HN user
Python developer, fullstackpython.com and plushcap.com creator and author.
AssemblyAI (YC S17) is currently the one that stands out in the WER and accuracy benchmarks (https://www.assemblyai.com/benchmarks). Though its models are accessed through a web API rather than locally hosted, and speaker diarization is enabled through a parameter in the API call (https://www.assemblyai.com/docs/speech-to-text/pre-recorded-...).
I worked at Twilio from 2014-2023 and had an incredible experience during those 9 years, including all of my interactions with Jeff.
One of my favorite moments was running into Jeff in the hallway at our Beale Street office just before he had to do a quarterly earnings call after we went public. I said "good luck on the earnings call" and Jeff said thanks then asked what happened to a link to a specific Stack Overflow question/answer from one of our docs pages that got removed. He had just been building an application and felt like that answer was particularly helpful for context in a certain programming language, and was surprised it was removed.
Jeff's always been software developer at heart even though he had a ton of non-dev responsibilities. There are definitely advantages and disadvantages in that mindset! But he created a tremendous company along the way and I wouldn't change working there for anywhere else during those years.
I'd recommend just trying the Colab in my comment above to test out how quick you can do what you want with LeMUR versus building your own. Piping in 100 hours of audio into an LLM can be a lot of work compared to an API call, but it'll depend on what you are building
Hey HN, Matt from AssemblyAI here. If you want to test out LeMUR one of the fastest ways is with our Google Colab: https://colab.research.google.com/drive/1xX-YeAgW5aFQfoquJPX...
I'm happy to answer questions about the API as well
Do other US financial institutions have the same exposures, or is this a one-off situation based on SVB's closeness to the US tech sector?
I used both Google's speech-to-text APIs and Assembly's APIs as well as some other ones to build Twilio Voice phone calling applications. The out of the box accuracy was way better with Assembly and its far easier to quickly customize the language model for higher accuracy in specific domains (for example programming language keywords). Generally I avoid using Google APIs whenever possible since they always seem overly complicated to get started with and have incomplete documentation even when I'm working in Python which should be one of the better supported languages.
Twilio Developer Evangelist here. Our Super SIM Product Manager @matjaxon had his comments held up since he has a new HN account so I'm passing this along from him: "You could send the same information to and from your devices using either SMS Commands or data but it would likely cost you a lot more to use SMS Commands. While our prices start at $0.10 per MB, we bill you in bytes.
For example, if each time your device checks in you exchange 50 KB of data, you could do that 20 times for $0.10. You can only communicate with your device twice using SMS Commands at $0.05 per SMS Command. SMS Commands are often used as a way of configuring some IoT hardware solutions where you can set configuration values over SMS or to communicate with your device if it seems like something is going wrong with its data connection. It's an extra way of interacting with your device that may be deployed into the field thousands of mile away from you."
Sure, no problem. Tweet @fullstackpython or @mattmakai and I can follow you for DMs.
My open source side project Full Stack Python (https://www.fullstackpython.com/) hit front page over July 4th weekend in 2014 (https://news.ycombinator.com/item?id=7985692). I've continued to write and grow the site from ~5k users per month to over 100k/month now.
The traffic bump and feedback was motivating and helpful to know I was on the right track with my content. I also learned there are some comments you just need to ignore and focus on your own vision :)
edit: my traffic was lower than I originally remembered, it was ~5k per month, not 25k in mid-2014
I wanted to keep this blog post as simple as possible so even beginner Python devs could get through it in one short sitting. However, it's definitely possible to integrate the command processing piece with backend APIs. I'd first try using the TextBlob library before an external service though: https://textblob.readthedocs.io/en/dev/
My colleague Rob also wrote a post that uses TextBlog with a similar bot-type application over SMS: https://www.twilio.com/blog/2014/06/using-natural-language-p...
Very cool, I hadn't previously seen Limbo. I added a link to it on the otherwise empty Full Stack Python Bots page: https://www.fullstackpython.com/bots.html
If you're in SF or traveling here sometime soon, you should definitely present this at the SlackDevs meetup where a bunch of us API users get together each month: http://www.meetup.com/SlackDevs/
There's nothing like big blocks of time to get some writing and coding done. Full Stack Python (https://www.fullstackpython.com) got some major updates with new pages, additional sections on current pages and new links to resources. The change log [1] and commit log [2] capture what's new.
[1] http://www.fullstackpython.com/change-log.html [2] https://github.com/makaimc/fullstackpython.com/commits/gh-pa...
Nice work by you and your team so far Leah! Quick question - what's the rationale behind having the two "Page TBD" top links in the nav? Is that to make sure the CSS/design works using placeholders or do you keep them knowing there are groups within the FEC that will definitely need 2 spots on the nav bar?
Thanks, it's encouraging to hear that the contracting officers are now getting better advice from technical folks who understand the small teams approach.
What do you see is being done to combat the issue of big contractors (Booz Allen, CGI, etc) who don't want the gov't building software with small teams because it would hurt their ability to obtain huge billion dollar contracts?
My favorite part of Kate's talk is that she not only gives stories that illustrate the issues, but also that she gives specific tangible advice for steps we can take to get closer to solving the problems. Definitely worth a watch.
Congrats to you folks. I had my first Spotted Cow Wisconsin-only unfiltered beer when I stopped by your Madison offices while roadtripping the US in 2013. Awesome to see things are still going well!
I agree Miguel's tutorial is fantastic for a step-by-step walkthrough. He also has a video to cover much of the material: http://pyvideo.org/video/2608/flask-by-example
If you're going down the Python route I'll throw in Full Stack Python http://www.fullstackpython.com/ I wrote FSP to provide beginners with context as to what these terms like web server, WSGI server and platform-as-a-service mean without assuming you have prior programming experience.
Tom, congratulations! This is exceptional work. It's exciting to see this release come to fruition via your Kickstarter campaign and a whole lot of work on your part along with the Django community.
not lame at all! definitely a mistake on my part and should be fixed now. thank you!
thanks, I'll make the change based on this link. I'm coming primarily from a Django & Flask background so my description of Pyramid isn't from a position of knowledge. I'll have to write a Pyramid project to get some experience with the framework.
thanks, it's great to hear! I created the site because while I was learning Python these were the resources I would've loved to have learned from.
Regarding Pyramid, I completely agree. It's a great framework and definitely one I need to adda page for it as soon as I get my hands dirty with a project so I can confidently write about the differences compared to Django/Flask/Bottle.
I agree that engagement has decreased from Twitter since the removal of photo embeds, but from less technical folks I talk to they are using Instagram increasingly from Facebook. that appears to balance out the decrease in usage from Twitter