# Why AI Agents Are Getting Their Own Infrastructure

URL: https://zunkireelabs.com/blog/ai-agents-are-getting-their-own-infrastructure/
Published: 2026-10-02
Summary: Agent swarms for security testing, agents inside messaging apps and calls for an AI-native OS: what this week's agent news means for businesses.

**In short:** AI agents are moving from demos to infrastructure. This week's news shows money flowing to the tools that test, deliver and connect agents, so businesses should plan for agent-friendly systems, messaging channels and human oversight, not just a chatbot.

## Key Takeaways

- Agentic AI takes actions toward a goal; a chatbot only answers questions.
- Armadin raised $255.5 million to run always-on agent swarms that test enterprise security.
- Photon raised $4.5 million to build agents that work inside messaging apps such as iMessage and WhatsApp.
- Airbnb's CEO says the industry still lacks a true AI operating system, and that apps and agents need to become interoperable.


## What Is Agentic AI, and Why Is It in the News Again?

**A simple definition:** agentic AI refers to software that does not just answer a question but takes actions toward a goal. It plans steps, uses tools, talks to other software, and works with limited supervision. A chatbot tells you how to book a flight. An agent books it.

Agents have been promised for a while. What changed this week is where the money and attention are going. Three separate stories, all reported on October 1, point the same way: the interesting work is shifting from the agent itself to the **infrastructure around agents**. That means how agents are tested, how they reach people, and how they connect to the apps and services businesses already run.

## Story One: Agent Swarms Are Being Used for Security Testing

According to [TechCrunch](https://techcrunch.com/2026/10/01/kevin-mandias-new-agent-swarm-security-startup-armadin-raises-255-5m-at-2-5b-valuation/), Kevin Mandia, best known as the founder of Mandiant, has raised $255.5 million for a new startup called Armadin at a valuation above $2.5 billion. The Series B round was led by Andreessen Horowitz and Accel, and it came just six months after a $190 million Series A, bringing total funding to more than $445 million.

What Armadin sells is the interesting part. Traditional penetration testing means hiring people to try to break into a company and report the weaknesses they find, usually at a point in time. Armadin instead runs **always-on agentic swarms** that chain vulnerabilities together to break in, with the aim of helping organizations find and fix holes before attackers, or even misbehaving AI agents, can exploit them.

The takeaway for business owners is practical. If attackers can use agents, defenders will need them too, and testing that happens once a year may stop being enough.

## Story Two: Agents That Live in the Apps People Already Use

The second story is about distribution. TechCrunch reports that a startup called [Photon held a literal funeral for mobile apps](https://techcrunch.com/2026/10/01/photon-held-a-funeral-for-mobile-apps-now-it-has-4-5m-to-help-replace-them-with-agents/) in San Francisco on September 17, and has raised $4.5 million in seed funding. Photon helps developers build agents that work over iMessage, WhatsApp, Telegram, SMS or RCS, email, and voice, using a unified API, a channel framework, a command-line tool and an observability suite.

The company says it has signed up more than 40,000 developers and grown revenue ten times in four months. Those are the company's own figures, so treat them as claims. Photon's CEO, Daniel Tian, also admitted the funeral was somewhat tongue-in-cheek and said the shift away from apps will take a while, though he believes it is "directionally inevitable."

The underlying idea is simple and worth taking seriously: people are tired of installing new apps, so a business may reach customers better through an agent inside a messaging app they already open every day.

## Story Three: Do Agents Need Their Own Operating System?

In a [TechCrunch interview](https://techcrunch.com/2026/10/01/brian-chesky-interview-ai-agents-need-their-own-operating-system/), Airbnb co-founder and CEO Brian Chesky argued that the industry still lacks a true operating system built for AI. Today, he said, people are building AI apps on top of iOS, macOS and Windows, but those are not "AI operating systems."

