Anyone who thinks they can keep a few GB of weights from getting out into the public is sorely misguided.
I find it fascinating how training is so hard, but inference is so trivial.
Smithers 0.23.0 is here
Everybody is talking about loops. Smithers users are already doing them
This release ships a brand new docs site, usage quota features, and quality of life features
1/ 🧵
I understand (kind of) why they do it, but I feel like it is a huge missed opportunity to not allow attendees to ask questions at ETHConf. So I'll use this tweet to ask questions. Hopefully get some answers.
Question for Taylor Lindman (wisely not on X).
The primary advantage of blockchain is the removal of trust from the equation for financial services.
x.com/EricForgy/status/20477…
For certain simple applications, it is reasonable to start out 100% decentralized on day 1. This is not realistic or even desirable for new protocols that have not been battle tested yet. Do you see a path for protocols to start out with some elements of centralization as long as there is a path to full decentralization without stifling innovation. For example, maybe an innovation sandbox?
One reason I understand why attendees aren't allowed to ask questions is that not all questions are softballs.
Question for @StaniKulechov.
A few weeks ago @aave was dead. It was saved by a community bailout. Bailouts are bailouts whether they are governments or communities. The end result is the same: morale hazard. A few days after KelpDAO, rates on Aave were essentially back to where the were before.
What are your thoughts on morale hazard?
Remembering Ryan Porter. Such a master of the trombone:
open.spotify.com/track/3t4DP…
Trivia: I used to play jazz trombone and recorded an album (cassette tape!) in an old school recording studio back in the day.