Nathaniel Whittemore on The AI Daily Brief declares the second quarter of 2026 the start of AI’s ‘second moment’ - the pivot from assistants to autonomous, workable agents. The evidence is in the revenue. Claude Code’s annualized run rate jumped from $1 billion to $2.5 billion in just two months.
Enterprises are buying agents to replace people, not just software seats. Block cut 40% of its staff as a portent, and Pulsia, a fully agentic company, hit $6 million in revenue with one founder. According to Ramp data cited by Whittemore, Anthropic now captures 70% of new enterprise AI buyers, leaving OpenAI with 25%.
"AI’s first moment was the chatbot. The second is the workable agent."
- Nathaniel Whittemore, The AI Daily Brief
This aggressive deployment creates a novel security crisis that legacy tools cannot solve. On No Priors, Onyx Security CEO Maxim Bar Kogan explains the context gap. If an AI agent is authorized to manage a database, a traditional security tool sees a ‘delete’ command as legitimate. It cannot distinguish between a requested cleanup and a hallucinated catastrophe, leaving companies one reasoning loop from a total data wipe.
The structural solution isn't another frontier model. Supervising a $1 task with a $2 guardian model is a financial non-starter. Bar Kogan’s architecture uses small, specialized ‘not smart’ models to flag anomalies in real-time, only invoking a powerful guardian for high-stakes decisions - a blitz chess approach to security.
A deeper trust barrier locks out the model labs themselves from providing security. Enterprises fear handing historical interaction data back to OpenAI or Anthropic for auditing, worried it will fuel future model training. Bar Kogan compares labs to car dealers - you don’t ask the seller to be the sole safety inspector.
"Structural independence is the primary hurdle for labs like OpenAI or Anthropic in the security space."
- Maxim Bar Kogan, No Priors
The national security rift has shattered any remaining neutrality. After Anthropic refused Pentagon demands to allow Claude’s use in autonomous weaponry, the Defense Secretary designated the U.S. firm a supply chain risk. OpenAI immediately signed a deal with the Department of War. The public reaction was tribal: ChatGPT saw a 775% surge in one-star reviews, while Claude hit number one in the App Store.
The enterprise AI stack is now bifurcating: agents from one set of vendors, security from another, all under the shadow of a geopolitical standoff.
