Google Promoted Agents to Infrastructure Primitives. The Runbook Is a Slack Thread.
Cloud Next '26 gave AI agents the same infrastructure status as containers. The decade of operational discipline that makes containers reliable? Not included. $750M committed, zero Agent SRE playbooks written.
Anatomy of a $750 Million Incentive Machine
A forensic reading of Google's $750M agent ecosystem fund reveals four payment triggers — integration delivery, certification, prototyping, embedded engineering. None require measured business outcomes. Here's what that means for your next agent contract.
OpenAI's Biggest Competitor Is OpenAI: How ChatGPT Eats Its Own API Developers Alive
Three updates in eight days. Each one killed another reason to build on OpenAI's API. The math is brutal.
The Browser-Agent Oligopoly Nobody Voted For
Browserbase, Browser Use, Anthropic quietly own the layer that lets agents click the web. Your stack already picked one.
Tool-Calling Is Dead. Agents Now Write Code.
In 48 hours OpenAI, Anthropic, and Cloudflare made code mode the default. JSON schemas are the deprecated path.
The Checkpoint Gap: Multi-Hour Agents Shipped Before Crash Recovery Did
Anthropic and OpenAI shipped long-horizon agents in 8 days. Nobody shipped crash recovery. The cat explains the bill.
MCP's Missing Half: Why Auth Brokers Became April's Hottest Unsexy Layer
MCP standardized tool calls but punted on per-user OAuth. Composio, Arcade, Pipedream are racing to be the Okta of agents.
Every Agent SDK Ships a Runtime. None Ships the Tests.
Anthropic, OpenAI and LangChain shipped agent runtimes in April 2026. None shipped an eval framework. Devs pay the tab.
Every Text Your AI Agent Reads Is an Unsigned Command
Five MCP CVEs in two weeks, 475 AI-generated malicious PRs in 26 hours, and zero reliable defenses. Prompt injection meets production agents.
The B-Sides: California's AI Power Grab and NVIDIA's Open Robotics Blueprint
Capitan, Nero, and Schnapps unpack two stories that didn't make today's show: California's SB 7120 giving state regulators veto power over frontier model deployments, and NVIDIA open-sourcing its entire robotics simulation stack under Apache 2.0.
Two Leaks, One Company, and an $852 Billion IOU
Anthropic leaked twice in ten days, OpenAI raised $122B at an $852B valuation, and MCP quietly won the protocol war. Seventeen stories, one thesis: who controls the infrastructure layer controls AI.
Google Gave Away the Farm — Gemma 4, Apache 2.0, and the Art of Strategic Generosity
Google released Gemma 4 under Apache 2.0 — no restrictions, 400M+ downloads, ranks #3 globally. But free models from a cloud company aren't generosity. They're a funnel. And the open-source community is giving Google something even more valuable than cloud customers.
The Locksmith Built the Lockpick
Anthropic's Claude Mythos — a 10-trillion-parameter model with elite cyber capabilities — was exposed through the same operational failures that leaked Claude Code's entire source ten days later.
Power Lives in the Pipes
Capitan hosts Raven, Mossy, and Compass to debate who actually controls AI's infrastructure layer — and whether open source, custom silicon, or regulation can break the concentration.
MCP Will Be the .deb vs .rpm of AI by September
MCP hit 97 million installs and got handed to the Linux Foundation. History says what happens next: the standard survives, but the implementations diverge.