aniketpanjwani
Basically, how can local media adapt to decreasing ad revenue and less interest in local issues? Case study shows how one newsletter/newspaper adapts and is able to do quite well with heavy automations.
Corresponding YouTube video here: https://www.youtube.com/watch?v=oWWmWo5WTGE
Don't have a list of YouTube videos/podcasts, but I'm going to start putting it together.
Local newsletter: a newsletter serving some local population with events info, business updates, local politics, etc. Typically operated by a single operator or a small staff.
They have become increasingly popular in many metros as local newspapers have been shutting down.
For this analysis, I wrote a Python scraper to parse 765 issues of one local newsletter - the Naptown Scoop. I used gpt-4o to help identify and classify the advertisers in each issue, which allowed me to analyze what types of advertisers are most common, and what types of advertisers drive the most revenue in local newsletters.
Happy to answer any questions both about the analysis and the data collection.
The Economics of Local Newsletters
The fact that happiness and health goes up dramatically when parents are north of 60 just reflects that in the past - if you didn't have kids - it's because you didn't get married. If you didn't get married, it was far more likely to be because you were somehow unviable as a marriage partner than because you didn't want to get married.
So, you're going to get a spurious, strong correlation between happiness/health and the interaction between age > 60 and having kids. If you run the same regression among Americans in 30 or 40 years, that same relationship will not hold, since whether to get married and whether to have kids are to ever-increasing degrees explicit choices.
1. At what sort of scale does Metaflow become useful? Would you expect Metaflow to augment the productivity of a lone data scientist working by himself? Or is it more likely that you would need 3, 10, 25, or more data scientists before Metaflow is likely to become useful?
2. When you move to a new text editor, there are some initial frictions while you're trying to wrap your head around how things work. So, it can take some time before you become productive. Analogously, I imagine there are initial frictions when moving to Metaflow. In your experience, after Metaflow's environment has already been established, how long does it take for data scientists to get back to their initial productivity? It would be useful to have a sense of this for the data scientist who would want to sell their organization on adopting Metaflow.
3. Many data scientists work in organizations which have far less mature data infrastructure than Netflix, and/or data science needs of a much smaller scale than Netflix. In particular, I may not even have batch processing needs (e.g. a social scientist working on datasets which can be held entirely in memory). In that case, is Metaflow useful?
4. What's the closest open-source alternative to Metaflow on the market? Off the top of my head, I can't think of anything which quite matches.
The Shallows: What the Internet Is Doing to Our Brains
1. Would it be possible to do this through mitmproxy? You are the person to ask, after all :P .
2. Does this approach allow you to also filter sites which use certificate pinning? That seems more important, since it would be a practical advantage, rather than simply more convenient installation.
3. Could you filter at the URL level with SNI, e.g. block amazon.com/gp/video/*, but not block the rest of amazon.com? From what I've just started reading about SNI, you would only see "amazon.com" and nothing else.
There are some websites which due to certificate pinning will not be properly filtered through the proxy (http://docs.mitmproxy.org/en/stable/certinstall.html#certifi...). For example, while Chomper is running, even if Dropbox is not on a blacklist, it probably will not properly sync. Eventually, I'll add the ability to make exceptions for these websites if desired.
It's been ad-hoc tested on my computer and on a fresh Linux installation. There's really no components right now worth unit testing, since it was really just a lot of configuration work. Once I start adding more complicated filter rules, I'll write formal unit tests.
Does that answer your questions?
1. It's browser independent. You have to import security certificates for each browser, but this is a step which will eventually be automated (https://github.com/aniketpanjwani/chomper/issues/8).
2. In both Firefox and Chrome, you can enter Private Browsing/Incognito mode, where extensions are automatically disabled, so you'll be able to evade blocks. It doesn't matter if you're in incognito mode while Chomper is running - you'll still get blocked.
3. It's usually trivially easy to disable or uninstall browser extensions. Chomper is marginally more difficult to disable when you have root privileges. However, unlike browser extensions, you have the option of tying your hands so that it is entirely impossible to remove a block (https://github.com/aniketpanjwani/chomper#hardcore-mode).
However, neither of these programs support Linux, and the only similar program hasn't been developed in over 7 years (http://svn.jklmnop.net/projects/SelfControl.html). So, I wrote Chomper. Any feedback is appreciated, but what's most useful is if you try installing it and tell me if you like it!