Data + AI Weekly Round Up 02
Things I've read and seen in the world of Data and AI
Hey y’all, I’m Hoyt!
Each week, I share news and findings in the world of Data and AI. I span the entire data stack and am trying to help seasoned data professionals move in to the new world of Data + AI.
Not Subscribed? Come join! ⊂(◉‿◉)つ
Happy Friday!
Here is the second installment of the Data + AI weekly round up. The beginning of the week was a little quiet and I began to wonder if I could actually get enough stuff to write about. That all changed within a couple days and WE HAVE A LIST!
Interesting things I found this week in Data and AI:
Something called ChatGPT5 came out yesterday by OpenAI (never heard of em). Jokes aside, the reviews have been mixed so far. Certain dev influencers like Theo were blown away by it while Andriy Burkov was not impressed.
I used to think that LLM Evals were just jargon by Product Manager's trying to give themselves work. But the more I've looked into it, the more it makes sense. There's human evaluation and then there is using an LLM to evaluate an LLM output. It still kinda confuses me in the sense that I don't know how it really scales, but I am starting to come around to it. This FAQ up by Hamel Husain is probably the best long form explanation I've ever seen on the subject.
ABC (@Ubunto on X) is a fierce first adopter of AI technology in Data Engineering. I found this X post giving me the best breakdown of what an MCP server needs to perform on multiple databases.
The resurgence of Semantic Layers as a useful tool with Data + AI has given Looker new life it feels like. Gemini has been added to it to allow for conversation analytics without needing to understand LookML. Maybe that will bring the tool back from the dead.
A
post on Substack let me know that Pew Research concludes that people don’t click on links as often when there is an AI response at the top. It seems obvious, but also trying to understand how Google eats its own tail and comes out on top with this one.I thought this one was really interesting. A guy is tracking the diffs in Anthtopic’s internal prompts for Claude Code. You can select different versions to see the difference. Just cool to see how it has changed over time to address different things.
I was actually more excited about a different model drop from OpenAI. Specifically they offered two open weight models that can be run on a single GPU for the big model or consumer hardware on a single machine for the small model. Both compare to o4-mini which is a model I love to use with my AI applications since I like to use AI in small use cases. But I wanna check these out in action.
I love seeing write ups about AI that aren’t just technical and theoretical but practical and being used in real life. So I enjoyed reading about a company’s GTM strategy using AI from GTM Strategist.
A ton read this week so have fun and learn along the way! I’m sure things are just getting started in the Data and AI ecosystems. Something is in the air!
Thanks for the shoutout!