Philip Nuzhnyi @callmephilip.com · Jun 25 Software is eating the world. 2.0 is eating 1.0 (e.g. neural nets replacing c++ code in tesla autopilot for information stitching together functionality). 3.0 is eating 2.0. Devs will now have to juggle all three and switch around when needed
0 likes 1 replies
?
Replies
LLMs have props of utilities: - CAPEX to train LLM - OPEX to serve (inference) - Access it metered ($/1M tokens) - Demand for low latency, high uptime, quality and consistency - Switching functionality (via OpenRouter) - When LLMs go down, intelligence "brownouts" occur (like an electric blackout)