The system you propose is non-sensical

This is how Singapore does it. Just because you don't know doesn't mean it's non-sensical.

Singapore famously does precisely what I suggested: they build state capacity by hiring the best civil servants and paying them accordingly.

And looking into it some, they themselves certainly agree on their digital services material that their government agencies need to be competent to oversee the work, including by having competent people. They don't just hire IBM and step away.

you need to fix the government

Exactly what I propose: you fix government by fixing its structure and incentives.

In either situation you must fix government to not be corrupt. We both agree on this.

Where we differ is that the system you propose is harder to remove corruption from. It is also frequently less efficient, not more efficient.

You use your personal experience being from a country with moderate corruption. Allow me to use my personal experience of having been the government representative overseeing contracted IT system development.

The primary obstacle is not that civil servants are good guys and contractors are bad guys. The primary obstacle is that it is hard for anybody to specify what to build in the first place (which popped up in this subreddit yesterday).

By introducing more steps between the solution developers and the end users (which fully offloading the work inherently does) you end up with situations where the users and government agencies get a bad product, if they get one at all.

And even where the contractor wants to make a better product (and I've worked with more than a few willing to flex in service of a superior deliverable), they get told they can't because they have to deliver what they were originally contracted to deliver.

The people who wrote that contractual requirement were government staff who don't understand software, and sometimes don't even understand the requesting agency because they are in a specialized contracting shop that does nothing but IT acquisition.

Stepping outside of government for a bit, there was a trend in the 90s whereby private sector companies would "offshore IT"... all of it. Help desk, software dev, anything that could be done by someone in India for 10% of the price, would be.

By the 2000s most organizations had to revert most (though not all) of those changes, because they kept running into the difficulty of how do you build your core business software when the people with 'hands on keyboard' don't actually understand your business? You can't!

But while the private sector was agile enough to offshore and then reverse that offshoring, the U.S. government at least "offshored" its digital functions to contractors in the 90s, but was never able to revert it. That led to frequent disasters like healthcare.gov, DIMHRS, the U.S. Navy's ongoing HR ERP disaster, and a multitude besides.

Conversely, things that worked, like IRS DirectFile and the USPS mass COVID test mailings, were done by government agencies with competent IT teams that were allowed to use their competence. And they still had contractors involved too, there's simply not enough gov't staff to go round, but they didn't simply try to offload everything to an IBM.