aleggg

6 ms·

1. 1.▲
   by aleggg 9mo ago

   Yes. This heavily subsidized LLM inference usage will not last forever. We have already seen cost cutting for some models. A model starts strong, but over time the parent company switches to heavily quantized versions to save on compute cos
2. 2.▲
   by aleggg 10mo ago

   GLM-4.5-Air AWQ Q4 is fantastic all around (including coding), and can run on 4 RTX 3090s easily.
3. 3.▲
   by aleggg 1y ago

   I definitely do not make a living from it, but after two years, my Bitcoin on-chain data and metrics API service has started to generate some revenue. It's basically a service where you can download highly processed Bitcoin data in CSV
4. 4.▲
   by aleggg 2y ago

   NOSTR!
5. 5.▲
   by aleggg 5y ago

   Blue light blocker glasses to wear at nigth. Changed my life in so many ways. I thought I slept well before, but boy I was so wrong!
6. 6.▲
   by aleggg 6y ago

   What does the "Satoshi Nakamoto 1A1zP1eP5QGefi2DMPTfTL5SLmv7DivfNa" on page 1 means?