How Headless Agents Will Change Work
Summary
This episode covers the intensifying compute race, highlighting OpenAI’s 30GW target and Google’s new dual-chip TPUs for training and inference. The main discussion explores "headless agents," analyzing how Salesforce, OpenAI, Microsoft, and Google are adapting enterprise software for autonomous AI users. The host examines the shift from per-seat to consumption-based pricing and the emerging opportunities for agent-native infrastructure and consulting services.
Topics discussed
Intro and show announcements
OpenAI triples compute goals to 30GW
Energy bottleneck and GE Vernova earnings
Dedicated inference chips and AI-generated code
OpenAI releases privacy filter model
Mistral potentially joining xAI
Sponsors: KPMG, Blitzy, ZenCoder, Granola
Salesforce announces Headless Agent Platform
OpenAI launches Workspace Agents
Anthropic, Microsoft, and Google agent updates
Shift from seat-based to consumption pricing
Future of SaaS and headless infrastructure
Listen ad-free on Castria