I pasted a subset of the Fil-C source code into your tool and it says 6 person years. I just pasted the compiler pass and the obvious parts of the runtime.
Note that I started the project in Nov 2023 and can only work on it maybe 1-2 hours a day because it's just a side project.
So I think your tool either estimates based on very bad programmers, or it's just wrong. Or maybe 10x programmers are real and I am him
These metrics necessarily have to underestimate programmer skills because those are not directly controllable. If there is any sort of rigor in these metrics (i.e. I don't know if COCOMO is one of them) they will probably assume, say, a mundane programmer whose performance is worse than 90/95/99% of all other programmers.
Here's more about the COCOMO model it uses: https://dwheeler.com/sloccount/sloccount.html#cocomo
Sounds like nonsensical pseudoscience
I don't take those results very seriously myself, but have you seen anything better?
No
To me this is a case where knowing that you don't have data is better than having data and pretending it means anything