From coding with LLMs → to systems that improve themselves.
Just explored AutoResearch by Andrej Karpathy — an idea where agents run experiments, test results, and keep getting better on their own in a loop.
Read this blog post :
https://t.co/vMetAug0FY
@krunalbuilds Yeah. Like in the case of cache miss, we should go with local handling. And in the cases like db fails unexpectedly, then we should propagate the error to the global handler.
In the multi-modal RAG I am building, for now, considering only PDFs as the source files.
Completed the first step : PRO PARSING the pdf.
Tables and images are then passed through vision models to generate meaningful captions.
Any tips guysss...
Building a multi-modal rag pipeline. I personally want to explore financial reports of companies, since I have a sole interest in trading and investing.
Let see how it goes.
Any tips guysss......
@amitkilhor Though this would be a great start, I think the real factor is exposure. It is the exposure to exceptional teachers, peers, ideas, and academic environments that helps really.
"A model that remembers everything eventually remembers nothing useful."
Suggest some methods for context compaction for systems like CLI based coding agent.
https://t.co/KQ7T3c4CiJ