I am interested in quantitative / probabilistic analysis on your thesis. Where should I go look?

nice prompt

troll

Does anyone know how to reliably use Firefox from command line to take screenshots? It used to work well a few years ago but now it does not. For one, it asks that Firefox is already running and I need to kill it. This is surprising -- why can't two Firefox processes run at the same time?

It is whining about the shared use of your main profile directory. You can give it a temporary profile directory to more clearly express your intentions via "--profile=/tmp/$(uuidgen)" or similar. I'd guess you could even just straight up point <<env HOME=$(tmpdir) firefox --screenshot...>> for even stronger isolation

The CLI has long had `-noremote` for a long time to tell it you want a second Firefox. It's useful with `-ProfileManager` or `-P $profileName` for multi-profile workflows, which are out of fashion this decade, especially with Multi-Account Containers being able to do most of those workflows in the same browser window now with different tags. But some of us still have ancient profiles for ancient reasons and ingrained habits regarding them.

Ancient documentation: https://www-archive.mozilla.org/docs/command-line-args.html

  PROFILEDIR="$(mktemp -d)"
  firefox --no-remote  --profile "$PROFILEDIR" --screenshot $PWD/output.png https://xkcd.com
  rm -r "$PROFILEDIR"

(You don't have to create and destroy a profile directory every time, but it's cleaner to do that way and you need one per instance you're going to run anyways)

Script with Playwright?

Use playwright?

Is this true for the trained pilots as well?

It's next to impossible to know at night.

The only thing you have visually is direction / angular velocity.

And if you're on a collision course the other vehicle will appear motionless, with no change in direction.

If the airliners was approaching faster than the helicopter it could have been approaching from behind. How large is the field of view of a UH-60?

"How large is the field of view of a UH-60?"

Massive, you're practically against the glass. Visibility below is more obstructed than visibility above or to the sides.

From the radar tracking, they appear to have been at least somewhat head-on.

(Also at a low altitude in a large city ... lots of lights.)

I wonder if there's a support element that would have obstructed the field of view over a narrow angle. This has been the cause of automobile accidents when cars approach at just the correct speeds to keep the other vehicle behind the pillar at the side of the windshield.

Yes.

I have no skin in the game and rarely code in C++. My answer is: no.

Somewhat relevant, I use this little script to open front page of NYTimes daily morning on my macbook:

d=$(date +%d)

m=$(date +%m)

y=$(date +%Y)

wget -q https://static01.nyt.com/images/$y/$m/$d/nytfrontpage/scan.p... && open ./scan.pdf

I made an RSS feed of these front-page images a while ago if anyone wants to use them:

https://nytonline.net/rss

See also https://www.frontpages.com/

https://www.frontpages.com/us-newspapers/

That's neat! I turned it into a one liner `rm todays_nyt.pdf; wget -q "https://static01.nyt.com/images/$(date +%Y/%m/%d)/nytfrontpage/scan.pdf" -O todays_nyt.pdf && open todays_nyt.pdf `

dmy=$(date +%d/%m/%Y)

"Very few things are an immediate success". Wise words.

This may be a good starting point: https://linuxupskillchallenge.org/

Thank you for your thoughtful suggestions. They are actionable and practical. Much appreciated.

More towards ML-powered products and services.

You might want to learn to find your way around

https://scikit-learn.org/stable/

the algorithms it uses are not the most fashionable but the stuff in

https://scikit-learn.org/stable/model_selection.html

and

https://scikit-learn.org/stable/modules/preprocessing.html

make the difference between successful and unsuccessful projects. Huggingface has nice (in a certain sense) tools for training and doing inference on small LLMs but is a train wreck when it comes to model selection and preprocessing. (To be fair a few years back I tried developing a general purpose trainer that worked for bigger models that scikit-learn would handle but did the model selection and preprocessing well and didn't like the answer I got)