Personally, I think fundamental knowledge of IAM should be a requirement for any engineer working with a manor cloud provider, and prefer IAM access/tokens to creating shared secrets and API tokens all over the place - but that isn’t the case for everyone.
Obviously, not everyone is accessing LLMs and other models through their cloud provider. However, I find it super intriguing that authentication to LLMs for folks on providers like Google Cloud could introduce additional mental burden.
One of the most interesting blockers I see to LLM/AI adoption at scale is the understanding of IAM-based access.
At least once a week I get asked for a Gemini API key, and when I reply with instructions on authenticating to Vertex AI with IAM - I’m often met with confusion 🤔
*my dad walking by the roomba*: “hey Romy!”
Burning all the OpenAI credits to fund of swarm of agents consuming fictitious marshmallows non-stop
It’s always the Kyles
Conway’s Law: something you periodically mention during the architectural design process to sound smarter
Wow I never noticed that either that’s awesome! But no worries it is also hilarious
I use comic sans everywhere so that it’s less serious when I do anything chromewebstore.google.com/detail/comic...
A couple weeks ago my brother-in-law asked if there were any apps that would let him read the NYT without any Tr*mp articles. I didn't see anything I liked, so I made something: https://icanteven.app (it'll actually work on any site).
Things to do with a book you co-authored
Latest Etsy Code as Craft article - this time, highlighting my team's work from this year's CodeMosaic on online, stateful machine learning! Our team, the So-Fine Real-time ML Paradigm (conveniently abbreviated SoFiReTiMaLe) piloted some really cool stuff www.etsy.com/codeascraft/...
Here's a promotional code for a free 30-day trial of the O'Reilly learning platform - use it to read a ton of books including The Machine Learning in Python Cookbook! Which I wrote! Some of! https://learning.oreilly.com/get-learning/?code=MLPCOOKBOOK23
It literally feels *so good* to have the Machine Learning with Python Cookbook going to print with @OReillyMedia after a year of asciidoc, and the npm library I found to turn them into Python notebooks. Buy it!
O'Reilly: lnkd.in/eH_6b2tQ
Amazon: lnkd.in/ezm9mWur
How do I view my liked skeets
I sure hope not?
weekend idea:
double feature
What AI people worry about: Releasing model weights means AGI comes to kill us all
What AI people should worry about: leftpad
Hot take: managing Python envs isn’t hard
Me when I update min replicas from 1->2
me when i scp data into my ec2 micro instance
Changed my weather app to Celsius awhile back so I’d feel smarter in other countries - and I’m just now realizing that I’m way more scared of the hot 32 degrees than the cold 32 degrees
Whether or not there’s a Makefile at the root of a repo tells me so much about a person
I love writing it makes me happy https://www.etsy.com/codeascraft/the-problem-with-timeseries-data-in-machine-learning-feature-systems
Copilot helping me with Python: 💪🏼
Copilot helping me with Scala: 🤷🏻♂️
I’m tired
Lemme know how it goes ❤️
Oh hey Vicki what’re your thoughts on web3?
