Why I Wanted a Backup

I work from home occasionally, and there’s nothing worse than your internet dying during an important work call, or right near the end of streaming a movie. After dealing with a few outages, I had it on my ‘backlog’ to find a reliable backup option that wouldn’t cost me another monthly subscription.

Normally I would use my cellular data plan as a hotspot, but unfortunately Verizon is extremely slow in my home.

Testing What Actually Works

Step 1: Check Coverage Maps

I started with OpenSignal, that shows crowdsourced network coverage data. In my area, it showed:

• T-Mobile: 147.70 Mbps average

• Verizon: 88.52 Mbps average

• AT&T: 73.96 Mbps average

These numbers looked promising, but they’re area averages. You need to verify coverage for your specific location before buying anything.

Step 2: The Free T-Mobile Test Drive

T-Mobile offers a free 30-day trial through their T-Life app. No credit card required. Here’s how it works:

1. Download the T-Life app

2. Sign up for the trial (uses eSIM on compatible phones)

3. Test T-Mobile’s actual coverage in your specific location

I did this to test the actual speeds in my apartment without spending money upfront.

Step 3: The Results

Unfortunately for me, the results were disappointing:

• Verizon hotspot: 2 Mbps (sometimes dropping to 600 Kbps)

• T-Mobile trial: Also about 2 Mbps, even by the window

Turns out, my building has terrible cellular reception regardless of carrier. I might still look into AT&T, but it’s most likely a reception issue with my building.

However, if your results actually look good, here’s what I almost decided to do.

The Best Options

1. T-Mobile Hotspot Device (If you have good coverage)

• Device: Franklin T9 Mobile Hotspot (~$20 used on eBay)

• How it works: Insert a T-Mobile SIM, creates WiFi network

• Data options: Prepaid plans starting at $10 for 2GB, up to $50 for 30GB

• Bonus: You can bring it with you when traveling

2. Dual SIM Phone Setup

• Keep your primary carrier

• Add a prepaid eSIM from another carrier

• Switch between them as needed

• No extra device to carry

Real-World Recommendations

1. Test before buying: OpenSignal showed great speeds in my area, but my actual apartment results were terrible. The free T-Mobile trial saved me from wasting money on equipment.

2. Building construction matters: Modern apartments with concrete and metal can block cellular signals, regardless of carrier.

3. Try near windows: If you’re getting poor signal, test near different windows - sometimes it makes a difference.

4. Consider alternative carriers: If T-Mobile doesn’t work, you could test AT&T or other carriers with cheap prepaid SIMs.

Summary

I hope you have better luck then I did. I may still end up buying the Franklin mobile hotspot, if not for home use then for when I’m traveling or working remotely for an extended period of time.

A used Franklin T9 hotspot is only around 20 bucks, with some prepaid data could save you a lot of frustration during an outage.

But before you spend any money, use a free trial or buy a prepaid sim if you have to. In my case, I discovered my building is just bad for cellular reception.

For those with better building reception than me, having that little hotspot device ready to go could be a lifesaver when your main internet inevitably goes down.

As of Mar 29, 2025, the best tool I’ve found to download videos is yt-dlp. Back in the good old days, the internet was very open and as long as you knew how to open chrome tools, you pretty much could save what you wanted.

Luckily yt-dlp mostly does the job, but you’ll also need a handy Chrome extension the stream detector to handle some edge cases.

Step 1: Install yt-dlp Using Homebrew

First, you'll need to install yt-dlp via Homebrew. Open your terminal and run:

brew install yt-dlp

✅ Once installed, you're ready to download supported videos.

Step 2: Download a Video with yt-dlp

Here’s an example:

~ » yt-dlp -o "scotland-4k.mp4" https://www.dailymotion.com/video/x82r9zl                                   07:26:52
[dailymotion] Extracting URL: https://www.dailymotion.com/video/x82r9zl
[dailymotion] Downloading Access Token
[dailymotion] x82r9zl: Downloading media JSON metadata
[dailymotion] x82r9zl: Downloading metadata JSON
[dailymotion] x82r9zl: Downloading m3u8 information
[info] x82r9zl: Downloading 1 format(s): hls-1080@60
[hlsnative] Downloading m3u8 manifest
[hlsnative] Total fragments: 201
[download] Destination: scotland-4k.mp4
[download] 100% of  541.08MiB in 00:01:44 at 5.18MiB/s
[FixupM3u8] Fixing MPEG-TS in MP4 container of "scotland-4k.mp4"

That means success. Pretty simple, but what happens if you try a less common video site and get this error?

