Uploaded July 2025 | Updated November 2026, 3 hours ago
Everyone talks about GPU costs in AI, but where is the real compute going?
UC Berkeley Professor and RunLLM Co-founder Joey Gonzalez explains how the innovation frontier is shifting—from massive pre-training to more iterative, accessible post-training techniques. He breaks down what makes post-training expensive, what’s GPU-bound (and what’s not), and why more teams should start experimenting with it now.
📺 Watch more LLMs on the Run:
Playlist → tinyurl.com/msxmhph7
🔮 Subscribe for more AI insights:
tinyurl.com/445y4sd3
🤖 Explore RunLLM, the AI Support Engineer built for advanced technical support: tinyurl.com/43vy999s
#LLMsOnTheRun #PostTraining #AITraining #LLMs #FineTuning #AIResearch #RunLLM #GPUCosts #RLHF #AIInfrastructure
Everyone talks about GPU costs in AI, but where is the real compute going?
UC Berkeley Professor and RunLLM Co-founder Joey Gonzalez explains how the innovation frontier is shifting—from massive pre-training to more iterative, accessible post-training techniques. He breaks down what makes post-training expensive, what’s GPU-bound (and what’s not), and why more teams should start experimenting with it now.
📺 Watch more LLMs on the Run:
Playlist → tinyurl.com/msxmhph7
🔮 Subscribe for more AI insights:
tinyurl.com/445y4sd3
🤖 Explore RunLLM, the AI Support Engineer built for advanced technical support: tinyurl.com/43vy999s
#LLMsOnTheRun #PostTraining #AITraining #LLMs #FineTuning #AIResearch #RunLLM #GPUCosts #RLHF #AIInfrastructure










