Current
Medium Reading
Reverse Engineer Discord
Convergence in DL
Understanding Q,K,V in Transformers
You’re doing RAG wrong
From 100,000 to Under 500 Labels: How Google AI Cuts LLM Training Data by Orders of Magnitude
Writing a LLM based telegram bot that texts like a person would
alphago explained
distil vs quant ML
cosine similarity not a silver bullet - reddit
Other reading
Coding Architecture
Bloom filters: the niche trick behind a 16× faster API
- Read it, very dense but cool. Basically hashing JSONB data to make filtering at access time faster.
- Their website is cool, I want to read more articles
Keeping it boring: the incident.io technology stack
Spiking Neural Networks (SNNs)
https://pub.towardsai.net/the-complete-guide-to-spiking-neural-networks-d0a85fa6a64
https://pmc.ncbi.nlm.nih.gov/articles/PMC9313413/
Resume
Music
besame mucho musescore
let u break my heart again piano
Local LLMs
Neuro
Evolutionary constraints on visual cortex architecture from the dynamics of hallucination - Butler et al.
OCR
https://www.datacamp.com/tutorial/fuzzy-string-python
https://www.paddleocr.ai/main/en/version3.x/installation.html
finetune?
https://hub.docker.com/layers/paddlepaddle/paddle/latest-dev-cuda12.3-cudnn9.0-trt8.6-gcc12.2/images/sha256-afc69baadc57cc151953aa31426028f1f60a1371629056a3457d080316be6cbb?context=explore