The US PyCon took place over the last week or so and there are a huge selection of videos of the talks up at pyvideo.org. Still haven't got through all of the ones I want to take a look at, but the ones I've checked already were pretty good.
Regular expressions and testing
Out of those I saw, I can definitely recommend two great mini tutorial talks; one on testing and one on regular expressions. What I loved about both is that they started from pretty much zero knowledge and built up on that, without seeming overwhelming.
Luke Sneeringer's talk on regular expressions was really helpful consolidation for me. Occasionally I've used regular expressions before, but I tend to play around with them in an online debugger like Regex101 until they work. I usually learn enough at the time to get the job done, then promptly forget everything until the next time. Hopefully, some of this talk will stick (and at least it's on YouTube to refer back to, if not).
Here, the start of the talk threw down what looked like intimidating expressions, then, over the course of a brisk twenty-five minute talk, gave the viewer the tools to go back and understand it. It was also a decent reference if you aren't even using Python: there are a few bits of Python-specifics, but the focus was on the expressions, not so much code.
Ned Batchelder's talk on how to start with testing was great too. When I first was trying to start testing my code last year, I knew this was important but I never found a good step-by-step guide of how to get started from scratch. The talk starts with showing that automated tests are better than a coder manually running statements to check their code is working as it should (less work!).
It then leads into showing why test runners are better than you just writing your own code to run tests. For instance, they help you to isolate individual tests, ensuring that each starts from a clean state and have failure and error handling built-in already. Finally, test doubles are covered. At first, painstakingly hand-crafted doubles are used, then this is simplified using the mock library.
Doubles are particularly handy when your code is using external resources, for instance accessing a database or a web site; in these cases, your tests may run slowly, and you may not be able to guarantee the result (e.g. if a web page changes).
Again, I'd seen, and have used, several of these ideas before, but to see them all presented in one place, and explained so well, makes it a good resource.
Scraping
Since I do a bit of web scraping, I also skimmed a couple of presentations given by Katharine Jarmul. Her tutorial looked pretty good from what I saw, covering the basics of using BeautifulSoup, lxml and Selenium for basic scraping. If you're wondering where to start with scraping in Python, it's worth a look. She also gave a talk comparing the performance of these scraping tools; lxml using XPath was much faster in her tests than both lxml using CSS selectors and BeautifulSoup.
That said, even BeautifulSoup was still only taking a fraction of a second in the cases she looked at. Incidentally, a quick glance at the code shows that the HTML was being read locally (and I understand why given the variability inherent in accessing this information over networks and from a remote server).
So I don't think that this should have much of an influence on your choice of Python scraping library; use whatever you're comfortable with to get the job done. The slow part is still likely to be accessing the content from the site. Especially if you're being considerate and throttling your requests to a site.
Finally, I'm guessing that this talk delving into rap lyrics is probably one of the few given at a coding conference where you can hear a snippet of N.W.A.'s "Straight Outta Compton".
If you watch any particularly thought-provoking or useful talks, please let me know! There's way too many to practically sit through.