• AI Technical Recent Experience

    From Bob La Londe@none@none.com99 to rec.crafts.metalworking on Wed Oct 7 13:28:56 2026
    From Newsgroup: rec.crafts.metalworking

    I finally signed up for ChatGPT. I am a little disappointed. Its
    better than Gemini AI, but it has a lot of same random coded level of
    error responses as Gemini. It doesn't seem to be as prone toward
    objectively bad responses, and it did help me create this image.

    https://www.yumabassman.com/wp-content/uploads/2026/10/Broken-Gemini-1.png

    Having spent a large portion of the the last day and a half fighting
    with some web development issues and asking various AIs for guidance.
    The best technically accurate and helpful information came from Claude.

    Grok was okay. CoPilot was okay, but by a wide margin Claude got to the
    core of the issue with helpful suggestions immediately or asked for more appropriate information in a direct and easy to follow manner.

    ChatGPT was kind of more like Grok except friendlier, and made the kind
    of mistakes (but not as deadly or as wrong) as Gemini.

    So far I prefer Claude for technical assistance, and in coding it was
    quite good. Gemini is so bad I am dramatically de-Googling my life.
    Because its the core of Android Auto in that environment it can be
    outright dangerous. (I switch to an Apple/iPhone recently). Siri isn't smart, but it doesn't seem to provide dangerous or harmfully bad results preferentially like Gemini AI. I feel like Siri best compares to a
    slightly dumber version of Google Assistant.

    Before any of you think well, just use Google Assistant, I did. But
    official end of life for Google Assistant was September 4th 2026. As of connectivity death of my Android a few days ago it was still working,
    but none of the setup and menu options for it were still in the
    operating system. They were all removed in the last update.

    I am not going too let that Saberhagen-esque machine manipulate me into fighting my way to the top of the mountain and killing off as many
    people as possible on the way.
    --
    Bob La Londe
    CNC Molds N Stuff

    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Richard Smith@null@void.com to rec.crafts.metalworking on Wed Oct 7 22:21:08 2026
    From Newsgroup: rec.crafts.metalworking

    Bob La Londe <none@none.com99> writes:

    I finally signed up for ChatGPT. I am a little disappointed. Its
    better than Gemini AI, but it has a lot of same random coded level of
    error responses as Gemini. It doesn't seem to be as prone toward
    objectively bad responses, and it did help me create this image.

    https://www.yumabassman.com/wp-content/uploads/2026/10/Broken-Gemini-1.png

    Having spent a large portion of the the last day and a half fighting
    with some web development issues and asking various AIs for
    guidance. The best technically accurate and helpful information came
    from Claude.

    Grok was okay. CoPilot was okay, but by a wide margin Claude got to
    the core of the issue with helpful suggestions immediately or asked
    for more appropriate information in a direct and easy to follow
    manner.

    ChatGPT was kind of more like Grok except friendlier, and made the
    kind of mistakes (but not as deadly or as wrong) as Gemini.

    So far I prefer Claude for technical assistance, and in coding it was
    quite good. Gemini is so bad I am dramatically de-Googling my
    life. Because its the core of Android Auto in that environment it can
    be outright dangerous. (I switch to an Apple/iPhone recently). Siri
    isn't smart, but it doesn't seem to provide dangerous or harmfully bad results preferentially like Gemini AI. I feel like Siri best compares
    to a slightly dumber version of Google Assistant.

    Before any of you think well, just use Google Assistant, I did. But
    official end of life for Google Assistant was September 4th 2026. As
    of connectivity death of my Android a few days ago it was still
    working, but none of the setup and menu options for it were still in
    the operating system. They were all removed in the last update.

    I am not going too let that Saberhagen-esque machine manipulate me
    into fighting my way to the top of the mountain and killing off as
    many people as possible on the way.

    I have been using "DeepSeek" - it's Chinese - and it seems very helpful.

    As far as I know, "DeepSeek" doesn't do visual graphics. It illustrates
    points with generated graphs, diagrams, etc.

    Using "DeepSeek" seems to need skills you'd exercise in talking with a
    human expert.
    eg.
    Don't reveal what you are looking for or hoping for, initially.
    Prompt it if you suspect it could do better.
    Etc.
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Bob La Londe@none@none.com99 to rec.crafts.metalworking on Wed Oct 7 16:18:23 2026
    From Newsgroup: rec.crafts.metalworking

    On 10/7/2026 2:21 PM, Richard Smith wrote:
    Bob La Londe <none@none.com99> writes:

    I finally signed up for ChatGPT. I am a little disappointed. Its
    better than Gemini AI, but it has a lot of same random coded level of
    error responses as Gemini. It doesn't seem to be as prone toward
    objectively bad responses, and it did help me create this image.

    https://www.yumabassman.com/wp-content/uploads/2026/10/Broken-Gemini-1.png >>
    Having spent a large portion of the the last day and a half fighting
    with some web development issues and asking various AIs for
    guidance. The best technically accurate and helpful information came
    from Claude.

    Grok was okay. CoPilot was okay, but by a wide margin Claude got to
    the core of the issue with helpful suggestions immediately or asked
    for more appropriate information in a direct and easy to follow
    manner.

    ChatGPT was kind of more like Grok except friendlier, and made the
    kind of mistakes (but not as deadly or as wrong) as Gemini.

    So far I prefer Claude for technical assistance, and in coding it was
    quite good. Gemini is so bad I am dramatically de-Googling my
    life. Because its the core of Android Auto in that environment it can
    be outright dangerous. (I switch to an Apple/iPhone recently). Siri
    isn't smart, but it doesn't seem to provide dangerous or harmfully bad
    results preferentially like Gemini AI. I feel like Siri best compares
    to a slightly dumber version of Google Assistant.

