Branson admits Virgin space mission was going the wrong direction

Yes, he did, but he did with an architecture incapable of doing anything more impressive than win the X-Prize.

I'll give you another example from the self-driving X-prize.

There were always two teams that were very close to winning. One team hand-coded every decision tree for the particular path the car might encounter, and the other team used machine learning and let the car figure it.

Now as it happened in the self-driving challenge, the machine-learning car won, but it was really, really close. For the limited purpose of the challenge, the hand-coded car was nearly good enough. Maybe if they'd hand-coded it just a little better, that car would have won.

But hand-coding doesn't scale from a closed challenge track to real-world self-driving cars. Only machine learning can do that. But because of the artificial constraints on the challenge, the non-scaling solution almost won.

Well for the Ansari X-Prize, the non-scaling solution DID win. That's the Rutan solution, which used wings, which work for 1 passenger but (as Branson has spent 20 years and a billion dollars proving) doesn't even scale to 10 passenger joy-rides, let alone orbit.

The wings on SpaceShipOne were pretty light and flimsy because they didn't need to be any stronger considering how slow SpaceShipOne's top speed was. It topped out at Mach 3, and you need to reach Mach 25 to achieve orbit. Wings strong enough to handle Mach 25 would be too heavy to ever reach Mach 25; Catch 22!

Yes, and this has been well understood for decades. The basic math is simple: per unit mass, the potential energy change required to get from the Earth's surface to orbital altitude (say 200 km, or 200,000 meters) is

gh = 9.8 * 200,000 = about 2 million Joules/kg

whereas the kinetic energy change required to get from rest on the rotating Earth, a speed of about 300 m/s at the latitude of the Kennedy Space Center, to orbital speed, 8000 m/s, is

1/2 (v2^2 - v1^2) = 1/2 (8000^2 - 300^2) = about 32 million Joules/kg

So the change in velocity dominates the energy requirement, which drives everything else.

It's a measure of change in velocity (hence the delta… v…, and in mathematical terms is literally written with the symbol delta) required to reach or escape x orbit.

For example you need a delta v of approx 9.3 km/s to go from Earth to LEO. From LEO to GEO you'd need a delta v of roughly half of that. Engineers in this sense would talk about a total delta v budget of ~13 km/s and some change to achieve Earth to GEO.

The question is then how do you design a rocket that can produce those numbers? Higher delta v requirements means higher mass requirements. You'd need more propellant to produce that thrust, and more structure to house it and hardware to control it. At some point it simply doesn't work anymore because it becomes counter intuitive to keep adding more mass.

Enter the concept of staging. Staging allows you to discard your useless mass after it's done it's job. For example you might need 9.3 km/s to achieve LEO, but you might also know it's not a linear relationship so that you need more of that upfront and less of it later. In that case, rocket engineers simply designed around that and created heavy first lift stages to produce the bulk of that delta v, and then discarding the heavy structure and hardware after the propellant has been exhausted.

I would suggest SpaceX freeze a design every 5-7 years that's good enough to scale while cheaper than it's completion and then have a more experimental path. If option A costs 20 million more, but has a 98+% success rate and option B saves 20 million then the market can chose which to focus on.

The real 'problem' spaceX is facing is humans simply don't send all that much stuff into space, so they are never get enough data get the experimental path stable.