And if I don't care about cost, what about actual performance?

I had put an example like "decision locked" in my CLAUDE.md and a few days later 20 instances of Claude's responses had phrases around this. I thought it was a more general model tic until I had Claude look into it.

It is funny how that works. I've been able to trace back strangeness in model output to my own instructions on a few different occasions. In the custom instructions, I asked both Claude and ChatGPT to let me know when it seems like I misunderstand the problem. Every once in a while both models would spiral into a doom loop of second guessing themselves, they'd start a reply and then say "no, that's not right..." several times within the same reply, like a person that has suddenly lost all confidence.

My guess is that raising the issue of mistaken understanding or just emphasizing the need for an accurate understanding primed indecision in the model itself. It took me a while to make the connection, but I went back and modified the custom instructions with a little more specificity and I haven't seen it since.

The ancient Hawaiians had two main seasons - winter and summer, marked by the rising of Pleiades in the East during the winter and the setting of the Pleiades in the West summer, which corresponds roughly to November and May. And lines roughly up with what the article proposes.

And?

Friend probably still sleeping

Yup. Happily retired.

Anyway, he would have been one of the folks signing the checks.

If he says something interesting, I might report it back. He sometimes just blows sunshine up my ass, but he's certainly one for interesting stories.

Worked with some of the most deplorable narcissists in history, and hardly ever has a bad thing to say about any of them. I can see how he did so well. They probably loved working with him.

Just to report.

He told me that they got paid plenty of money, but their T-shirt sales were enormous. He said they still sell like hotcakes. Apparently, Nirvana has also done well in that department.

Of course, I can't speak for the general health of each band. They made money, but I'm not sure how well things worked out for them... I don’t think there’s any Ramones, left.

> Norton Antivirus for Android

I had no idea that was a thing

This is the take when you haven't really tried driving these tools with much practice

I don't think it's that

Here's an example from Gemini with some Lua code:

    label = key:gsub("on%-", ""):gsub("%-", " "):gsub("(%a)([%w_']*)", function(f, r) 
      return f:upper() .. r:lower() 
    end)

    if label:find("Click") then
      label = label:gsub("(%a+)%s+(%a+)", "%2 %1")
    elseif label:find("Scroll") then
      label = label:gsub("(%a+)%s+(%a+)", "%2 %1")
    end

I don't know Lua too well (which is why I used AI) but I know programming well enough to know this logic is ridiculous.

It was to help convert "on-click-right" into "Right Click".

The first bit of code to extract out the words is really convoluted and hard to reason about.

Then look at the code in each condition. It's identical. That's already really bad.

Finally, "Click" and "Scroll" are the only 2 conditions that can ever happen and the AI knew this because I explained this in an earlier prompt. So really all of that code isn't necessary at all. None of it.

What I ended up doing was creating a simple map and looked up the key which had an associated value to it. No conditions or swapping logic needed and way easier to maintain. No AI used, I just looked at the Lua docs on how to create a map in Lua.

This is what the above code translated to:

    local on_event_map = {
      ["on-click"] = "Left Click",
      ["on-click-right"] = "Right Click",
      ["on-click-middle"] = "Middle Click",
      ["on-click-backward"] = "Backward Click",
      ["on-click-forward"] = "Forward Click",
      ["on-scroll-up"] = "Scroll Up",
      ["on-scroll-down"] = "Scroll Down",
    }

    label = on_event_map[key]

IMO the above is a lot clearer on what's happening and super easy to modify if another thing were added later, even if the key's format were different.

Now imagine this. Imagine coding a whole app or a non-trivial script where the first section of code was used. You'd have thousands upon thousands of lines of gross, brittle code that's a nightmare to follow and maintain.

This sounds like moving the goal posts but Gemini is generally not a good model. Try Sonnet

They've been doing this since well before genAI

> grow some veggies and raise chickens.

Maybe Claude will be able to do that soon, too.

Smells like LLM written. Maybe useful advice though. Who actually writes the hyphen in compound modifiers (or even knows what that is)? E.g. "team-level" instead of "team level".

> I couldn't have been happier.

And phrases like this.

Definitely, although maybe it's ok here. I'm not sure. The opening paragraph doesn't feel like it. The rest could be condensed to 1 or 2 paragraphs and it would better communicate the idea.

It's not too bad because it's not too long, but I think it's worse than if the human had just written a second part to the post about the length of the first part.

Notice there's all these needless sections that have the LLM-form.

> What "Taking the Position" Actually Looks Like

> Why Sustained Performance Is What Counts

> The Responsibility-First Mindset

And each has this opening paragraph setup then a one sentence paragraph contradicting or reinforcing it, which is simultaneously punchy and pure fluff.

> I couldn't have been happier.

> And I mean sustained.

> But that's backward.

I can't really tell how much the author cares about any given bit of text beyond the starting paragraph because it's all expressed with too many words that don't say anything, but just evoke the marketing/linkedin-speak, giving everything too much weight.

good observations of AIGC form here

Continually sad to see each day that the vast majority of users cannot classify

Honestly that just seems like normal English.

To refer to something as "team-level" seems so absurdly unspectacular, relative to the other kinds of signals that exist for sussing out AI writing, that I'm surprised it was worthy of mentioning at all.

It's proper grammar but I have only ever seen journalists use it. It is a strong signal in my experience

Maybe so but in that case hats off to the LLM for writing a pretty good article.

Uh. I'd use a hyphen there. I'm not an LLM. At least, I don't think I am. But, who knows?

"808 all day" as an example