I presume what you're trying to do is implement a user-friendly interface for implementing M of N transactions
Yes. But on the tech side it might also involve some "smart contracts" features: transaction replacement (emulation), nLockTime etc.
Are you thinking of going further and investigating an implementation of smart property (http://en.bitcoin.it/wiki/Smart_Property), or at least an interface where smart property contracts can be registered and an API the smart property itself can interact with?
It is more aligned with what BitcoinX does, as smart property can be implemented using colored coin software. But, yes, I want to experiment with this one way or another.
The outcome of some M of N contracts could be actually reduced to algorithms (e.g. betting on the value of MtGox_7day_USD which, I suppose, can be obtained programatically) - would you implement an interface where the writer of the contract can specify the algorithm which decided the outcome?
I don't think it is a good model. Outcome can indeed be controlled by algorithms, but it all depends on who will run such algorithms, i.e. you need to trust the service which does this.
I think at this point it is better to run derivative/prediction markets as a service, where service operator will determine which contracts are in demand, will standardize such contracts and will control settlements. BitContracts.org might create software for such prediction market operators, but it won't run anything itself, since running such exchanges involves a lot of risks.
As for flexible custom contracts, that's certainly interesting, but I doubt there is a lot of demand for that, compared to demand for, say, standardized BTC/USD forwards.
But certainly such flexible contracts might align well with business model of a prediction market operator, so there is certainly a possibility that it will be implemented in future.
I wonder which type of language that would use.
Oh, I can answer this question
. I'm a fan of Lisp family of programming languages, and it is really easy to implement a restricted version of Lisp which will be safe and deterministic (but not Turing-complete). Of course, some things need to be provided as fixed functionality, for example, access to MtGox prices. Letting algorithm to do networking on its own isn't a good idea I'd be interested in such a project, but I think it will be quite some time before there would be any real world example of it being necessary. I mean, bitcoin would have to take off before these complex features of bitcoins will be required, I think.
Yes, but people already use BTC/USD CFDs and futures.