Picture yourself in a folding chair somewhere in the middle of a packed conference hall. The lights dim. Someone from Anthropic is on stage, and later, someone from OpenAI. These are the two labs whose products you have probably argued about with a coworker. You have a lukewarm coffee in one hand and your phone in the other, wondering whether you should be taking notes or just listening.
That scene is scheduled to happen at TechCrunch Disrupt 2026. Anthropic and OpenAI are both joining the AI stage. And if you are someone who uses AI tools without building them, this is a bigger deal than it might first appear.
Why both labs showing up matters
Anthropic and OpenAI are competitors. They build models that compete for the same customers, the same developers, and the same headlines. When two competing labs appear at the same event, they are not just presenting products. They are staking out positions in front of an audience that includes press, investors, and the people who decide what software their companies buy next year.
For non-technical folks, conference talks like this are one of the few moments where you get to hear how these companies describe themselves in plain language. Product pages are marketing. Research papers are dense. But a stage conversation, with a host asking follow-up questions, tends to produce clearer answers about what a tool actually does and where it falls short.
My honest advice if you follow the coverage: pay less attention to the demos and more attention to how each company talks about limits. Anyone can make a model look impressive in a rehearsed five-minute segment. What tells you more is how a speaker responds when asked what their agents cannot handle yet.
The Replit signal
Amjad Masad, CEO and co-founder of Replit, is also joining the Disrupt Stage. Replit sits in a different spot than the big labs. It is a place where people build software, increasingly with AI doing a large share of the typing.
That combination on one program is worth noticing. You have the companies making the underlying models, and you have a company making the environment where those models get put to work. Those are two halves of the same story. The labs build the engine. Companies like Replit build the car, the dashboard, and the seatbelt.
For anyone trying to understand AI agents without a computer science degree, that split is one of the most useful mental models you can carry. When someone says “an AI agent built this app,” there are usually at least two companies involved, and their incentives are not identical.
Agents that run the business, not just start it
One other item from the same news cycle stuck with me. Runable raised $21 million on the bet that AI agents can go from building businesses to growing them.
Read that framing carefully, because it describes a real shift in ambition. Building something is a bounded task with a clear finish line. Growing something is open-ended. It means making judgment calls repeatedly, over long stretches of time, without a human checking each step.
Those are very different problems. Most AI agent products you can use today are good at the first kind of work and shaky at the second. So when you see funding aimed squarely at the second kind, that is a signal about where the industry thinks it is heading, not a description of what works right now. Keeping those two things separate in your head will save you a lot of disappointment.
You get a vote
Two smaller notes from the Disrupt announcements are easy to skip, but they change who gets to participate.
- Applications are open to host a Side Event at TechCrunch Disrupt 2026, which means smaller organizers can put together their own sessions around the main program.
- Audience Choice voting is open to decide which sessions make it to the TechCrunch Founder Summit.
That second one is a small piece of real influence. Conference agendas usually get set by whoever has the biggest budget or the loudest launch. A public vote nudges that slightly toward what people actually want to hear about. If you have ever sat through a session that felt like an ad, voting is a low-effort way to push for something better.
How I would approach it
If you plan to follow Disrupt 2026 coverage, go in with questions rather than expectations. What tasks are these agents genuinely reliable for? Who is accountable when one makes an expensive mistake? What does it cost at scale, not in a demo?
The people on stage are smart, well-prepared, and there to make their companies look good. That is not a criticism, it is just the format. Your job as an audience member, in the room or reading about it later, is to listen for the parts that are not in the script.
🕒 Published: