Daily Reading List – January 23, 2025 (#478)

First trips of the year are coming up. Hopefully I’ll see many of you at one of these stops. I’ll be in Sunnyvale next week, India in February, Paris/Istanbul in March. Then Vegas for Cloud Next in April. Unless something changes, which is always entirely possible 🙂

[blog] 2024: A year of extraordinary progress and advancement in AI. Wow, what a post. I’m seeing so much cool AI work happening everywhere. And a robust amount is happening within Google.

[blog] SmolVLM Grows Smaller – Introducing the 250M & 500M Models! Hugging Face shipped a pair of small vision language models that perform well and can be run in resource constrained environments. Very cool.

[article] Google Cloud links poor credentials to nearly half of cloud-based attacks. If you want an instant change to your security posture, get serious about credentials! Weak credentials is how many attackers are getting into your environments.

[blog] How do large language models get so large? Educational post if you’re not entirely clear what a “model” is made up of.

[article] Employees See Middle Managers as an Organization’s Moral Compass. Managers don’t get a lot of glory. They get blasted as the “frozen middle” or blockers to change. Some may be, but many are heroes wthin an org who get things done and shape team norms.

[blog] Understanding Google IAM (Identity and Access Management) and Best Practices. Credential management and role based access matters, as mentioned above. This is a deep dive into our cloud’s IAM service.

[article] More AI, More Problems for Software Developers in 2025. Every solution creates different challenges. AI will be no different. Don’t skip it because of that. The benefits will grossly outweigh the cost if done well.

[blog] Model fine-tuning made easy with Axolotl on Google Cloud Batch. Outstanding post from Karl that walks us through a solution that makes it simpler to fine-tune an open LLM.

[blog] Hey Ollama, how about running on Vertex AI? I like when folks can’t find existing walkthroughs for a given scenario, so they write their own.

[blog] Scaling LLMs with Golang: How we serve millions of LLM requests. Sure, Python rules in the LLM world. But this post shows off three features of Go that make it a great choice for AI apps.

Want to get this update sent to you every day? Subscribe to my RSS feed or subscribe via email below: