Great and very balanced blog post. I really appreciate the human aspect and your approach helping mentees becoming very good at what they - and have fun during the process. I always make it a point to hire for attitude, train for skills.
Thanks for the post Nathan! Despite having senior level applied scientist / MLE experience + a prior research background (not in LLMs), I nonetheless find myself anxious about how to move from doing applied work towards more research-oriented / fundamental work.
Any thoughts on how to find good open-source opportunities / communities to contribute to in order to make this leap? Do you think good open-source opportunities are also becoming scarce in the age of VC-backed, GPU-hungry, and mentorship-scarce AI industry?
Thanks for the post! Could you expand on your point of middle author papers being a negative signal? I assume, e.g., 3 first-author papers + 5 middle-author would be better than no middle author, but maybe 15 middle author would be worse?
When you start questioning “what do they actually do,” but yes the scenario you proposed is the gist of it. People need a clear list of why they did first.
Extremely informative. I’m new go this blog and glad I found it.
Welcome! Thanks for reading.
Great and very balanced blog post. I really appreciate the human aspect and your approach helping mentees becoming very good at what they - and have fun during the process. I always make it a point to hire for attitude, train for skills.
Really appreciate this post. Puts to words a lot of how I've been feeling on both sides of this.
Every day, I’m thankful that I’m retired and don’t have to look for a job in this crazy environment.
Thanks for introducing me to NanoGPT speedrunning, and reintroducing me to GPU Mode. Gotta get cracked 🏃🏻💨
Some great pointers for juniors
This was a helpful look into the other side.
This captures the ‘fog’ perfectly.
The point about senior impact compounding with agents is sobering and useful for career planning.
Thanks for the post Nathan! Despite having senior level applied scientist / MLE experience + a prior research background (not in LLMs), I nonetheless find myself anxious about how to move from doing applied work towards more research-oriented / fundamental work.
Any thoughts on how to find good open-source opportunities / communities to contribute to in order to make this leap? Do you think good open-source opportunities are also becoming scarce in the age of VC-backed, GPU-hungry, and mentorship-scarce AI industry?
Thanks for the post! Could you expand on your point of middle author papers being a negative signal? I assume, e.g., 3 first-author papers + 5 middle-author would be better than no middle author, but maybe 15 middle author would be worse?
When you start questioning “what do they actually do,” but yes the scenario you proposed is the gist of it. People need a clear list of why they did first.
Makes sense. Thank you!
So it sounds like:
1) no one ever cancelled obsession with learning new things
2) trying new ways to do various things
3)continue to build systems
4) don't get very discouraged in this age of fast AI development even though coding has become really easy to do.