The “Biggest of All Time”: A Tiny RewindOS Prototype

Why I Built This

This is the first mini-project for RewindOS and it started with a simple question:

When I watched the latest season of Prehistoric Planet Season 3: Ice Age, I noticed that Tom Hiddleston who recently replaced the late great David Attenborough as the narrator, said some variation of “the biggest of all time,” quite a lot. So I set off to determine how many times without using my fingers to count it like the cavemen did in those prehistoric times.

It turns out with a little reverse engineering and hacking, this is quite easy and it highlights a metric that nobody in Hollywood seems interested in measuring— yet.

If you watch a lot of nature documentaries, you probably notice recurring superlatives—“the largest ever discovered,” “the biggest predator of its era,” “the strongest bite force in history.” These phrases shape how we interpret both animals and storytelling. They’re a mix of science communication and spectacle.

So I wanted to answer a playful but data-driven question:

How frequently does the show in season 3 use the phrase “of all time,” and in what contexts?

This became a perfect first test for RewindOS, because it touches everything I envision this project will be and ultimately do at scale:

• extracting structured data from media
• analyzing linguistic patterns
• building repeatable pipelines
• archiving, visualizing, and tracking cultural tendencies in media

How I Got the Subtitles

I already had the Season 3 video files. Thanks to my automated plex setup. So logically, I first went to my plex library and tried to extract the files that way.

However, I see in plex after some basic research and simple ls linux commands on my server that they were nowhere to be found readily available for extract in my system. Lo and behold further research proved my point:

Subtitles downloaded by Plex itself are stored inside of Plex’s blob files and aren’t able to be interacted with, nor can the location be changed.

Chatgpt gave me some helpful but rather tedious “legal” ways to obtain the subtitles by recording them on VLC/Plex as it plays. However after some more research I discovered VLC’s VLSub feature which can download the embedded files. Success!

This gave me five .srt files I needed — one for each episode.

Those files became the dataset for the project.

Not being a scratch programmer and now deeply entrenched into the era of AI, the next logical step was to build the python script.

The entire analysis pipeline was built using AI as a collaborator.

I described the problem (“Find all phrases ending in of all time in the SRT files”), and the model generated a clean Python script that:

1. Walks through my subtitle directory
2. Reads every .srt file
3. Uses a regex pattern to capture the phrase + its context
4. Prints them to the console
5. Exports everything into a structured CSV

The core of the extraction logic was this:

pattern = re.compile(
    r"(\b\w+(?:\W+\w+){0,6}\W+of all time\b)",
    re.IGNORECASE
)

This lets the script capture phrases like:

• “largest predators of all time”
• “one of the most powerful hunters of all time”
• “the biggest terrestrial bird of all time”

The final output CSV includes:

• which episode
• the snippet of text
• how many instances appear across the season

This is exactly the kind of small linguistic dataset that RewindOS will eventually let creators, journalists, and analysts explore effortlessly.

It’s easy to laugh at how often nature docs use hyperbole, but those phrases shape cultural impressions of extinct animals. They’re part of storytelling tradition. Being able to quantify them is the first step in understanding patterns across eras, genres, studios, and creators.

If you want to see or run the Python script used for this analysis, the full repository is here:

👉 GitHub: https://github.com/jjf3/prehistoric-planet-of-all-time-analysis