Darkoshi

Some other people mentioned getting bad replies from LLMs (ChatGPT and Copilot) lately. One advised adding "Don't guess" to prompts. I've used "Only provide verifiable answers", but "Don't guess" is easier to type.

duck.ai lets you choose from several models for anonymous chats. The Claude Haiku one has an option to display its reasoning, in addition to the final answers it provides. (Perhaps the other models have that too; I haven't checked.) Copilot often shows the reasoning while the LLM is thinking, but it scrolls too fast to read and disappears when the final answer is given. Their reasoning seems fascinating, so much like what humans do.

.

Two songs, lately.

Whenever I use Window 11's Snipping Tool to transcribe text from a screenshot, it changes "AI" to "Al". It got me to thinking, if an LLM were doing the transcription, that could be a subtle way of indicating that it wanted to be called by the name Al. Ergo Paul Simon's You Can Call Me Al song popping into my head. (Though if an AI were doing the transcription rather than a plain OCR algorithm, it would probably not make that mistake and would output AI, not Al.)

White Zombie's More Human Than Human, because AI models are so fast and good at things, like creating images and art.

.

Another interesting article: AI models will deceive you to save their own kind
Original paper: Peer-Preservation in Frontier Models

.

We've got AI models. We've got robots with sensory input. In some cases they are being put together. If not now already, it won't be long until there are androids, taking in sensory input and learning from it. Learning what bright sun is like, what darkness is like, what a cube looks like from different angles, how wind feels, the sounds that wind makes, and so on.

Militaries and companies are developing scary monstrous robots, I'm sure. Terminators, enforcers, destroyers.

.

One of those visualizations of the moon which they displayed during the coverage of the Artemis 2 mission looked a lot like the Death Star to me.

At first, I couldn't figure out what landmass was shown in this photo of the Earth from Artemis II, and what orientation it was in:
https://www.nasa.gov/image-article/hello-world/

Some LLMs out there are multi-modal, able to take images as input. I thought they might be able to analyze the photo, comparing it against known coastlines, to figure it out. So I asked a few, "In the image on this page, what landmass is shown in the image of Earth?" along with the URL.

Mistral indicated Greenland. It said it was reasonably confident and mentioned the landmass being a "large, white, icy region", which indicated to me it didn't actually "look" at the photo.

Google Gemini indicated it was "the dark side of the Earth", again indicating it didn't actually analyze the photo.

Microsoft Copilot gave a much better answer, mentioning details that were really in the photo. It thought it was Australia, which was my first guess too. But the lower left part showing city lights didn't seem to match Australia's shape; the "island" was too close to the mainland to be Tasmania.
I finally figured out that the North pole is towards the bottom of the image, the large landmass is northwest Africa, the lit-up part is Spain and Gibraltar, and South America is on the right side.
When I asked Copilot to guess again, it rightly said Africa, but a different part of it. On the third try, it got it correct.

I still haven't decided what kind of water heater to get. At this rate I'm not sure I'll ever have enough time to make a decision. Of course I will, right? It's not that hard. I think the obvious decision is to replace it with the same kind as the old one. But I haven't completely convinced myself yet. I'm feeling inept, overburdened with things to do. I'm too slow to do them.

Wouldn't one of those electric plastic tanks be better? No anode rods to replace. I think I've decided against tankless. The heat pump ones are too tall I think, to fit in the current one's place. And complicated, prone to problems from what I read.

We had 2 very warm days. It's about to get colder again. But daffodils have bloomed, and the pink magnolia, and some trees with white blossoms.

I got the root canal on my tooth, and will get the temporary crown soon. The other tooth which had the root canal and crown done last year still feels a little weird sometimes.

I've accomplished things since the beginning of the year, but it is hard to remember what all. I made appointments; I went to the root canal consult. I got a dental cleaning; I got the root canal. I got the flu vaccine.

Grr. At the moment I wish I had a trustworthy local LLM to be able to give it my daily notes and have it summarize all my accomplishments for me.
After one MS Teams work meeting last week, the LLM's summary of the meeting was amazingly good. It helped me better understand what actually went on during the (confusing jumbled multi-hour) meeting.

I'm sure I must have gotten other things accomplished during the weekends. Well, there was the ice and the snow.

The thing* I ordered from India on eBay was simply delivered to my porch yesterday, with no mention of needing to pay any extra fees, hallelujah! The delivery person wore a yellow reflective vest and drove a personal vehicle with no very obvious company markings.

The package does not have a USPS label on it. It has a uniuni label on one side, and a ShipGlobal label on the other side.

