[baka367]

Repositories are rarely worth much.

Sure, some algorithms there might save you some time, but its often the design and the data where the money lies (what this guy focused on).

Clone google's repo and you'll likely struggle forever to get anything of substance running on a rando vm/docker/etc. not to mention about spinning the entire stack with interconnected services, certificates, shitty code, and layers upon layers of hacking that can only be resolved by relying on the tribal knowledge on whomever built the darn thing.

Compared to that - detailed design docs, a team of motivated Chinese dudes/ettes with some monetary support from the local party, and you can have a close-enough copy running natively on the Alibaba cloud in a few months.

[xyzzyz]

Source code repo is like a very extremely detailed doc. You might not be able to actually easily run it due to all of the dependencies etc, but with couple of weeks of reading, you should be able to tease back out the high level design.

[avidiax]

I've done enough code archaeology to say that looking at the code to understand the design is a good way to understand that the two halves of the bridge didn't mate up, but there was a deadline, so...

The design from a design doc can be replicated at almost any company. The actual code is specific to the company and their exact stack.

The company's business position is similarly hard to duplicate. You can understand a company's current capital, customers and money flows. Your new company has to either outcompete for those same flows or create or capture alternative flows, and do this with different capital. Having, say, the entire source code for FedEx doesn't make it easy to launch a competitor. It's practically irrelevant compared to the network of capital investments, corporate goodwill and contracts, etc.

[fragmede]

A copy of Google3 would take an outsider eons to replicate Borg for any of it to run on.

[paulddraper]

There's probably some deep science AI-type stuff.

Or maybe useful for security exploits.