• =?UTF-8?Q?How_fast_is_EyeProlog=2c_do_we_need_=cf=80-WAM=3f_[Budget?= =?UTF-8?Q?_AI_Laptop]?=

    From Mild Shock@janburse@fastmail.fm to comp.lang.prolog on Wed Aug 12 01:02:32 2026
    From Newsgroup: comp.lang.prolog

    Hi,

    How fast is EyeProlog, do we need -C-WAM?
    Ok, here is a test:

    between2(L, L, R) :- !, L = R.
    between2(L, _, L).
    between2(L, H, X) :- J is L+1, between2(J, H, X).

    test :-
    between2(1, 100, _),
    between2(1, 100, _),
    between2(1, 100, _),
    fail.
    test.

    I get for EyeProlog and Dogelog:

    /* EyeProlog */
    0.476 sec
    https://eyereasoner.github.io/eyeprolog/playground

    /* Dogelog */
    % Zeit 446.020 ms, GC 0.000 ms, Lips 6794 k
    https://www.herbrand.ai/

    Thats not bad. Now watch -C-WAM, the first
    AI accelerated Prolog subset, with
    JavaScript Workers:

    :- ensure_loaded(library(misc/brainfog)).

    test2 :-
    execute((between(1,100,_),
    between(1,100,_),
    between(1,100,_))).

    ?- time(test2).
    % Zeit 52.675 ms, GC 0.000 ms, Lips 9 k
    true.

    BTW: With JavaScript WebGPU the AI acceleration
    is even an itch stronger, if the problem can
    be parallelized.

    Have Fun!

    Bye

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Mild Shock@janburse@fastmail.fm to comp.lang.prolog on Wed Aug 12 20:06:22 2026
    From Newsgroup: comp.lang.prolog

    Hi,

    Its interesting that we already have
    like for 50 years the Curry Howard
    isomorphims, as proof justifications terms,

    Still it seems extremly hard to generate
    proof certificates for Prolog, isn't it?
    How it started:

    Ai x Big Gas. What could go wrong?
    https://www.youtube.com/watch?v=kON2ZI2BNj8

    How its going:

    eq(Z,Z).
    why eq(f(X,a),f(b,Y))
    eq(f(b, a), f(b, a)).
    why(
    eq(f(b, a), f(b, a)),
    proof(
    goal(eq(f(b, a), f(b, a))),
    by(fact("<input>", clause(1))),
    bindings([binding("Z", f(b, a))])
    )
    ).
    https://eyereasoner.github.io/eyeprolog/playground

    Truely some bullshit authorized by Sarah Connor.
    So far there is no danger of an upcoming
    Skynet that will gained self-awareness.

    Bye

    Mild Shock schrieb:
    Hi,

    How fast is EyeProlog, do we need -C-WAM?
    Ok, here is a test:

    between2(L, L, R) :- !, L = R.
    between2(L, _, L).
    between2(L, H, X) :- J is L+1, between2(J, H, X).

    test :-
    -a-a between2(1, 100, _),
    -a-a between2(1, 100, _),
    -a-a between2(1, 100, _),
    -a-a fail.
    test.

    I get for EyeProlog and Dogelog:

    /* EyeProlog */
    0.476 sec
    https://eyereasoner.github.io/eyeprolog/playground

    /* Dogelog */
    % Zeit 446.020 ms, GC 0.000 ms, Lips 6794 k
    https://www.herbrand.ai/

    Thats not bad. Now watch -C-WAM, the first
    AI accelerated Prolog subset, with
    JavaScript Workers:

    :- ensure_loaded(library(misc/brainfog)).

    test2 :-
    -a-a execute((between(1,100,_),
    -a-a-a-a-a-a-a-a-a-a-a between(1,100,_),
    -a-a-a-a-a-a-a-a-a-a-a between(1,100,_))).

    ?- time(test2).
    % Zeit 52.675 ms, GC 0.000 ms, Lips 9 k
    true.

    BTW: With JavaScript WebGPU the AI acceleration
    is even an itch stronger, if the problem can
    be parallelized.

    Have Fun!

    Bye


    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Mild Shock@janburse@fastmail.fm to comp.lang.prolog on Wed Aug 12 20:17:27 2026
    From Newsgroup: comp.lang.prolog


    Whoever payed for this nonsense, had his money
    lavishly thrown out of a window. They could use the
    australian gas to directly burn some dollar notes.

