> Generally speaking I agree with you.  I even have a research project I’m trying to start on a related topic (to allow the sharing of a codebase by programmers with different language preferences… email me privately if you want to know more).

Sure! I will.

> The point I am less convinced with is that adding static typing to Scheme is the best next step towards this goal.

Would you suggest first interfacing to other dynamic languages? That's a
fine goal, but it gives us less information about the more difficult
aspects of the work. I feel it would be beneficial to tackle some hard
problems early.

> Also I see a problem with adoption by the community of a “universal language” that can express (most of) what can be expressed with existing languages. I expect the performance to suffer (because implementing a single specific language can take advantage of the missing features to enable various optimizations). So for the first 10-20 years it takes to develop and implement an efficient compiler for the universal language, few people will see a benefit in using it. “Worse is better”…

Existing compilers aren't going to disappear; people can still use them.

A key feature of the proposed unified model is that implementations
should make it easy to take subsets of it. It should be possible to take
a lexical block of code, and say what features are enabled for that
block. This is the key to code re-use, optimization, automated
translation, and language-based security. The full model will be
difficult to reason about (though still possible). Subsets will make it
easier.

The security field has the principle of least privilege. High quality
production code should have a similar "principle of least language
features".

> Your vision may be “the right thing”, but you may be 100 years ahead of the times. [I’m not saying the direction is bad, just that the road may be very lonely]

It is lonely now :) I hope it will be less so in the coming years.

According to
https://en.wikipedia.org/wiki/Standard_Model#Historical_background the
standard model of physics took about 20 years to put together
(1954-1974) once most of the ingredients were in place. We have most of
the ingredients for the programming language model so a similar 20 year
timeline might work.