ERROR: Unsupported URL: https://9gag.com/

Answer: Use the Stream Detector addon & yt-dlp

1. Install the extension from the Chrome Web Store:
   The Stream Detector on Chrome Web Store

2. Once installed, open the webpage with the video you want to download.

3. Click the extension icon, and you’ll see a list of urls.

4. click on the ‘master’ link.
   ⚠️ It won’t give you a confirmation when you click, but don’t worry—it’s already copied to your clipboard.

Go back to your terminal and run:

yt-dlp -o "video-name.mp4" <master-url>

And that’ll get you the video you wanted. I usually leave the video in my browser playing while it downloads, but that may not be needed.

Note: I tried at first writing a script using python and playwright to automatically collect the right streaming urls. But using the stream detector (or your network tab) seems to be more reliable.

Initial Setup for MacBook Pro with Apple Silicon (M4)

Prerequisites

1. Install Miniconda (if not already installed)

brew install miniconda

2. Initialize conda for your shell

conda init

You may need to restart your terminal or source your shell config

3. Create a dedicated conda environment

conda create --name whisperx python=3.10 -y

4. Activate the environment (conda does this similar to pipenv and tools like it)

conda activate whisperx

If you have activation issues, you may need to first run:

source /opt/homebrew/Caskroom/miniconda/base/etc/profile.d/conda.sh

5. Install PyTorch (optimized for Apple Silicon), ffmpeg & finally Install WhisperX

pip install torch torchvision torchaudio
brew install ffmpeg
pip install whisperx

Hugging Face Authentication for Speaker Diarization

Speaker diarization in WhisperX uses Hugging Face models that require authentication:

1. Create a Hugging Face account if you don't have one at huggingface.co

2. Critical Step: Accept the license terms for **both** required models:

   1. Visit https://huggingface.co/pyannote/speaker-diarization-3.1 and accept terms

   2. Visit https://huggingface.co/pyannote/segmentation-3.0 and accept terms

3. Generate a Hugging Face access token

   1. Go to https://huggingface.co/settings/tokens

   2. Create a new token with read access

Running WhisperX with Speaker Diarization

Finally you’re ready to run the command.

Tip: You may want to try with a very short audio file before something larger, in case you still run into any issues to debug

Basic command structure:

whisperx /path/to/your/audio.mp3 \

 --model large-v2 \

--compute_type float32 \

 --output_format txt \

--diarize \

 --hf_token YOUR_HUGGING_FACE_TOKEN

Creating a Reusable Script (Optional)

Save this as `transcribe.sh`:

#!/bin/bash

set -euo pipefail

# Replace with your actual token

export HF_TOKEN=your_token_here

whisperx "$1" \

 --model large-v2 \

--compute_type float32 \

 --output_format txt \

--diarize \

 --hf_token $HF_TOKEN

Make it executable & use it:

chmod +x transcribe.sh

./transcribe.sh your_audio_file.mp3

Common Issues and Troubleshooting

"Could not download" errors: Make sure you've accepted the terms for both required models on the Hugging Face website

Authentication errors: After accepting terms, always generate a fresh token

Performance issues: For large files, consider testing on a small clip first to verify everything works

Output Formats

I’ve tried txt and json outputs that both seem to work well. You may also want to try (untested by me):

* `srt`: Subtitles with timestamps

* `vtt`: Web video text tracks

Specify your desired format with the `--output_format` parameter.

There’s a meme around social media that Europe can no longer innovate, that the EU is ‘too old’ to make new companies, or that the whole continent is ‘in decline’.

This is nonsense. America right now clearly has a great economy and a lot of that is thanks to unicorns that have popped up over the last 10-20 years. But I didn’t buy this idea that EU* countries don’t have their own major players.

