I’m very excited to announce that I’ll be joining Perceptron AI (perceptron.inc?) as a researcher and founding member of the technical staff. I’ll be working with Akshat Shrivastava and Armen Aghajanyan to create the world’s first visual language foundation models specifically

I no longer find Cursor's tab auto completions to be very useful now that I'm working outside of legacy tech company codebases...I think it's relying too much on the underlying API model having been trained on the external libraries being imported (e.g. pyspark), better RAG

My first and only Twitter beef in my life was with that certain someone who hacked his prompts to top the benchmarks and then wouldn’t cooperate with open source researchers to reproduce them haha 🤔🤔

Anyone else finding that GPT4.5 has a hard time following instructions and has a tendency towards being lazy, i.e. doing only part of a task then asking you to finish it yourself?

Great work from the Databricks research team. The combination of the AI gateway + ground truth-free training methods is such a powerful tool for adapting models to real world applications

fun debugging journey w/Akshat Shrivastava: be careful around FP8 w. activation checkpointing activation checkpointing works under the assumptions that different calls of forward give similar results which we move away from the more we quantize. when you re-quantize in activation

🤦‍♂️this is my quant