Latent Space (Swyx) analyst · 23h ago · 2/5
[AINews] Tasteful Tokenmaxxing ↗
Google introduced eighth-generation TPUs split into TPU 8t for training and TPU 8i for inference, claiming 3x compute per pod and ability to scale to a million TPUs in a single cluster. The company also launched Gemini Enterprise Agent Platform with Agent Studio, 200+ models via Model Garden, and integrations across Workspace products. Across the industry, OpenAI, Google, and Cursor are converging on cloud-hosted agent architectures with shared team context, approvals, and long-running execution, while developers gain model flexibility through bring-your-own-key support in VS Code.
you want to go for depth (e.g. do more serial autoresearch loops) than go for breadth (e.g. solve a problem by kicking off 5, 10, 50, 500 parallel runs of the LLM slot machine).