🕶️ Pentagon to Anthropic: pick a side

A Google startup lead says the era of “LLM + simple UI + Stripe button” startups is ending: companies that just wrap GPT or Gemini in a basic interface, or act as middlemen routing users to the “best model,” will struggle as model providers add those features themselves.

As OpenAI, Google, and others ship native tools and multi-model access, margins shrink, switching gets easier, and power shifts back to the platforms—similar to how AWS squeezed resellers. The startups that will survive are those with proprietary data, deep workflow integration, strong vertical focus, or real technical infrastructure; simply aggregating and marking up APIs is unlikely to be defensible, and consolidation is expected as investors favor businesses with real moats.

You have a startup idea.
It has billion-dollar potential.

But how much is it actually worth right now?

In this episode, Caya breaks down how early-stage startup valuations really work; and why most founders think about them the wrong way. We explain pre-money vs post-money, how dilution actually happens (bricks vs pie charts), why most pre-seed rounds translate to 15–20% equity, and how raising more money can paradoxically make your company “worth” more.

We also walk through real 2025–2026 funding benchmarks, typical seed round math, stock option pools, and what happens to founder ownership over time. By the time most startups reach seed, founders typically own around 56% combined, and that’s normal.

This isn’t about chasing unicorn valuations.
It’s about not losing control and understanding the math before you negotiate.

Defense Secretary Pete Hegseth has called Anthropic CEO Dario Amodei to the Pentagon over limits Anthropic placed on how the military can use its Claude AI model. The Pentagon wants broad rights to use AI for lawful military purposes, including weapons and surveillance, but Anthropic pushed back. Now Hegseth is threatening to label the company a “supply chain risk” — a designation usually reserved for foreign adversaries — which could jeopardize a reported $200M Defense Department deal and force contractors to remove Claude from their systems. Claude was reportedly used in a January raid that captured Venezuela’s Nicolás Maduro, highlighting how high the stakes are.

If that designation goes through, AI companies seeking DoD contracts may need fully on-prem, air-gapped, and tightly controlled systems to avoid sudden bans. Competitors would likely rush in with government-compliant offerings, while procurement teams factor political risk into their vendor decisions.