Your confusion is warranted, the system has twisted incentives. I am a scientist working at a rich university. I still mostly use scihub unless the article is on arxiv, even for articles which have my name in the byline. All this to say, I consider pirating these papers the moral highground and you should not feel bad doing it. Thankfully, I have the job security and funding to demand my work be open access (a whole other can of worms).

i think a possible startup idea would be to start a "Github for science" (replace "science" with any academic or interest area really). Encourage folks to post their papers in latex or markdown. Post their data, code, mechanical and electrical drawings, videos and pictures of their experimental setup, videos explaining their research, videos performing the experiment?, and so on. Allow others to comment on research with questions, concerns, suggestions for enhancement, or link it to their own supporting work, etc. Allow other researchers to fork people's research (but all forks link back to the original to give full attribution). The core focus is open up access to academic papers and research, foster better communication between people, and set a higher bar for communication of research.

Going the non-profit route: I think it would be cool for the site to be federated/distributed so universities & countries could host replicated nodes. It would be best if somehow the software system development and hosting costs could be underwritten by governments and universities.

Going the for-profit route: The site could have an inexpensive PRO account to help support the site and software development. The site could be add supported, but zero tracker, with adds that would be small and content-based purely on the content of the paper topic.

Add Examples: Reading an article on Julia numerical programming... maybe show an add for the upcoming JuliaCon or Julia Computing

Reading a metallurgy analysis paper... maybe show a small add for an x-ray fluorescent system for metal composition analysis or advertise an upcoming metallurgy conference

... the idea being that these adds are (1) not targeted at users, just targeted at content, (2)small, dismiss-able, and not annoying (3) hopefully actually useful and interesting to the reader.

The site could also sell "advanced API access" to interested parties for knowledge management/search on fast trending research and topical areas, research community graph analysis (maybe for recruiting?)... kind of the LinkedIn layer of the site.

The problems with this are overcoming the startup inertia and coming to a common agreement about how to implement it.

In my view, the latter is by far the most difficult and failing to come up with a common objective or view is likely to either scuttle the project or render it ineffective or inefficient.

Let me give you a real world example: Linux. We've so many distros that it's hard for anyone of them to get real traction. As we've seen for decades, Linux has never managed to take on Windows on the desktop for the reason that there are so many disparate views about how tackle the problem. Both businesses and ordinary users want certainty not constant variation.

We see the same problem in other areas of the net, for instance copyright reform. The only way to tackle that is on a worldwide basis - as the forces against it are aldo already very well organized on a worldwide basis and have been so for some 140 years (since the original Berne Convention).

I've often thought that we need a worldwide union of computer users to tackle these issues and take on Big Tech, etc. but how does one start something as big and significant as this let alone hold it altogether for any length of time?

I wish I knew.

Just use sci-hub. It always works, and it's very easy. There's no good reason to give the journal your money to read an article they had no hand in writing or funding.