上屋抽梯
Draw someone into a position they can't retreat from, then take away the way back. Commit them past the point of return — an enemy into a trap, an ally into a decision, or your own people into fighting with no line of escape.
The principle
People fight and commit hardest when retreat is impossible; and people also blunder when an easy-looking advance quietly closes behind them. Both are this stratagem. Aimed at an enemy, you bait them forward with an apparent opening — undefended ground, an easy prize — and once they're committed, you spring the trap and cut the return. Aimed at your own side, you deliberately remove the fallback so hesitation is no longer an option and everyone commits fully. Aimed at a reluctant partner, you engineer a situation where the only way out is through the decision you wanted. The ladder is the bait that makes the climb look safe; removing it is what converts position into commitment.
Historical origin
Applied today
To end endless hedging on a hard call, remove the fallback: set a public deadline, commit the budget, retire the old system on a date. Once retreat is off the table, the team stops debating and starts making it work.
Watch for the deal that lures you in with an easy start, then removes your alternatives — the free integration that becomes lock-in, the platform you build on that later changes the terms. The easy ladder up is often the trap.
In practice
Invented scenarios showing this stratagem at work — illustrations, not reports of real events.
Eleven variants, no export
Training credits were free for the first year and the tooling was pleasant, so by the second year eleven fine-tuned variants sat inside production workflows. The weights were not exportable and the evaluation history lived in the provider's console. The new pricing was not unreasonable in itself. It simply arrived once the only alternative was rebuilding a year of work.
Use it Before fine-tuning, confirm you can export weights and evaluation history — and keep the prompts and datasets that produced them.
The ladder was taken first
The encryption began at two in the morning, but the decisive work had been done four days earlier: backup jobs quietly reconfigured to fail, and retained copies deleted from the same console that managed them. By the time anyone read a ransom note, the way down had already been removed, and the newest surviving copy was eleven weeks old.
Use it Keep one backup copy offline and immutable, and alert on backup deletion — recovery is what gets attacked first.
After everyone was upstairs
The free tier was unusually generous for two years, long enough for the team to build reporting, alerting and three internal workflows on top of it. The pricing change arrived afterwards. The cost of leaving was no longer the subscription; it was the four systems that had been built assuming the ladder would stay.
Use it Before you build on something free, price your exit — and keep an export path you have actually tested.
Starting before the ink
The client was enthusiastic, the start date was tight, and the agency began work on a verbal go-ahead while the paperwork was finished. Three weeks in, with two people staffed and a launch already announced internally, the contract came back with payment terms doubled and the scope broadened. Refusing had become expensive. The climb was encouraged; the ladder came away afterwards.
Use it Do no billable work before signature — every week of unpaid commitment weakens the terms you can still afford to refuse.