breakingcups

From "The Psychology of Computer Programming":

    After months of effort, a particular application was still not working, so a consultant was called in from another part of the company. He concluded that the existing approach could never be made to work reliably. While on his way home he realized how it could be done. After a few days work he had a demonstration program working and presented it to the original programming team.
    Team leader: How long does your program take when processing?
    Consultant: About 10 seconds per case.
    Team leader: But our program only takes 1 second. {Team look smug at this point}
    Consultant: But your program doesn't work. If the program doesn't have to work then I can make it as fast as you like.

There's a reason Microsoft is aggressively deprecating "older" CPU's that work perfectly fine. Heck, I have one laptop with Windows 11 that worked great, but won't update from 22h2 to 24h2 because CPU support was dropped between versions, leaving me with only the glib suggestion from the Windows Update UI to "Buy a new device".

Ironically, installing Windows 10 and activating ESU would lead to longer hardware life.

Of course, I didn't. Instead, I installed Linux on that laptop too. My partner had no issues switching.

Microsoft had a lot of practice with this with the recent Xbox hardware & software.

You have no idea whether your server is currently actively compromised and participating in a botnet.

Luckily you have NVIDIA to fall back on, with their affordable, consumer-focused LLM chips...

I will say that a "Critical Security Vulnerability in flyctl, update now: https://bad-link/to/update.zip" tweet will have very serious consequences for a portion of your userbase, despite not directly compromising your own infra.

Thus far it has always been right for me, but you can tweak the settings to offer you a manual skip if you prefer to lean into it more slowly.

You can set up an alphabetical PIN, which is, for all intents and purposes, exactly the convenient password you're looking for.

Not that that should keep you on Windows, of course.

You haven't read the article in which they explain that that is the specific hole they're closing.

10 is no longer supported in 7 days, unless you activate ESU. Officially this requires a Microsoft account, but there's always the Massgrave way.

You can! Windows 11 Enterprise.

Sponsorblock took care of that for me

It might help your goal of a PSA if you source your claims, so this doesn't become inflammatory.

It then continued: "I'm sorry, but as an AI language model..."

It reminds me of Janet malfunctioning in the TV show The Good Place.

And we wonder why LLMs can be such an accelerant for people suffering from delusions.

Thank you. How long have you been on your current medication?

I've heard that a lot, to the point where the advice was not to listen to the advice of people who have been on their ADHD medication for less than a year due to the initial boost.

Actually, I believe the IA respects robots.txt retroactively, eg. putting something on the disallow list now removes the same page scrapes from a yeaer ago from public access in teh Wayback Machine, but I'd love to be corrected on that.

Specifically, because some pirates will put advertisements for other illicit services at the beginning or end of movies and tv shows in its subtitle data where there's a suitable gap. Usually those gaps have silence.

Companies incorporating subtitle data as transcription source of truth training data will thus train their models to output facsimiles of these messages whenever they're encountering prolonged stretches of silence.

I've read enough nuanced takes on ADHD medication that I'm very reluctant to even seek a diagnosis because I'm very sure I don't want to depend on them, for a variety of reasons.

It's made me feel like it's not even worth getting a diagnosis because the only clear path forward is medication or eternal struggle with various much-less effective coping strategies. Anyone who can weigh in with their perspective on this is welcome and thanks in advance.

Anything that is vaguely present in their training data (see, a wine glass filled to the brim [1][2]), that is.

https://medium.com/@joe.richardson.iii/the-curious-case-of-t... https://medium.com/@joe.richardson.iii/openai-slaps-a-band-a...

Wonder if it's possible to change the POP server without changing the credentials... You could set up your own sniffer server to retrieve the password that Gmail sends to you.

My pet theory without any strong foundation is because OpenAI and Anthropic have trained their models really hard to fit the sycophantic mold of:

    ===============================
    Got it — *compliment on the info you've shared*, *informal summary of task*. *Another compliment*, but *downside of question*.
    ----------
    (relevant emoji) Bla bla bla
    1. Aspect 1
    2. Aspect 2
    ----------

    *Actual answer*

    -----------
    (checkmark emoji) *Reassuring you about its answer because:*

    * Summary point 1
    * Summary point 2
    * Summary point 3

    Would you like me to *verb* a ready-made *noun* that will *something that's helpful to you 40% of the time*?
    ===============================

It's gotta reduce the quality of the answers.

That's a lot of carrots

Cool to see Ladybird get some corporate love. I wish Firefox got more varied sponsoring from multiple sources, too.

Google Gemini had an ongoing issue for at least 3 weeks with empty responses and the thinking context filled with absolute nonsense. It was even acknowledged on social media by their head of product but no official communication at all. No status page updates, nada. Apparently it was even a known issue for much longer but they only started fixing it after a config change made the problem much more prevalent.

How do you use GPT-5? I've wanted to try it from Cline, but to be allowed to use GPT-5 through the API I need to submit my biometrics to OpenAI and that's not happening.

Tangential, but I've only heard about BigQuery from people being surprised with gargantuan bills for running one query on a public dataset. Is there a "safe" way to use it with a cost limit, for example?

That seems made up, do you have a credible source for that?