Barrett Studdard

Exploring all things data science, machine learning, visualization, and BI. Opinions are my own.

[Barrett Studdard @Data_Stud]

@misraetel Inevitable once AI handles more pieces of self improvement/training cycles independently.

[Barrett Studdard @Data_Stud]

@joelgrus @Adversarial_L Really excited for #4!

[Barrett Studdard @Data_Stud]

@hwchase17 @thdxr And what is the preferred alternative

[Barrett Studdard @Data_Stud]

@langchain Any langraph devs in the St. Louis area interested in organizing something?

[Barrett Studdard @Data_Stud]

@alexdupler Very cool, love this idea.

[Barrett Studdard @Data_Stud]

Check out my most recent article - using large language models to generate podcast episodes.

https://t.co/qiuB4mx2FX
@langchain & @elevenlabs libraries doing a lot of the heavy lifting here.

[Barrett Studdard @Data_Stud]

@creeeeeeeeeeg @pathofexile Right on, I think im going to try tactician mercenary this season

[Barrett Studdard @Data_Stud]

@creeeeeeeeeeg @pathofexile What build are you playing this season?

[Barrett Studdard @Data_Stud]

@joelgrus Don't give in 😅 Apps are ram hogs nowadays, probably need 32gb

[Barrett Studdard @Data_Stud]

@marcorus It would however be interesting to have something to allow llms to interact with pbi semantic models (think API or MCP released by msft). This would allow both scripting and custom llms to interact with a semantic model outside of the predefined chat interface.

[Barrett Studdard @Data_Stud]

@marcorus Great points, thanks for sharing. I played around with something adjacent (llm dax generator) prior to the msft pbi copilot launch. Overall, like the podcast mentioned, not going to be able to build something as competitive as Microsoft.

https://t.co/j8pOdRQR9Y

(1/2)

[Barrett Studdard @Data_Stud]

@karpathy Then another meta layer (reflect on the day's reflections) to identify opportunities to add to the global system prompt.

[Barrett Studdard @Data_Stud]

@karpathy I was skimming through Claude's system prompt today as well. Perhaps some type of self reflection - given an answer/chain of thought and either direct user feedback or inferred feedback based on follow up questions, infer summary heuristics about the situation and response.

[Barrett Studdard @Data_Stud]

@tiangolo @FastAPI FastAPI is awesome, well deserved.

[Barrett Studdard @Data_Stud]

@tunguz Yes

[Barrett Studdard @Data_Stud]

@LiorOnAI Been playing with the same, yeah it's awesome

[Barrett Studdard @Data_Stud]

@aaditsh Great book

[Barrett Studdard @Data_Stud]

Exploring MCP python servers with a PydanticAI client.

https://t.co/4rxPtgEa5P

[wordgrammer @wordgrammer]

“Okay Claude, I just added $2000 to my account. Please spin up five agents to download Cuda”

[Barrett Studdard @Data_Stud]

Playing around with some PydanticAI abstraction around streaming. The library and API are fun to explore and work with so far.

https://t.co/KNMn50lbfw