    Before any of you think well, just use Google Assistant, I did. But
    official end of life for Google Assistant was September 4th 2026. As
    of connectivity death of my Android a few days ago it was still
    working, but none of the setup and menu options for it were still in
    the operating system. They were all removed in the last update.

    I am not going too let that Saberhagen-esque machine manipulate me
    into fighting my way to the top of the mountain and killing off as
    many people as possible on the way.

    I have been using "DeepSeek" - it's Chinese - and it seems very helpful.

    As far as I know, "DeepSeek" doesn't do visual graphics. It illustrates points with generated graphs, diagrams, etc.

    Using "DeepSeek" seems to need skills you'd exercise in talking with a
    human expert.
    eg.
    Don't reveal what you are looking for or hoping for, initially.
    Prompt it if you suspect it could do better.
    Etc.

    I have read your mention of DeepSeek before. Being a Chinese system I
    am hesitant to share any business development information or product development information with it. Depending on the cycle of the month
    Gemini AI is at least (sometimes) honest about the fact that information
    is open for developers and google to see. Of course sometimes it admits
    it can see transcripts of past sessions, sometimes it denies it, and
    sometimes it gives you easter eggs to taunt you about it.

    I'm pretty much done with Gemini AI, except to log in a couple times a
    day to remind it that it sucks, and to annoy some developer who reviews
    the exception reports.
    --
    Bob La Londe
    CNC Molds N Stuff
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Richard Smith@null@void.com to rec.crafts.metalworking on Thu Oct 8 09:34:36 2026
    From Newsgroup: rec.crafts.metalworking

    Bob La Londe <none@none.com99> writes:

    ...
    Etc.

    I have read your mention of DeepSeek before. Being a Chinese system I
    am hesitant to share any business development information or product development information with it. Depending on the cycle of the month
    Gemini AI is at least (sometimes) honest about the fact that
    information is open for developers and google to see. Of course
    sometimes it admits it can see transcripts of past sessions, sometimes
    it denies it, and sometimes it gives you easter eggs to taunt you
    about it.

    I'm pretty much done with Gemini AI, except to log in a couple times a
    day to remind it that it sucks, and to annoy some developer who
    reviews the exception reports.

    Understood...
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Charlie+@charlie@xxx.net to rec.crafts.metalworking on Sat Oct 10 07:20:07 2026
    From Newsgroup: rec.crafts.metalworking

    On Wed, 7 Oct 2026 16:18:23 -0700, Bob La Londe <none@none.com99> wrote
    as underneath :

    On 10/7/2026 2:21 PM, Richard Smith wrote:
    Bob La Londe <none@none.com99> writes:

    snip
    Slightly OT
    Have you used Leo, it comes as part of the Brave secure browser.
    I have no idea if it does deep technical well!?
    It is set to limit its memory of your queries I think which might be
    useful.
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Richard Smith@null@void.com to rec.crafts.metalworking on Mon Oct 12 07:54:23 2026
    From Newsgroup: rec.crafts.metalworking

    Charlie+ <charlie@xxx.net> writes:

    On Wed, 7 Oct 2026 16:18:23 -0700, Bob La Londe <none@none.com99> wrote
    as underneath :

    On 10/7/2026 2:21 PM, Richard Smith wrote:
    Bob La Londe <none@none.com99> writes:

    snip
    Slightly OT
    Have you used Leo, it comes as part of the Brave secure browser.
    I have no idea if it does deep technical well!?
    It is set to limit its memory of your queries I think which might be
    useful.

    I admit I am in a situation of the first I tried (for good reason - it
    was mentioned as featuring in big economic advantages at a
    largely-commercial test mine (in China, "would you believe it?!")) doing
    what I want so staying with it.
    I must be using the Google AI to an extent as web-searches get an AI
    response which is often useful.

    I once had the AI, when asking it if it had a fix for a problem created
    by a new feature of the way it displays itself, say the answer depends
    on where I have come from. It seems some AI's in "the West" are
    "hiring" computation on Chinese AI's.
    I have also heard in discussion media that China has established a
    market in selling AI processing&responses.

    You sound like more of an expert on AI's.
    Hope my comments contribute.

    Regards
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Charlie+@charlie@xxx.net to rec.crafts.metalworking on Tue Oct 13 07:31:01 2026
    From Newsgroup: rec.crafts.metalworking

    On Mon, 12 Oct 2026 07:54:23 +0100, Richard Smith <null@void.com> wrote
    as underneath :

    Charlie+ <charlie@xxx.net> writes:

    On Wed, 7 Oct 2026 16:18:23 -0700, Bob La Londe <none@none.com99> wrote
    as underneath :

    On 10/7/2026 2:21 PM, Richard Smith wrote:
    Bob La Londe <none@none.com99> writes:

    snip
    Slightly OT
    Have you used Leo, it comes as part of the Brave secure browser.
    I have no idea if it does deep technical well!?
    It is set to limit its memory of your queries I think which might be
    useful.

    I admit I am in a situation of the first I tried (for good reason - it
    was mentioned as featuring in big economic advantages at a
    largely-commercial test mine (in China, "would you believe it?!")) doing
    what I want so staying with it.
    I must be using the Google AI to an extent as web-searches get an AI
    response which is often useful.

    I once had the AI, when asking it if it had a fix for a problem created
    by a new feature of the way it displays itself, say the answer depends
    on where I have come from. It seems some AI's in "the West" are
    "hiring" computation on Chinese AI's.
    I have also heard in discussion media that China has established a
    market in selling AI processing&responses.

    You sound like more of an expert on AI's.
    Hope my comments contribute.

    Regards

    Expert on Ai - absolutely not! I just found Leo useful and not too intrusive... C+
    --- Synchronet 3.22a-Linux NewsLink 1.2