Re: 6 months free Claude Max 20x for working on Maxima

--===============4176698941868649853==
Content-Type: multipart/alternative;
 boundary=----2TH0TYGRCHGHMP3JFN2U6Q47S6643L
Content-Transfer-Encoding: 7bit

------2TH0TYGRCHGHMP3JFN2U6Q47S6643L
Content-Type: text/plain;
 charset=utf-8
Content-Transfer-Encoding: quoted-printable

Two months ago I have asked Claude to review the architecture of wxMaxima=
=2E It has split the large files into small functional groups and I thought=
 that is for readability purposes=2E But then it identified invariants and =
wrote test bench entries for it, identified which places aren't hit by the =
test suite and added tests for them, went through the git log and identifie=
d what bugs are recurring - and after everything had a test started on chan=
ging the architecture so it avoids the bugs in the first place=2E When it f=
inally was sure it understands the code it even added cool documentation=2E

Then I asked it to do an efficiency review=2E It changed many places that =
were O(n=C2=B2) complete to loops that are O(n) complete=2E It told me it o=
nly dares to do so because all the tests are already in place and if one of=
 those changes didn't come with a test it implemented it=2E=20

One other non-mathematical thing would be asking it to clean up the build =
dependencies so incremental builds actually yield the result one expects an=
d don't forget dependencies or hit circular ones=2E=20

We could ask it to clean up our mess with global variables as a first step=
 towards multi-threading=2E=20

We could ask it to perform a stability review=2E

I don't know how good it is in maths=2E But when it can change stuff so it=
 is testable, add tests and then test the results of its work it really exc=
els=2E Or we could ask it what needs to be done so it can port Rubi=2E It n=
ormally generates a structured plan for such things, maps dependencies of s=
tuff and even if we don't put that plan in action we will learn something w=
hen asking it such questions=2E

Kind regards, =20

   Gunter=2E


On 21 July 2026 15:04:03 CEST, Michel Talon <talon@lpthe=2Ejussieu=2Efr> w=
rote:
>Hello,
>
>having been a user of the grobner package, i think that it seems to work =
OK as it is=2E Maybe it is slow, there are=C2=A0 faster algorithms implemen=
ted elsewhere notably the F4 and F5 Faug=C3=A8re algorithms in Maple and Ma=
gma, but it is a limitation of the concept that for a complicated enough pr=
oblem, the grobner computation will take "infinite" time=2E Also the grobne=
r basis obtained will depend of the ordering of the polynomial set, so chan=
ce is a factor here=2E
>
>In fact the author Rychlick has written an extension of his code, long ag=
o, introducing colored variables and such stuff which are useful for solvin=
g algebraically problems in geometry, a well known application of this busi=
ness=2E At the time i took a copy of this program, you can take it at:
>
>https://www=2Elpthe=2Ejussieu=2Efr/~talon/rychlik_CGBLisp=2Etgz
>
>There is another implementation of grobner basis in maxima, that can be f=
ound in share/affine, due to the late William Shelter=2E A long long time a=
go it seemed to work, slower than the Rychlick implementation, but since th=
e package affine was written for doing algebraic geometry on an affine map =
of an algebraic variety, for non commutative variables, it is possible it w=
as working for non commutative polynomials=2E However when i have tried to =
play with this stuff more recently almost nothing worked and there is essen=
tially zero documentation=2E=C2=A0 Maybe this is extremely interesting but =
short of understanding all the code, which probably requires a lot of work =
done by specialists of the subject=2E
>
>
>Le 21/07/2026 =C3=A0 02:44, Robert Dodier a =C3=A9crit=C2=A0:
>> (1b) as part of that, maybe fix bugs, extend, unify, and update the
>> existing stuff for Groebner bases=2E I don't know where that stands at
>> present=2E
>
>--=20
>Michel Talon

------2TH0TYGRCHGHMP3JFN2U6Q47S6643L
Content-Type: text/html;
 charset=utf-8
Content-Transfer-Encoding: quoted-printable

