from @AravSrinivas (perplexity CEO): hybrid agentic inference is how we maximise token value per watt per user local model on intel core ultra series 3 -> handles confidential files

@tengyanai·2026年6月2日