    LoL

    Mild Shock schrieb:
    Hi,

    Its interesting that we already have
    like for 50 years the Curry Howard
    isomorphims, as proof justifications terms,

    Still it seems extremly hard to generate
    proof certificates for Prolog, isn't it?
    How it started:

    Ai x Big Gas. What could go wrong? https://www.youtube.com/watch?v=kON2ZI2BNj8

    How its going:

    eq(Z,Z).
    why eq(f(X,a),f(b,Y))
    eq(f(b, a), f(b, a)).
    why(
    -a eq(f(b, a), f(b, a)),
    -a proof(
    -a-a-a goal(eq(f(b, a), f(b, a))),
    -a-a-a by(fact("<input>", clause(1))),
    -a-a-a bindings([binding("Z", f(b, a))])
    -a )
    ).
    https://eyereasoner.github.io/eyeprolog/playground

    Truely some bullshit authorized by Sarah Connor.
    So far there is no danger of an upcoming
    Skynet that will gained self-awareness.

    Bye

    Mild Shock schrieb:
    Hi,

    How fast is EyeProlog, do we need -C-WAM?
    Ok, here is a test:

    between2(L, L, R) :- !, L = R.
    between2(L, _, L).
    between2(L, H, X) :- J is L+1, between2(J, H, X).

    test :-
    -a-a-a between2(1, 100, _),
    -a-a-a between2(1, 100, _),
    -a-a-a between2(1, 100, _),
    -a-a-a fail.
    test.

    I get for EyeProlog and Dogelog:

    /* EyeProlog */
    0.476 sec
    https://eyereasoner.github.io/eyeprolog/playground

    /* Dogelog */
    % Zeit 446.020 ms, GC 0.000 ms, Lips 6794 k
    https://www.herbrand.ai/

    Thats not bad. Now watch -C-WAM, the first
    AI accelerated Prolog subset, with
    JavaScript Workers:

    :- ensure_loaded(library(misc/brainfog)).

    test2 :-
    -a-a-a execute((between(1,100,_),
    -a-a-a-a-a-a-a-a-a-a-a-a between(1,100,_),
    -a-a-a-a-a-a-a-a-a-a-a-a between(1,100,_))).

    ?- time(test2).
    % Zeit 52.675 ms, GC 0.000 ms, Lips 9 k
    true.

    BTW: With JavaScript WebGPU the AI acceleration
    is even an itch stronger, if the problem can
    be parallelized.

    Have Fun!

    Bye



    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Mild Shock@janburse@fastmail.fm to comp.lang.prolog on Sun Aug 23 00:29:28 2026
    From Newsgroup: comp.lang.prolog

    Hi,

    ergoAlp at least tries to provide not only
    a why, but also a whyNot. But already its
    why is not very convincing. I guess the

    conjunction order got scrambled. I get:

    ?- why(ancestor(albert, dan)).
    ruA ancestor(albert,dan)
    ruA ancestor(bob,dan)
    ruA ancestor(carol,dan)
    ruA parent(carol,dan)
    ruA parent(bob,carol)
    ruA parent(albert,bob)
    https://ergoai-studio2-wasm.interprolog.com/

    Using my library(kbref) in Dogelog Player,
    I get the more convincing:

    ?- why(ancestor(albert, dan)).
    ruA ancestor(albert, dan)
    ruA parent(albert, bob)
    ruA ancestor(bob, dan)
    ruA parent(bob, carol)
    ruA ancestor(carol, dan)
    ruA parent(carol, dan)
    true;
    fail.

    And unlike ergoAlp, I don't need to call
    solve/2 inside show/3, there is enough
    information in the clause references to

    reconstruct head bindings. Source code:

    :- ensure_loaded(library(edge/kbrefs)).

    why(G) :-
    solve(G, T),
    show(G, T, 0).

    solve(true, true) :- !.
    solve((A,B), (J,K)) :- !,
    solve(A, J),
    solve(B, K).
    solve(H, (J,K)) :-
    clause(H, B, J),
    solve(B, K).

    show(true, true, _) :- !.
    show((A,B), (J,K), O) :- !,
    show(A, J, O),
    show(B, K, O).
    show(H, (J,K), O) :-
    compiled(J, H, B),
    tab(O), write(' ruA '), writeq(H), nl,
    O2 is O+3,
    show(B, K, O2).