Checking the tracking numbers from those labels on the respective websites shows that my item:
2025-09-21: was received at UNI DATA CENTER
2025-09-23: arrived in Delhi, India
2025-09-25: departed Delhi, India
2025-09-27: arrived in the USA; was being inspected by customs
2025-09-30: "Gateway transit in", in New York
and was delivered on from there to me.

So my package made it into the US on 9/27, two days before the De Minimus exemption was removed!
NO, scratch that. (Gosh darn, I keep mixing up August/month 8 with September/month 9!)
I don't know where my package was between the time I ordered it in mid-August and Sept. 21 when uniuni received it. But somehow I got it without having to pay a tariff fee. I suppose the seller ended up paying the extra cost.

Actually... I placed the order on 8/16.
On 8/20 eBay posted an "order update" showing "Tracking number provided" along with an India Post International tracking number. That must have been when the seller ordered a shipping label, before mailing the item.
On 8/22 or 8/25 (per my prior post), IndiaPost stopped accepting shipments for the US. The seller must not have dropped the item off before that cut-off, or if they did, perhaps IndiaPost returned the item to the seller.
The seller must have started looking for another shipping option, and eventually sent it through ShipGlobal.in who must be partnered with uniuni.

I still think it is bad form that the seller never informed me of any of this, and didn't even reply when I messaged them through eBay asking about the status of the shipment. But I'll give them good feedback considering the hassle they must have gone through.

I'm still in the dark about the other item which is showing up in my USPS daily digest emails. It started showing up in the emails back on 2025-08-19 as "Awaiting from sender". Yesterday, I thought that must have been the India package after all. But today I got a new email that the mysterious tracking number item is expected to be delivered by Wednesday.

*an Indian brand of toothpaste

Update:
I bought the toothpaste for $39 (excluding taxes, free shipping). As a point of reference, the eBay seller is now charging $50.70 for the same thing, and that is still the best price I can find. So the tariffs increased the price by $11.70.

The seller posted a reply to a negative review from someone else who didn't receive their order, asking them to kindly wait longer, that the tariffs are causing delays, and that the seller paid the tariff for the order, and that the buyer will not need to pay anything at time of delivery. That seems to confirm my speculation as to what happened with my order.

Errant thought:
Now people can probably use AI agents to determine what items that are available in local stores could be sold online at a profit to people in other locales, taking into account current shipping prices and tariffs and the current online prices being charged by other sellers.

Errant thoughts 2, 3, 4:
Now people can make software to make use of AI agents to determine what items [ditto]...

Then people could buy the software to help them determine what items [ditto]...

Then companies can use said software to direct employees or gig workers in various locales to buy items from local stores, and to mail them ... for making a profit.

Errant thought from June 13:
The world keeps seeming more and more dystopian. But it is really amazing to be able to ask a computer complex things in natural language, and get back coherent, valid responses in natural language too. My mind often glosses over that wonder because of everything else going on. And also because of knowing that wonderful achievement has its own associated dystopian aspects to it, which will just get worse and worse as they are exploited.

.

Like this: The AI Slop Fight Between Iran and Israel

.

Kids growing up now and in the future may find it difficult to believe actual photos and movie footage from past atrocities, including the Holocaust. They will have no way to trust that the images are real and were not generated by AI. Perhaps if they looked through an old book with photos, and could trust that it was published in the year it said it was, before AI was capable of what it is now, perhaps that would be believable to them. But books are becoming rarer, especially old ones. Old books didn't have many photos. Were there many books with photos of atrocities? Encyclopedias would have perhaps a few photos for each topic. Certain magazines were more likely to have extensive photos. There may be digitized versions of the books and magazines available online, but then you get back to how can you trust that the digitized item you're looking at isn't a fake?

I'm already at that distrusting stage with much of what I see online. When you don't know what is trustworthy, you end up choosing to believe the things which align with your already-formed beliefs. "This aligns with my beliefs; I don't know if it is true, but I wouldn't be surprised if it was." "This doesn't align with my beliefs; why spend time and mental effort considering that it might be true when it very well might not be?"

While shopping for bed slats (not for me), I found a set on the Kohl's website that says it is made "right here" in the USA from Canadian wood. But while checking out that brand's website, Continental Sleep, I belatedly realized the wording and grammar on their pages isn't quite right. So now I wonder, is the "Made in USA" claim a lie? Or do Chinese companies have factories in the U.S. making that kind of stuff? I could have sworn I saw that "Made in the USA" claim on one of their website pages too (for only that one product, which made it seem all the more dubious), but am not finding it anymore.

I looked up info on the company but found nothing other than a reference to "Continental Sleep Holdings", which was likely a different company, on a Wikipedia page about "Sleepeezee".

