Weeknotes 2022-01-31
Worked on using the new deployment context feature of fastdeploy to automate deploying cast_hosting instances for my podcast hosting SaaS. It finally worked and the hardest part of this was getting the traefik config right. If users can choose domain names by themselves, you have to use wildcard certificates. This means that you have to use an API key for your DNS provider allowing traefik to manipulate DNS records for the dnsChallenge, which is the only one supporting wildcard DNS. But then you still have to get both this part and your httpChallenge based certificates to work well together. Ok, here's the static config part that made traefik work for me.
Additionally, I did two programming streams last week, which is an improvement. And then there was an interesting discussion on the PyDDF telegram channel about automatic speech recognition. I haven't looked at this topic since mid-2019 and wow, a lot is going on there. Maybe it's even possible to train an asr model for german just by using existing labeled training data like the Mozilla common voice dataset plus lots of unlabeled audio data (podcasts maybe?).
Articles
Useful Websites
Software
Videos
Podcasts
-
Mike Fisher of Etsy talks AI and E-commerce The Robot Brains Podcast
-
Episode 93: Launching Python, Virtual Environments, and Locking Dependencies With Brett Cannon The Real Python Podcast
-
Improve Your Productivity By Investing In Developer Experience Design For Your Projects - Episode 349 Podcast.__init__
-
Episode 100 - Daniel Terhorst-North - SOLID vs. CUPID Software Architektur im Stream
-
Episode 496: Bruce Momjian on Multi-Version Concurrency Control in Postgres (MVCC) Software Engineering Radio
-
UKW090 Corona Weekly: Mit dem Angriff von Omikron wird das alles in Ordnung kommen Unsere kleine Welt
-
David Heinemeier Hansson on Rails 7.0, Hotwire, and the future of Rails Remote Ruby
Always nice to hear that the RoR community is doing well and also moving in the "render html on the server"-direction
-
What matters in tech according to Benedict Evans The Robot Brains Podcast
-
Einzigartig? - unsere Individualität Das philosophische Radio
Papers
-
wav2vec: Unsupervised Pre-training for Speech Recognition
Maybe we don't need that much labeled data? The progress in speech recognition in the last few years looks very impressive...