So I did a little research, and there’s actually even more highly impactful companies that I had initially remembered. To me high impact meant two things: it was fairly well known and valuation reached 1B (at least at one point). Now, I’m also including the UK and the nordics as origin countries. But since a lot of people casually include those in “Europe” I think it’s fine for this purpose.

Note: The Following table renders well on desktop/mobile web, but you may need to click the link to DataWrapper to see all entries if using the Substack mobile app

So here it is, a handy reference to share if you see this meme going around. Technically Novo Nordisk was founded in 1923 but since Ozempic has exploded in popularity and will very likely have a major impact on people you know, I thought it fair to include it.

I’m a big fan of the GDPR regulations the EU passed a few years ago. If you haven't yet, it's a good idea to find out what information companies have been keeping about you. You can use the site Your Digital Rights to help you do it, and it’s fascinating (worrying?) to see what the exports reveal.

Afterward, you can do basically the same thing to have companies delete what they have on you. I deleted my Facebook account the other day, and it was a great feeling—even though I still have Instagram.

But what about accounts like Reddit or Twitter? It would be really annoying to lose all the accounts/spaces you’ve followed over the years. But I also don’t want every single post to live forever.

So I found a script online that automates the deletion of all your Reddit posts and comments, and then I went ahead and turned it into a simple python app in case you don’t want to deal with any code: https://reddit.scottnelson.xyz/, it’s open source as well. Be sure you want to delete your old content before you hit submit.

Doing this made me think: we really should have more controls over our data. GDPR was a good start, but it seems like all social media should have some sort of expiration setting for posts. It’s why Snapchat became so popular: most chatter doesn’t live forever, it’s low-stakes because it’s ephemeral; it’s more natural when you know there isn’t a public record of every word you’re about to say. It’s also why people get ‘writers block’ but not ‘talkers block’.

Unfortunately for Twitter users it seems like any app that cleaned up all posts automatically were broken back in 2023. Frankly these kind of changes are why it’s value has dropped 80% in the last few years.

After enormous pressure, Instagram is finally changing its user experience for teens and kids, according to the Journal. This comes barely a week after Australia plans to ban it altogether for kids. This is likely the start of a new trend.

Frankly I think this is good news, which I never thought I’d be saying. As a teen I would have hated this, and think it totally unfair. But we now know social media like instagram is uniquely dangerous, even lethal, for young kids. And more broadly, the more you use social media the more harmful it is, for anyone.

This is at first kind of surprising. It seems harmless, and initially it ‘feels good’. And that’s why it’s so pernicious, and why comparisons to cigarettes are so apt. They both deceptively cause you harm in the long run.

But even despite this knowledge, and anti-social media evangelists like Cal Newport and Jonathan Haidt criticizing it publicly nonstop, I still use stuff like Instagram. And I don’t think we’re ever going to completely ban it - instead we should change it.

Control The Feed

One of the things that’s not talked about enough is how little control you have over your own news feed.

There’s a number of problems with that. First of all, it’s currently engineered to illicit angry, depressing, and other negative emotions - because scientifically that’s what increases engagement and therefore ad revenue over time. Big tech companies hire teams of PhD’s to make the content more addictive than a slot machine, and rile you up.

Second, you have virtually no control over what kind of content you’re getting. Maybe you’re energized and want to see political news. Or maybe you’re exhausted after work and just want some silly memes. Doesn’t matter - once you start scrolling the algorithm will do whatever it takes to keep you there.

Isn’t this bizarre? When I’m watching TV, I simply change channels whenever and to whatever I want. My rewatch of The Office isn’t suddenly interrupted with news in Syria (or vice versa). But that’s basically every news feed.

This wouldn’t be too hard to change, since we encourage tech reform all the time. There are already some solid ‘right to repair’ laws, copyright laws, GDPR privacy laws, etc that have been popular and successful. So it’s not hard to imagine some more user rights regarding personalized control over news feeds.

Imagine being able to tap a button to temporarily turn off political or depressing content. Or show only funny memes for a bit. Or even dial your feed to be less addictive for a day, so you can do the productive things you actually want to get done. That sounds a lot better to me.