Several of his points are useful for anyone planning an agent project:

- **Chatbots are not the right interface for everything.** He said they are poor for browsing and shopping because they show a few options at a time and need many turns to reach a result. Airbnb launched AI-powered search this week as part of its fall update.
- **Apps may become agents.** He described a future where an app's sections each have an agent, with an eventual "macro" agent for the company, connected to other agents through MCP, a standard that lets AI agents connect with other software.
- **Interoperability is the prize.** In his words, agents are like "apps with faces," and when agents talk to each other, apps that were never integrated by a business deal can become interoperable.
- **Consumer AI is not solved.** He said he has used Airbnb through consumer agents and that it "doesn't work very well" yet.

## What These Stories Have in Common

Taken together, the three stories show a market maturing from "look what an agent can do" to "how do we run, reach and connect agents safely." For a business, that suggests four practical priorities:

1. **Make your business agent-friendly.** Chesky said Airbnb wants to be more agent-friendly, and he sees agents as potential lead generators. Clean data, clear policies and well-documented services make it easier for agents to work with you.
2. **Meet customers where they already are.** A messaging channel can be a better first step than building a new app.
3. **Plan for testing and oversight from day one.** Agents that act need monitoring, limits, and a way for a person to step in.
4. **Prefer open connections over lock-in.** Standards such as MCP exist so agents can work across tools instead of inside one vendor's walls.

## Where to Start

You do not need an operating system for agents to begin. A sensible first project is a single, well-defined workflow, such as answering customer questions from your own documents or booking appointments, with a human review step and clear success measures. If you want background on agent-led commerce, see our guide to [what agentic commerce is](/blog/what-is-agentic-commerce/), and for the managed-service model, read about [agentic as a service](/blog/understanding-agentic-as-a-service-a-comprehensive-guide/). If you are comparing partners, our checklist for [choosing an AI development company](/blog/how-to-choose-ai-development-company/) covers what to ask.

## FAQ

**What is agentic AI?**
Agentic AI refers to software that does not just answer a question but takes actions toward a goal. It plans steps, uses tools, talks to other software, and works with limited supervision. A chatbot tells you how to book a flight; an agent books it.

**What is an agent swarm?**
In TechCrunch's report on the security startup Armadin, agent swarms are groups of AI agents that run continuously and chain vulnerabilities together to try to break into an enterprise. The idea is to find and fix weaknesses before attackers do, as an alternative to point-in-time penetration tests.

**What is MCP?**
MCP is a standard that lets AI agents connect with other software, as described in TechCrunch's interview with Airbnb CEO Brian Chesky. He expects agents connected through MCP to make apps interoperable that were never integrated through a business deal.

**How should a business start with AI agents?**
Start with one well-defined workflow, such as answering customer questions from your own documents or booking appointments, with a human review step and clear success measures. Expand only when the results justify it.

## Related Insights

- [AI Coding Agents: What Microsoft's Copilot Rethink Means](/blog/ai-native-software-and-coding-agents-what-microsofts-copilot-rethink-means/)
- [Gemini 4 Argon: How to Read a Frontier Model Release](/blog/gemini-4-argon-how-to-read-a-frontier-model-release/)

## Sources

- TechCrunch, [Kevin Mandia's new 'agent swarm' security startup Armadin raises $255.5M at $2.5B valuation](https://techcrunch.com/2026/10/01/kevin-mandias-new-agent-swarm-security-startup-armadin-raises-255-5m-at-2-5b-valuation/), October 1, 2026
- TechCrunch, [Photon held a funeral for mobile apps. Now it has $4.5M to help replace them with agents](https://techcrunch.com/2026/10/01/photon-held-a-funeral-for-mobile-apps-now-it-has-4-5m-to-help-replace-them-with-agents/), October 1, 2026
- TechCrunch, [Brian Chesky interview: AI agents need their own operating system](https://techcrunch.com/2026/10/01/brian-chesky-interview-ai-agents-need-their-own-operating-system/), October 1, 2026