This is very similar to Doug Lenat's work on Automated Mathematician & later on Eurisko, and later Ken Haase's follow up work on representation languages.<p><a href="http://oai.dtic.mil/oai/oai?verb=getRecord&metadataPrefix=html&identifier=ADA155378" rel="nofollow">http://oai.dtic.mil/oai/oai?verb=getRecord&metadataPrefix=ht...</a><p>There were severe sticking points around the cultivation of an idea of "interesting" properties and the performance issues around evaluating a combinatoric space of possible manipulations. There hasn't been serious work along those lines since the early 90s or so.<p>It's annoying because especially Haase's work has some very practical insights, but Wolfram seems to be loathe to ever admit he's building off of someone else's work.