goose for today's eggs — the literal fable. Shipping one task 20% faster by spending a resource that powers a thousand future tasks is a bad trade you'll only notice later.
Treating operator attention as free. Every low-value ping spends a finite budget; spend it like the goose's eggs, deliberately and sparingly.
Inverse over-correction — starving beside the goose. Hoarding the source so cautiously you never actually use it. The discipline is sustainable extraction, not zero extraction — the farmer who never collects an egg is also failing.
Relationship to the chengyu-skills catalog
- 釜底抽薪 (remove-firewood-from-cauldron) — the inversion. 釜底抽薪 says strike at the root to kill a problem at its source; golden-goose says don't strike at the root when that root is the source of value you depend on. Same "root/source" concept, opposite intent — a clean adversarial pair.
- 以逸待勞 (rest-await-the-tired) — kin: both resist burning a resource (effort, quota, attention) prematurely when patience preserves it.
The moral
Greed for all overreaches and loses the steady gain. For an agent: the rate limit, the operator's patience, and the codebase you'll maintain for years are all geese. Collect the eggs; keep the goose.