As with LMs, modern Diffusion models rely heavily on Attention. This improves quality but requires patching to scale. Working with Apple, we designed a model without attention that matches top imagenet accuracy and removes this resolution bottleneck.
https://t.co/E5MvNxgJsO
@buffalu__ Tbh this is a supply chain attack mitigation. If we relax the “=“, we get a bigger atacking surface given usually 1000+ dependencies for Sol project. I suspect there are already teams trying to attack.
@Davyegld @SolBulletin @solana I feel “diversity” might be the metric we care about. E.g. what’s the average (and median) # of tx for an address daily, no matter bot or not.
We hope you enjoyed this thread on confidential transfers. Confidential transfers are new and exciting technology, but critical to get right the first time.
Lastly, for more in-depth Web3 security research, check out our blog! https://t.co/fVM4Kz2NTp
@raggedsec Sol dependency is in hell mode due to these projects pin solana-* and spl-* crates too tight, often use ~ or even = in the version requirement. To make it worse, cargo error message is very obscure.What is do finally is forking every projects and manually align their dependencies
If you have enjoyed this explanation, share it with your friends and give me a follow! I regularly post deep-dive explainers such as this.
Understanding mathematics will make you a better engineer, and I want to help you with that.
https://t.co/HgMOIyRBPT
https://t.co/8OyNlMnHZR
https://t.co/SJLYfUOiyl
For the first one, we all thought it was a bad coder. But thrice in a row.. seems like an oracle attack. @switchboardxyz, you guys using on-chain data for oracle price feed, right?