Sentient Engineering Code Breaking
It is considered 'possible' (if impractical) for a being to give enough information about itself over a period of time that its thought processes can be modeled solely from its external actions.
In some certain circumstances (cf - the Llewellyn affair, Gupta 3 bureaucracy scandal, etc) this has been proven to be a successful if truly involved method to gather initial cryptographic keys and similar critical information. This usually (as in the Llewellyn affair) involves transapient intellect(s), but occasionally (as in the Gupta 3 incident) merely requires a large number of baselines working efficiently together.
In each case, a model of the targeted being(s) is developed and refined over a period of weeks to years, and then this model is utilized in predicting future actions of the sentient. If the predictions appear highly accurate to the producers (which may take multiple iterations to 'get it right') the model can then be used to begin predicting thought processes, often involving prediction of specialized information - when a conceptual breakthrough may occur, what a specific passcode might be, etc.
Text by John B (inspired by Dr William Sims Bainbridge's work on 'personality capture')
Initially published on 31 December 2007.