    The predicates clause/3 and compiled/3 are from
    library(edge/kbref) and not found in the ISO
    core standard. Also I made parent/2 and

    ancestor/3 dynamic to be able to access it.

    Mild Shock schrieb:
    Hi,

    Its interesting that we already have
    like for 50 years the Curry Howard
    isomorphims, as proof justifications terms,

    Still it seems extremly hard to generate
    proof certificates for Prolog, isn't it?
    How it started:

    Ai x Big Gas. What could go wrong? https://www.youtube.com/watch?v=kON2ZI2BNj8

    How its going:

    eq(Z,Z).
    why eq(f(X,a),f(b,Y))
    eq(f(b, a), f(b, a)).
    why(
    -a eq(f(b, a), f(b, a)),
    -a proof(
    -a-a-a goal(eq(f(b, a), f(b, a))),
    -a-a-a by(fact("<input>", clause(1))),
    -a-a-a bindings([binding("Z", f(b, a))])
    -a )
    ).
    https://eyereasoner.github.io/eyeprolog/playground

    Truely some bullshit authorized by Sarah Connor.
    So far there is no danger of an upcoming
    Skynet that will gained self-awareness.

    Bye

    Mild Shock schrieb:
    Hi,

    How fast is EyeProlog, do we need -C-WAM?
    Ok, here is a test:

    between2(L, L, R) :- !, L = R.
    between2(L, _, L).
    between2(L, H, X) :- J is L+1, between2(J, H, X).

    test :-
    -a-a-a between2(1, 100, _),
    -a-a-a between2(1, 100, _),
    -a-a-a between2(1, 100, _),
    -a-a-a fail.
    test.

    I get for EyeProlog and Dogelog:

    /* EyeProlog */
    0.476 sec
    https://eyereasoner.github.io/eyeprolog/playground

    /* Dogelog */
    % Zeit 446.020 ms, GC 0.000 ms, Lips 6794 k
    https://www.herbrand.ai/

    Thats not bad. Now watch -C-WAM, the first
    AI accelerated Prolog subset, with
    JavaScript Workers:

    :- ensure_loaded(library(misc/brainfog)).

    test2 :-
    -a-a-a execute((between(1,100,_),
    -a-a-a-a-a-a-a-a-a-a-a-a between(1,100,_),
    -a-a-a-a-a-a-a-a-a-a-a-a between(1,100,_))).

    ?- time(test2).
    % Zeit 52.675 ms, GC 0.000 ms, Lips 9 k
    true.

    BTW: With JavaScript WebGPU the AI acceleration
    is even an itch stronger, if the problem can
    be parallelized.

    Have Fun!

    Bye



    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Mild Shock@janburse@fastmail.fm to comp.lang.prolog on Sun Aug 23 00:59:07 2026
    From Newsgroup: comp.lang.prolog

    Hi,

    BTW, a whyNot explanation can have success branches:

    ?- whynot(ancestor(bob, albert)).
    rLo ancestor(bob, albert)
    rLo parent(bob, albert)
    ruA parent(bob, carol)
    rLo ancestor(carol, albert)
    rLo parent(carol, albert)
    ruA parent(carol, dan)
    rLo ancestor(dan, albert)
    rLo parent(dan, albert)
    ruA parent(bob, charlie)
    rLo ancestor(charlie, albert)
    rLo parent(charlie, albert)
    true.

    See also my Dogelog Player Literate Programming tech
    report, it has a vanilla interpreter for success and failure.
    Tree explanations can be derived from it:

    solve(true) :- !.
    solve((A,B)) :- !, solve(A), solve(B).
    solve((\+ A)) :- !, falsify(A).
    solve(H) :- clause(H, B), solve(B).

    falsify(true) :- !, fail.
    falsify((A,B)) :- !, forall(solve(A), falsify(B)).
    falsify((\+ A)) :- !, \+ \+ solve(A).
    falsify(H) :- forall(clause(H, B), falsify(B)).

