programming

The differences between the APIs of numpy, cupy and torch are so fascinating....

```
>>> import torch as np
>>> np.arange(10)
tensor([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
>>> a = np.zeros(20)
>>> a[19:-1:-2] = np.arange(10)
```

Torch doesn't let you have ranges that go backwards 🤣 

Would love more consistency

Ponder is now 10-50x faster during historical indexing

0.4.37 introduces batched database writes & an in-memory cache for frequently accessed records

Apps that use a remote database will be much faster

Make it run
Make it right
Make it fast     ←

New way to encode a profile picture dropped:

https://x.com/Ethan_smith_20/status/1801493585155526675

320 bits is basically a hash. Small enough to go on chain for every user.

AI as lossy compression. With a big model you might be able to do quite a lot with a few bits.

You could maybe stack this on top of ZK for extreme levels of onchain compression and offchain reconstruction.

Ho🎭🔮🍖🍌⚡️🎩🔥

@giveup