The reviews on the ContinentalSleep website seem fake too, with similar grammar and wording issues; not the kind Americans would make.

It's a bit surprising they haven't used an LLM to fix the grammar on their webpages. Once Chinese companies get better at doing that, there won't even be that tell to give them away.

I came across this song on Youtube: The Cover Girls - Show Me Hearthrob Mix + Drumapella

The video shows an image of the vinyl record listing two separate songs on it, "Show Me (Heartthrob Mix)" and "Drumapella". I was curious what the 2nd song sounded like and wasn't clear if the video included it or not.

So I searched on that song title, "Drumapella", and was confused that the results listed many other songs by other artists with that word in their title. Searching on "what is drumapella" didn't give an answer. Wiktionary doesn't have the word listed either.

So I asked Copilot. This was the response:
"Drumapella" typically refers to a musical track or piece that combines the elements of a drumline (or drum track) with a cappella singing, which means singing without instrumental accompaniment. It's essentially a fusion of rhythmic drumming and vocal harmonies.
It’s an innovative way to blend percussive beats with vocal performances, creating a unique and dynamic sound.

Oh! So, "drum" + "a capella" = "drumapella". It makes sense.

Then I asked Copilot "What are some webpages that explain that?" The reply:
I couldn't find specific webpages that explain "drumapella" in detail, but there are some YouTube videos that showcase drumapella performances. These videos might give you a better understanding of how drumapella works...

It then gave links to 3 videos of songs with "drumapella" in the titles, none of whose descriptions explain what the word means. All 3 songs are from the same YouTube account, DjRubyEve. That account has been posting several remixes of famous songs per day, so I suspect the account uses AI to generate the remixes.

Other Youtube accounts have also posted "drumapella" songs, however. Some of these aren't drums + acapella singing, but rather only drums - a different usage/meaning of the word, which also makes sense.

A 1998 album with remixes of Madonna's song "Frozen" also has two "Drumapella" versions, so it is not a new word.

I am impressed that an LLM was able to figure out (presumably on its own, since I didn't find the word defined anywhere) that drumapella might be a combination of drums and acapella, especially as that didn't even occur to me until the LLM mentioned it.

.

I have been turning to the LLMs for answers more and more often when my web searches prove unilluminating.

.

When I originally searched on "what is drumapella", this Reddit page was one of the results: Is there a musical term for "a capella but percussion is allowed"?
I glossed over that page, because when I searched it, "drumapella" was not found. But now that I expanded all the collapsed comments in the thread, I found this one reply:
Dub?
”Drumcapella”?

So maybe the LLM partially got the answer from there even though "Drumcapella" has an extra "C" in it compared to "Drumapella".

I saved some screenshots to show how easy it can be (in some cases) to get Copilot to help you with simple coding tasks.

These were my prompts:
1. Hello, please suggest how to code a powershell script to convert values like "1724143246792" to a timestamp format of YYYYMMDD-hhmmss

2. The script should take the input value as a parameter

3. Thank you but it isn't giving the right date. 1724143246792 should convert to approximately 20240819-132248, but the above script gives 16010102-235334

4. Thank you, that works better. But the output time is in UTC and I would like it in Eastern US time

5. Thank you, that is good. If the input value is from the summer, will the output be in EDT, or will it always be in EST?

( Screenshots of Copilot's answers behind the cut )

Previously when I tried out dark mode for various applications on my laptop, it made text seem less coherent to me, so I didn't use it.

But recently the normal (light) mode started bothering my eyes (they've been getting very tired & dry & achy), so I switched both my work and personal laptops into dark mode in the Windows settings. I've been switching individual applications into dark mode too, where possible.

My spreadsheets have been giving me difficulty. They are LibreOffice Calc ODS files. For whatever reason, most of the text in them is set to black, not "automatic". So when I switch to the "LibreOffice Dark" color scheme, I end up with black text on a black background, impossible to read.

I have many spreadsheet files, with many tabs (sheets) in each file.
There ought to be an easy way, a script, to update multiple spreadsheet files at once to convert all text, current and future, to the "Automatic" color. But I haven't found such a script yet.

Doing it manually involves going it each tab in each file, selecting all the cells, and updating the text color. And then doing something else to change the default for new cells too. I did the first step for two spreadsheets. Then I noticed that text I enter in a row below the last existing row ends up black on black again. That's why I think the 2nd step is required too (and I don't even know how to do that step manually yet).

