Wrote a two-part series on HNSW, the vector index behind fast similarity search. Mental model in part 1, Rust implementation in part 2.
Part 1: https://t.co/2TzAhXy1nq
Part 2: https://t.co/THaTCVtaca
#Rust#VectorDatabases#HNSW
Before going to sleep I set up multiple alarms on all my three devices, naa too much work. Just gave claude to build a centralised server which can set it up on all. Issue is ipad, I can't let my server set the alarms on it, cause native, 99$. No. I tested apple shortcuts, it can make http request, that's all I need.
Any other work around?
Built Alarmsync - set an alarm once, it rings on my iPad, phone, and laptop.
Any of the three can create or edit alarms for all devices, with a custom delay per device.
Apple Shortcuts are really useful. Otherwise, I had to build another client for iPad.
Before going to sleep I set up multiple alarms on all my three devices, naa too much work. Just gave claude to build a centralised server which can set it up on all. Issue is ipad, I can't let my server set the alarms on it, cause native, 99$. No. I tested apple shortcuts, it can make http request, that's all I need.
Any other work around?
Suggest more features. I have one - create MCP to connect with my claude, voice input, it will create alarms with all customisation per device in one go 😭😭
@aniket_ai Uggh, xcode requires macos, there are tools which let you build on Linux or on cloud, and then I plug in my ipad and install, but runs only for 7 days, too much work setting up all. Found another hack via shortcuts where I don't need an app on ipad, via email recieved event.
Sparse Retrieval, Part 2: SPLADE is here.
How BERT turns a doc into a sparse vector of weighted words, including ones the doc never uses.
A lot of this was new to me, tried to cover it as simply as possible. This part covers encoder architecture from tokens and embeddings to self-attention, the MLM head, and FLOPS regularization.
https://t.co/pSV3L4oSEm