The following is much more useful, and covers many more cities across the world:
Most people aren’t going to stay in Greenwich village/SOHO/UES if they are in NY, or similar elsewhere. Prices reflected in the MIT calculator seem to take all areas of living into account, that too by county.
Edit: Numbers on numbeo seem to have improved a lot since the last time I saw them. Taking back my previous opinion, but I find both to be useful resources.
You can change the assumptions in numbeo too, like for example if you plan to live outside the city center.
It's partly due to their data collecting methodology. They rely on crowd-sourced user inputs and manually collected data. I also have doubt about their analysis methodology, it's seems very ad hoc and amateurish.
You should treat Numbeo with a grain of salt. It's honest attempt with small budget and better than nothing, but you can't treat it as authoritative source.
What happens at gitlab if you start off in SF then move to another city? You're a remote employee and are free to work wherever... Do they then demote your pay to reflect the new city or did you just game the system?
What about the opposite case, a worker in India at gitlab moving to SF did they just lose a lot of money?
In general this cost of living approach to remote employment is a cop out to pay lower salaries and save money. Zapier offers bay area salaries to all remote employees and gitlab discriminates... Which would you want to work for?