One month has passed and the tabs are piling up. And what a month it was.
Very interesting find - this manifesto served as a cultural touchstone, reinforcing the company’s identity during a period of explosive growth. The article offering insights into how storytelling can effectively scale culture within a rapidly growing organization. But as last month has shown, a company does not exist in a societal vacuum.
Another enticing github repo to get my hands dirty. I find the comparison to a distro very apt: A Llama Stack Distribution (or “distro”) is a pre-configured bundle of provider implementations for each API component. I am very excited since it does support Amazon Bedrock as well.
This idea really hits the sweet spot for my interests in APIs and AI. It’s all about creating an open protocol to standardize how apps give context to LLMs. I love the analogy of MCP being like a USB-C port for AI apps. Just like with APIs, the value isn’t in the interface but in the service it connects, and having a standard way to link services to LLMs could be huge. I think the enterprise integration space is ready for smarter solutions, especially with how common machine-readable OpenAPI specs are for documentation. Very much related but for content sites is the llm.txt proposal circulating the web. There is even a curated llm.txt directory for some real-world examples.
Very much in the similar spirit is this article showing an real-world example of connecting a LLM to real world REST APIs, here Spotify and TMDB. The authors write that they are using the OpenAPI spec to organize the API call parameters on the fly.
I had to brush up on OAuth2 for a customer issue and found this awesome, super easy-to-understand primer on it. Definitely saving this one for when those annoying auth problems pop up again!
Another interesting twist in the agent patterns with a linked tutorial at the end on how to build an AI Email Assitant.
The author offers an interesting perspective on OKRs, suggesting they shouldn’t just track routine tasks. Instead of focusing on your day-to-day work, OKRs should push you toward new challenges and strategic projects. This made me rethink some of my own assumptions about how OKRs are used and how they should drive real progress.
Another self-paced guide on my list. Another one is this video tutorial Building LLMs from the Ground Up.
I still keep standing by my POV’s that LLM’s are currently a bunch of smart interns. But even if something like this can generate only 30% of a valid architecture, I see huge potential in it.