John Berryman’s Post

I wrote a short story about AI 8 years ago. I just found it and decided to update it. Want to take a read and make comments before I publish it? https://lnkd.in/edc7BZsT (feel free to comment in the document)

This Claude 3.5 Sonnet response is fascinating. It got the answer wrong, but it caught itself and attempted a correction. This is the first time I've seen this behavior. Is it: A) Emergent behavior from a smarter model? B) Creative training to make models realize mistakes?

LangChain, DSPy, etc. make it easy to abstract away the actual LLM so that you can easy try different LLMs. But I don't want the whole framework. Is there anything out there that _just_ abstracts away the LLM so that I can swap out an Anthropic model for OpenAI, etc?

Prediction: in the year 2030 there will be 6 varieties of code: 1. Python written by LLM 2. Rust written by LLM 3. Golang written by LLM 4. JavaScript written by LLM 5. Java written by LLM 6. COBOL written by people that died in the 20th century

So I'm not saying that I'm starting a consulting company for prompt engineering and LLM application development. But if I was, what would be a good name for it?

I am so excited to see Copilot Extensions hit production. I think it's the start of a huge shift in how users interact with GitHub, and it's also the first example of a huge shift in how customers interact with ALL online applications.

Why, dear community, did we call it RAG? Retrieval Augmented Generation? Wouldn't it be better, cooler, and more accurate if we had called it RAP? Retrieval Augmented Prompting 😎

Check out my new office setup. I'm using a Breaking Bad motif.