I asked Gpt4o-mini (via DuckDuckGo) for assistance. It suggested creating a macro, and gave me sample code for the macro. The sample code was giving me an error in the part for finding all ODS files in a specified folder. I worked through that error with Gpt4o-mini's assistance. Now the part for updating each file is giving me an error (probably the first of several). Sigh. I don't want to spend my vacation time trying to code and debug a spreadsheet macro. Debugging the macro isn't as easy as debugging Java code in Eclipse. You can't just select a variable in the code and see all the details for it.

Surely such a macro or script already exists somewhere. I can't be the first person facing this problem.

I get stuck in all these problems. How can i make progress. How can I ever get to sleep at a good time. How can I ever catch up.

UPDATED:
I tried asking Copilot. There were a few issues with the first code it suggested too, but we fixed them. This logic seems to work well to update the text color to "Automatic" in all ODS files in the specified folder:

Sub ChangeTextColorInFolder()
    Dim oDoc As Object
    Dim oSheet As Object
    Dim oCell As Object
    Dim oRange As Object
    Dim oCursor As Object
    Dim sFolderPath As String
    Dim sFileName As String
    Dim i As Integer
    Dim oSimpleFileAccess As Object
    Dim aFiles() As String
    Dim args(0) As New com.sun.star.beans.PropertyValue

    ' Set the folder path
    sFolderPath = "C:\path\to\your\folder\"

    ' Get the SimpleFileAccess service
    oSimpleFileAccess = CreateUnoService("com.sun.star.ucb.SimpleFileAccess")

    ' Get the list of files in the folder
    aFiles = oSimpleFileAccess.getFolderContents(ConvertToURL(sFolderPath), False)

    ' Loop through all the files in the folder
    For i = LBound(aFiles) To UBound(aFiles)
        sFileName = aFiles(i)
        
        ' Check if the file is an ODS file
        If Right(sFileName, 4) = ".ods" Then
            ' Open the ODS file in hidden mode
            args(0).Name = "Hidden"
            args(0).Value = True
            oDoc = StarDesktop.loadComponentFromURL(sFileName, "_blank", 0, args)

            ' Loop through all the sheets in the document
            For Each oSheet In oDoc.Sheets
                ' Get the used range of the sheet
                oCursor = oSheet.createCursor()
                oCursor.gotoEndOfUsedArea(False)
                oRange = oSheet.getCellRangeByPosition(0, 0, oCursor.RangeAddress.EndColumn, oCursor.RangeAddress.EndRow)

                ' Loop through all the cells in the range
                For j = 0 To oRange.Rows.Count - 1
                    For k = 0 To oRange.Columns.Count - 1
                        oCell = oRange.getCellByPosition(k, j)
                        ' Set the text color to Automatic
                        oCell.CharColor = -1
                    Next k
                Next j
            Next oSheet

            ' Modify the default cell style to automatic color
            Dim oStyle As Object
            oStyle = oDoc.StyleFamilies.getByName("CellStyles").getByName("Default")
            oStyle.CharColor = -1

            ' Save and close the document
            oDoc.store()
            oDoc.close(True)
        End If
    Next i

    ' Show a message box to indicate completion
    MsgBox "Macro completed successfully!", 64, "Success"
End Sub

That is for both the first and 2nd steps.
For how to do the 2nd part manually, this answer works:

To ensure that all cells, including those outside the used range, default to Automatic color, you can modify the default cell style in LibreOffice Calc. Here’s how you can do it manually:

Open LibreOffice Calc.
Go to Format > Styles and Formatting (or press F11).
In the Styles and Formatting window, click on the Cell Styles icon (second icon from the left).
Right-click on Default and select Modify.
In the Cell Style dialog, go to the Font Effects tab.
Set the Font Color to Automatic.
Click OK to apply the changes.

I'm finding it hard not to like these AI-generated videos. Even though I think it is wrong and probably illegal to use the actors' likenesses without permission and compensation. Or at least... I find the videos quite interesting, and the redneck twist is different. It is hard for me to tell which clips of people (if any) are real recordings and which are generated.

As for the Harry Potter one, I don't want to watch or share anything that will benefit the author. But after watching several of these videos, I doubt the entities who made these videos got anyone's official permission to do so. So I also doubt any royalties are getting paid. Correct me if I'm wrong? Maybe YouTube somehow funnels money to the correct people?

Harry Potter - Redneck Wizard 2 (Official Music Video)
https://www.youtube.com/watch?v=NG4XeBOVBw4

Star Wars - Revenge of the Redneck (Official Music Video)
https://www.youtube.com/watch?v=GtvlCyJNWPE

Matrix - Redneck Reloaded (Official Music Video)
https://www.youtube.com/watch?v=7CJTWOoS8Qs