    Bye

    Mild Shock schrieb:
    Hi,

    ergoAlp at least tries to provide not only
    a why, but also a whyNot. But already its
    why is not very convincing. I guess the

    conjunction order got scrambled. I get:

    ?- why(ancestor(albert, dan)).
    -aruA ancestor(albert,dan)
    -a-a ruA ancestor(bob,dan)
    -a-a-a-a ruA ancestor(carol,dan)
    -a-a-a-a-a-a ruA parent(carol,dan)
    -a-a-a-a ruA parent(bob,carol)
    -a-a ruA parent(albert,bob)
    https://ergoai-studio2-wasm.interprolog.com/

    Using my library(kbref) in Dogelog Player,
    I get the more convincing:

    ?- why(ancestor(albert, dan)).
    -aruA ancestor(albert, dan)
    -a-a-a ruA parent(albert, bob)
    -a-a-a ruA ancestor(bob, dan)
    -a-a-a-a-a-a ruA parent(bob, carol)
    -a-a-a-a-a-a ruA ancestor(carol, dan)
    -a-a-a-a-a-a-a-a-a ruA parent(carol, dan)
    true;
    fail.

    And unlike ergoAlp, I don't need to call
    solve/2 inside show/3, there is enough
    information in the clause references to

    reconstruct head bindings. Source code:

    :- ensure_loaded(library(edge/kbrefs)).

    why(G) :-
    -a-a solve(G, T),
    -a-a show(G, T, 0).

    solve(true, true) :- !.
    solve((A,B), (J,K)) :- !,
    -a-a solve(A, J),
    -a-a solve(B, K).
    solve(H, (J,K)) :-
    -a-a clause(H, B, J),
    -a-a solve(B, K).

    show(true, true, _) :- !.
    show((A,B), (J,K), O) :- !,
    -a-a show(A, J, O),
    -a-a show(B, K, O).
    show(H, (J,K), O) :-
    -a-a compiled(J, H, B),
    -a-a tab(O), write(' ruA '), writeq(H), nl,
    -a-a O2 is O+3,
    -a-a show(B, K, O2).

    The predicates clause/3 and compiled/3 are from
    library(edge/kbref) and not found in the ISO
    core standard. Also I made parent/2 and

    ancestor/3 dynamic to be able to access it.

    Mild Shock schrieb:
    Hi,

    Its interesting that we already have
    like for 50 years the Curry Howard
    isomorphims, as proof justifications terms,

    Still it seems extremly hard to generate
    proof certificates for Prolog, isn't it?
    How it started:

    Ai x Big Gas. What could go wrong?
    https://www.youtube.com/watch?v=kON2ZI2BNj8

    How its going:

    eq(Z,Z).
    why eq(f(X,a),f(b,Y))
    eq(f(b, a), f(b, a)).
    why(
    -a-a eq(f(b, a), f(b, a)),
    -a-a proof(
    -a-a-a-a goal(eq(f(b, a), f(b, a))),
    -a-a-a-a by(fact("<input>", clause(1))),
    -a-a-a-a bindings([binding("Z", f(b, a))])
    -a-a )
    ).
    https://eyereasoner.github.io/eyeprolog/playground

    Truely some bullshit authorized by Sarah Connor.
    So far there is no danger of an upcoming
    Skynet that will gained self-awareness.

    Bye

    Mild Shock schrieb:
    Hi,

    How fast is EyeProlog, do we need -C-WAM?
    Ok, here is a test:

    between2(L, L, R) :- !, L = R.
    between2(L, _, L).
    between2(L, H, X) :- J is L+1, between2(J, H, X).

    test :-
    -a-a-a between2(1, 100, _),
    -a-a-a between2(1, 100, _),
    -a-a-a between2(1, 100, _),
    -a-a-a fail.
    test.

    I get for EyeProlog and Dogelog:

    /* EyeProlog */
    0.476 sec
    https://eyereasoner.github.io/eyeprolog/playground

    /* Dogelog */
    % Zeit 446.020 ms, GC 0.000 ms, Lips 6794 k
    https://www.herbrand.ai/

    Thats not bad. Now watch -C-WAM, the first
    AI accelerated Prolog subset, with
    JavaScript Workers:

    :- ensure_loaded(library(misc/brainfog)).

    test2 :-
    -a-a-a execute((between(1,100,_),
    -a-a-a-a-a-a-a-a-a-a-a-a between(1,100,_),
    -a-a-a-a-a-a-a-a-a-a-a-a between(1,100,_))).

    ?- time(test2).
    % Zeit 52.675 ms, GC 0.000 ms, Lips 9 k
    true.

    BTW: With JavaScript WebGPU the AI acceleration
    is even an itch stronger, if the problem can
    be parallelized.

    Have Fun!

    Bye




    --- Synchronet 3.22a-Linux NewsLink 1.2