<!doctype html>
<html>
  <head>
    <meta http-equiv=3D"Content-Type" content=3D"text/html; charset=3DUTF-=
8">
  </head>
  <body><div dir=3D"auto">Two months ago I have asked Claude to review the=
 architecture of wxMaxima=2E It has split the large files into small functi=
onal groups and I thought that is for readability purposes=2E But then it i=
dentified invariants and wrote test bench entries for it, identified which =
places aren't hit by the test suite and added tests for them, went through =
the git log and identified what bugs are recurring - and after everything h=
ad a test started on changing the architecture so it avoids the bugs in the=
 first place=2E When it finally was sure it understands the code it even ad=
ded cool documentation=2E<br><br>Then I asked it to do an efficiency review=
=2E It changed many places that were O(n=C2=B2) complete to loops that are =
O(n) complete=2E It told me it only dares to do so because all the tests ar=
e already in place and if one of those changes didn't come with a test it i=
mplemented it=2E <br><br>One other non-mathematical thing would be asking i=
t to clean up the build dependencies so incremental builds actually yield t=
he result one expects and don't forget dependencies or hit circular ones=2E=
 <br><br>We could ask it to clean up our mess with global variables as a fi=
rst step towards multi-threading=2E <br><br>We could ask it to perform a st=
ability review=2E<br><br>I don't know how good it is in maths=2E But when i=
t can change stuff so it is testable, add tests and then test the results o=
f its work it really excels=2E Or we could ask it what needs to be done so =
it can port Rubi=2E It normally generates a structured plan for such things=
, maps dependencies of stuff and even if we don't put that plan in action w=
e will learn something when asking it such questions=2E<br><br>Kind regards=
,=C2=A0 <br><br>=C2=A0=C2=A0 Gunter=2E<br></div><br><br><div class=3D"gmail=
_quote"><div dir=3D"auto">On 21 July 2026 15:04:03 CEST, Michel Talon &lt;t=
alon@lpthe=2Ejussieu=2Efr&gt; wrote:</div><blockquote class=3D"gmail_quote"=
 style=3D"margin: 0pt 0pt 0pt 0=2E8ex; border-left: 1px solid rgb(204, 204,=
 204); padding-left: 1ex;">

    <p><font size=3D"4" face=3D"monospace">Hello,</font></p>
    <p><font size=3D"4" face=3D"monospace">having been a user of the grobn=
er
        package, i think that it seems to work OK as it is=2E Maybe it is
        slow, there are&nbsp; faster algorithms implemented elsewhere nota=
bly
        the F4 and F5 Faug=C3=A8re algorithms in Maple and Magma, but it i=
s a
        limitation of the concept that for a complicated enough problem,
        the grobner computation will take "infinite" time=2E Also the
        grobner basis obtained will depend of the ordering of the
        polynomial set, so chance is a factor here=2E <br>
      </font></p>
    <p><font size=3D"4" face=3D"monospace">In fact the author Rychlick has
        written an extension of his code, long ago, introducing colored
        variables and such stuff which are useful for solving
        algebraically problems in geometry, a well known application of
        this business=2E At the time i took a copy of this program, you
        can take it at:</font></p>
    <p><font size=3D"4" face=3D"monospace"><a class=3D"moz-txt-link-freete=
xt" href=3D"https://www=2Elpthe=2Ejussieu=2Efr/~talon/rychlik_CGBLisp=2Etgz=
">https://www=2Elpthe=2Ejussieu=2Efr/~talon/rychlik_CGBLisp=2Etgz</a></font=
></p>
    <p><font size=3D"4" face=3D"monospace">There is another implementation
        of grobner basis in maxima, that can be found in share/affine,
        due to the late William Shelter=2E A long long time ago it seemed
        to work, slower than the Rychlick implementation, but since the
        package affine was written for doing algebraic geometry on an
        affine map of an algebraic variety, for non commutative
        variables, it is possible it was working for non commutative
        polynomials=2E However when i have tried to play with this stuff
        more recently almost nothing worked and there is essentially
        zero documentation=2E&nbsp; Maybe this is extremely interesting bu=
t
        short of understanding all the code, which probably requires a
        lot of work done by specialists of the subject=2E<br>
      </font></p>
    <p><font size=3D"4" face=3D"monospace"><br>
      </font></p>
    <div class=3D"moz-cite-prefix">Le 21/07/2026 =C3=A0 02:44, Robert Dodi=
er a
      =C3=A9crit&nbsp;:<br>
    </div>
    <blockquote type=3D"cite" cite=3D"mid:CAAsY_sT9W5evntJYMfe_760+LSO_Bzy=
nECwpLMGNMpvROdzXOw@mail=2Egmail=2Ecom">
      <pre>(1b) as part of that, maybe fix bugs, extend, unify, and update=
 the
existing stuff for Groebner bases=2E I don't know where that stands at
present=2E</pre>
    </blockquote>
    <pre class=3D"moz-signature" cols=3D"72">--=20
Michel Talon</pre>
  </blockquote></div></body>
</html>

------2TH0TYGRCHGHMP3JFN2U6Q47S6643L--


--===============4176698941868649853==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline


--===============4176698941868649853==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
Maxima-discuss mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/maxima-discuss

--===============4176698941868649853==--