On 10/29/2018 05:28, Noel Chiappa via cctalk wrote: My best recollection as a distant observer, for the neural/cognitive modeling it was conceived for, there weren't enough labs with enough grant money to support the company. And as part of the "general" supercomputer segment there were limited problems you could address with the original 1-bit CPU model, and producing code for it was a big challenge coming from more traditional systems. A lot of that may have been alleviated with the later designs built on mainstream (SPARC, Alpha) CPUs, but by then I think there were more alternatives chasing those dollars. And the company's, er, management issues were not helping. Some links for further reading: "Rise and Fall of Thinking Machines" - https://www.inc.com/magazine/19950915/2622.html MIT CSAIL "Final Report" - http://courses.csail.mit.edu/6.972/TMC%20Corp.html That's an interesting question. The same challenges for other big iron come into play, including critical dependence on a FEP from Symbolics, DEC, Sun, HP, etc. --S.