Everyone has a workflow for a given process. Most astronomers it's work frantically against a deadline and collapse. I'm not a fan.
For literature resources, you also need a workflow. The smoother the better. The less you have to think about where that paper was or how to organize your references, the better. It makes it much much easier to write the thing.
So here is a short intro how *I* organized my literature. Eventually.
It began with Ron Allen suggesting I'd pick a system. He was vague as to what system. His system involved a giant drawer in his office with physical folders. I got started that way. It devolved into a mess.
A better way is a reference manager some kind. I settled on bibdesk fairly quickly (this is back in 2000) but I made a new .bib bibtex file for each paper/project I was working on. This also devolved into a mess.
I need one file. One BibTex file to rule them all.
So I made one. Bibliography.bib.
That's it...that's...my bibliography file. I spent a day or so merging all the files and weeding out all the duplicates. Set up so that pdfs of the papers automatically go to a folder. Name+ last two digits of the year as the identifier and it works? It lives in my dropbox so I always have it?
I was quite happy with that for a quite some time. It's where I'd dump new astro-ph papers in (for later reading hahahah anyway) and so often I'd have the astro-ph already or I could refer to the most recent work.
And then I moved to Overleaf. Aaaah. I *love* it. Version control? No problem! Every project tagged and searchable.
But How To Do The BibTex?
Here comes my second to last trick: I have one project called Up To Date Bibliography Here.
Yup. I always upload my bibtex there. As a rule I export it in bib desk as "minimal bibtex" so that it's small. And then I can import or refresh it in the project I'm working on.
And the last tip: there is a command line tool called ads2bibdesk and when I call it with an astroph number of an article identifier, it loads everything in...it's done.
At the end of every paper there is just: