Category: Programming

My Wordle result on Christmas Day this year looked like this:

⬛⬛🟩🟩⬛
⬛⬛🟩🟩⬛
⬛🟩🟩🟩⬛
⬛🟩🟩🟩🟨
🟩🟩🟩🟩🟩

It almost looked like what I wanted, which was a Christmas tree with a star, like this:

⬛⬛🟨⬛⬛
⬛⬛🟩⬛⬛
⬛🟩🟩🟩⬛
🟩🟩🟩🟩🟩

This got me thinking a little bit. For one, is such a pattern possible? It probably is given the large amount of accepted Wordle guesses. But more importantly, does there exist a sequence of guesses which “makes sense”, and still produces the desired pattern.

It turns out producing Wordle art isn’t a new thing. Frankly, this is not a difficult thing to accomplish: simply loop through the word dictionaries until the desired pattern is matched and move on to the next pattern. This is not some big data exercise as the number of words is less than 13000.

The second question, can we make a sequence which makes the art while still making “sense”, is also pretty easy. I define making sense in the Wordlebot manner:

If a sequence of guesses strictly reduces the number of possible answers.

The reason we ask for strict inequality is that if the nth guess results in only 1 remaining possible solution (say “carol”), then the next guess must be “carol.” This means that, with possibly more guesses than available, one can always find the answer.

This was not difficult to code up in a brute force manner, and, spoilers, for the solution on 12/25/2025, PRISM, one could’ve guessed

FORTH

QUIET

BRISK

PRISM

and gotten the nice Christmas tree.

Code located here.

Bob’s Burgers has been the ambient track to my life for awhile now. It provides a comforting background noise from which I can choose to tune in to or out of at ease due to the irrelevant (but good humored) nature. Day to day, I would pull up Hulu, and just resume… however, this is not ideal. I only have the ad supported version of Hulu meaning no offline or minimized playing, and it has ads.

“But Marshall, Bob’s Burgers is only available on Hulu. Where else are you watching it?”

Good question. For awhile now, there have been people uploading virtually full episodes onto YouTube. As this is against copyright, these videos all have various ways to skirt around the initial screening: cropped videos, truncated episodes, slightly distorted sounds. They still get taken down relatively quickly. For an explanation of why this is happening, see posts like this. But I would also use these videos, as they generally don’t have shuffled episodes too, meaning I can be surprised by what comes up next. I can also play this very much like a podcast due to having Premium.

Recently, I got a Garmin smartwatch with the capabilities to play stored music. Looking to curb my phone dependence, I thought about putting Bob’s Burgers on the watch in an audio format. However, I wanted a few constraints:

1. Since I listen to this sometimes to zone out/nap, as much as I like the introductory jingle, it needs to go.
2. The audio should also generally be faded out before the credits, as it’s usually a long jingle which is too distracting.
3. Audio cannot be too “jumpy.”

This turned out to be very easy due to the structured nature of the episodes. Suppose one has the files of the episodes in “.mkv” files in a legal manner, then the following lines

for file in *.mkv; do
echo "Processing $file..."
filename="${file
ffmpeg -i "$file" -map 0:a:0 \
-ss 00:00:20 -to 00:20:20 \
-filter:a "afade=t=in:st=20:d=5,afade=t=out:st=1215:d=5" \
-metadata title="$file" -metadata artist="The Bobs" \
-id3v2_version 3 "${filename}.mp3"
done

creates mp3 files of the audio which simply fades in/out the credits and out song. The last point about normalizing the audio turned out to not be important, but using the following line into the for loop might help

  ffmpeg-normalize --progress -c:a libmp3lame "${filename}.mp3" -o normalized/${filename}.mp3

By moving this onto Garmin, I know can shuffle through various episodes!