Small weights that pack a punch. Very parameter efficient
Low key: we pack weights well.
Core Automation
104 posts
Core Automation
@CoreAutoAI
Building systems to automate and optimize
- Maybe even more than thatReplying to @MatthewJBarI think you’re wrong and there’s 1,000x efficiency gains leftover in deep learning research that could lead to much smarter faster more agentic models given the same inputs
- The path has been clear all alongStep 1: Buy a sh*tload of GPUs Step 2: ? Step 3: Profit
- Research is like poker
- When we say "full stack lab" we mean shortest path of feedback from megakernel microcode to agentic swarm and back. Homegrown and integrated
- Every single neuron is weak. But together they constitute a force to be reckoned with
- Today is a good day for vibecoding
- You can fit a lot of automation in this one
- Replying to @CoreAutoAII will take "what is the best method to learn complex programs from data from a perspective of hardware-level acceleration" for 500
- If deep learning is the solution what is the problem?
- BeautifulWe will free researchers of systems so they can dream of research again - illustration by @SaroufimSarah
