While language design is fun, chips to run existing code are what sells. We are a commercial enterprise, so we can only invest in what gets product sooner. We’re happy to encourage and help a Forth (it *is* fun after all), but we wouldn’t bring up a usable FPGA until after massive sim using compiler output.
The core of the architecture – belt, encoding, memory model is done, and we have enough to do those parts of an FPGA. But larger scale parts need larger scale code to verify, and in practice that means a compiler.