My idealistic part says that a combination of AI-driven technical orchestration (much more than just coding) and orbital/langrange manufacturing facilities could, perhaps, get somewhere in the not ridiculously distant future (centuries rather than millenia)

A more pragmatic me would point out that the required energy and materials needed would mean we would need breakthroughs in space-based solar capture and mining, but this is still not New Physics.

I think the solution will come from exponentially advancing self-assembling machines in space. These can start small and, given the diminishing cost of getting things to space, some early iterations of the first generation could be mere decades away. There are several interesting avenues for self-assembling machines that are way past napkin-sketch phase. Solar arrays are getting bigger and we have already retrieved the first material from an asteroid.

The quality and reliability of AI agents for processes orchestration and technical reflection is now at a stage where it can begin to self-optimise, so even without (EDIT) a "take-off" scenario, these machines can massively outperform people in manufacturing orchestration, and I would say we are only some years from having tools that are good enough for much larger scale (i.e. planetary) operations.

Putting humans there is a whole other story. We are so fragile and evolved to live on Earth. Unsurprisingly, this biological tether doesn't get much of a look-in here. Just being on the ISS is horrible for a person's physiology and, I am guessing there would be a whole host of space sicknesses that would set in after a few years up there or elsewhere. Unless we find a way to modify our biology enough so we can continually tolerate or cure these ailments, and develop cryo-sleep, we're probably staying local - both of these are much more speculative that everything above, as far as i can tell.