The AI isn't doing the problem solving. It is using the computer algebra system as a problem solver. This is similar to the way William Stein's Sagemath connects to other systems to supply answers. The key difference is that MCP is an open standard that can be used by anything, including LLMs.
The MCP idea is to set up a server that advertises the ability to answer mathematical questions and provide a client that provides the actual answers. The various computer algebra systems provide their own client. All of the LLMs can use MCP. They can know the services provided by the server (e.g. computer algebra) and will ask the server for answers. This is different than asking ChatGPT to provide answers. This is standards based and much more general than having a system like Hyperdoc. A new Hyperdoc would communicate with an MCP server to provide a "hyperdoc-like" page. This is the same argument used for HTML, JSON, and other general standards behind a browser. Tim On Fri, Jun 27, 2025 at 6:55 AM 'Nasser M. Abbasi' via FriCAS - computer algebra system <fricas-devel@googlegroups.com> wrote: > " Suddenly all of the AIs know how to do computer algebra." > > When chaptGPT 3.5 came out in 2022, I run one integration test file (MIT > integration problems) using > chaptgpt 3.5. At the time just to try it. > > Had to type each problem by hand one by one into the command and copy > the result. May be these days this can be automated with the new API's. > > But at the time chaptGPT did not do well. It scored lowest in percentage > solved. I am sure it has improved by now. > > Here is the HTML > <https://12000.org/my_notes/CAS_integration_tests/reports/summer_2022/test_cases/11_MIT/reportchapter1.htm> > (there is also PDF file) > > This was the result (percentage solved) > > Mathematica 99.38 > Fricas 96.57 > Maple 95.33 > Rubi 94.08 > Maxima 92.52 > Giac 91.59 > Mupad 90.03 > Sympy 82.24 > ChatGPT 14.33 > --Nasser > On Monday, June 23, 2025 at 5:25:20 PM UTC-5 axio...@gmail.com wrote: > >> It seems that it might take a day of hacking to connect to Claude AI >> or other AI systems. The clearest example with code is at >> https://www.youtube.com/watch?v=N3vHJcHBS-w&ab_channel=ShawTalebi >> >> So if you want to try to connect to an AI to answer integration questions >> this should provide sufficient background to get it running quickly. >> >> An MCP server could even be customized to read and format database >> information to answer questions about available domains, functions, >> and documentation. It could even be formatted similar to Hyperdoc output. >> >> Suddenly all of the AIs know how to do computer algebra. >> >> Tim >> >> >> On Saturday, June 21, 2025 at 12:50:14 AM UTC-4 cl...@jhcloos.com wrote: >> >>> so, obviouly my recall is not as good as it used to be... >>> >>> (and it has been a couple of decades since i did much with the server- >>> side fonts.....) >>> >>> so some corrections to my statements: >>> >>> after further review, i see that the 18pt adobe-courier strikes are in >>> the 75dpi/fonts.dir which (obviously) was not in this box's font path. >>> >>> and Type1/fonts.scale (also was not in my fp) has: >>> >>> cour.pfa -ibm-courier-medium-r-normal--0-0-0-0-m-0-iso8859-1 >>> courb.pfa -ibm-courier-bold-r-normal--0-0-0-0-m-0-iso8859-1 >>> courbi.pfa -ibm-courier-bold-i-normal--0-0-0-0-m-0-iso8859-1 >>> couri.pfa -ibm-courier-medium-i-normal--0-0-0-0-m-0-iso8859-1 >>> >>> which can fully scale. >>> >>> so perhaps the best change would be s/adobe/*/ so that the ibm-courier >>> pfb file get chosen when Type1 is in the fp, but the adobe-courier still >>> can get used when only bfd or pcf fonts are in the fp. >>> >>> i thought i recalled that xorg removed the type1 server-side renderer. >>> but perhaps that was only proposed and rejected? or reversed? >>> >>> too many decades. and two many strokes. >>> >>> -JimC >>> -- >>> James Cloos <cl...@jhcloos.com> >>> OpenPGP: https://jhcloos.com/0x997A9F17ED7DAEA6.asc >>> >> -- > You received this message because you are subscribed to a topic in the > Google Groups "FriCAS - computer algebra system" group. > To unsubscribe from this topic, visit > https://groups.google.com/d/topic/fricas-devel/mb_O0196RMQ/unsubscribe. > To unsubscribe from this group and all its topics, send an email to > fricas-devel+unsubscr...@googlegroups.com. > To view this discussion visit > https://groups.google.com/d/msgid/fricas-devel/0fafdeb1-f723-4f4b-8f80-f1f7758d8b05n%40googlegroups.com > <https://groups.google.com/d/msgid/fricas-devel/0fafdeb1-f723-4f4b-8f80-f1f7758d8b05n%40googlegroups.com?utm_medium=email&utm_source=footer> > . > -- You received this message because you are subscribed to the Google Groups "FriCAS - computer algebra system" group. To unsubscribe from this group and stop receiving emails from it, send an email to fricas-devel+unsubscr...@googlegroups.com. To view this discussion visit https://groups.google.com/d/msgid/fricas-devel/CAJn5L%3DKEHpmcJHqkz_SyHaGwyhwDghzbV_SDKA_0R7MExu8nRA%40mail.gmail.com.
