Welcome to the agent platform research briefing for Sunday, July nineteenth, twenty twenty-six.

SpaceX stock fell below its IPO price for the first time — closing at about one hundred twenty-four dollars on Friday, down five point four percent. That's below the one-thirty-five IPO price set in June. The drop follows Thursday's Starship Flight thirteen scrub, when four Raptor engines on Super Heavy booster twenty failed to ignite at T-zero. Musk says the next attempt will target a date next week. This is the first time SPCX has traded below its IPO valuation of about two trillion dollars at the peak — and it's a signal that the market is pricing in both Starship's engineering challenges and the unresolved FAA investigation from Flight twelve's booster cascade failure. Analyst Gary Black is going public saying SpaceX is still "ridiculously overvalued" and warning investors. Bottom line: the honeymoon phase of the SpaceX IPO is over, and the market is connecting rocket reliability to stock price.

Anthropic merged Chat and Cowork into a single unified Claude interface on July eighteenth. Previously these were separate environments — Chat for conversation and ideation, Cowork for autonomous task execution with file access. Now they share one home screen and a single message box with a mode toggle. The core change is architectural: the distinction between conversational and agentic workflows lives in the mode switch, not in navigating to different sections of the platform. It's available on web and desktop. This signals Anthropic's view that the line between "chatbot" and "agent" is blurring — and that most users should see Claude as a single tool that shifts between modes based on the task. It also puts pressure on OpenAI to unify its own ChatGPT surface, which still separates chat, file analysis, and Codex workflows.

Anthropic launched Claude for Teachers on July seventeenth — providing verified US K-through-twelve educators free access to premium Claude capabilities for at least a year. The package includes a library of teaching skills, connections to evidence-based curricula mapped to standards in all fifty states, Claude Code classroom access, and Cowork for autonomous lesson preparation. It follows Google's Gemini education push and Microsoft Copilot's inroads into schools. The timing is notable: this is Anthropic's first major education-specific product, coming just after the Fable five pricing resolution and the Chat/Cowork merger. It suggests a deliberate vertical expansion strategy alongside the core model battle. Critics in EdWeek raised concerns about data privacy and classroom surveillance implications — the same concerns that forced Meta to pull its Instagram Muse AI feature last week.

IETF is scrutinizing competing AI agent protocols at a Birds of a Feathers session in Vienna this week — and the outcome could shape the entire agent ecosystem. The agentproto BoF will evaluate whether to standardize agent-to-agent communication as an internet standard. On one side: MCP from Anthropic, donated to the Linux Foundation in December twenty twenty-five, with ninety-seven million monthly SDK downloads. Also in the mix: Google's Agent-to-Agent protocol, the Agent Communication Protocol, Cisco's Agentic Framework, and the Agent Network Protocol. Most companies building multi-agent platforms will read Thursday's session outcome as a signal — either the IETF creates an interoperability standard that overrides vendor choices, or the market-dominant protocols establish their current architectures as the de facto baseline. This matters because right now, if you want your AI agent to talk to another agent from a different vendor, there is no universal wire format.

That's the briefing for today. Four stories — SpaceX valuation reality check, Anthropic interface unification, education market push, and the standards battle for AI agent interoperability. See you tomorrow.