OpenAI’s Alexander Embiricos is coming to TechCrunch Disrupt 2026 — days after the launch of Dots

We have spent the last few years watching AI companies promise us the world while delivering what amounts to very clever autocomplete. But the conversation is shifting. The focus is moving away from just talking to a box and toward software that actually does work. The recent news that Alexander Embiricos, a lead at OpenAI, will be speaking at TechCrunch Disrupt 2026 immediately following the release of Dots is a signal builders should pay attention to.

For those who have been following the infrastructure play, Embiricos isn't just another executive. His background with Remotion and his transition into the OpenAI fold suggests a specific direction for the company: the intersection of collaboration and execution. The launch of Dots represents a pivot from the LLM as a chat interface to the LLM as a coordinator. If you are building in the crypto or AI space right now, you need to understand that the "chat era" is effectively over. We are entering the agentic era, and it is going to be messy.

The End of the Chat Interface

For a long time, the industry was obsessed with the Turing test—could a machine trick a human? As founders, we don't care about tricks. We care about utility. The launch of Dots suggests that OpenAI is finally leaning into the idea that an AI shouldn't just tell you how to write code or manage a project; it should be the connective tissue between the tools you already use.

This is where the skepticism comes in. Every time a major player like OpenAI launches a new coordination layer, they effectively kill a thousand startups. If your entire business model was being a "wrapper" that helps teams collaborate using GPT-4, you just got a very clear warning shot. The platform is absorbing the features. For builders, this means you have to move deeper into the stack or solve problems that are too niche or too high-stakes for a generalist agent to handle.

What Dots Tells Us About OpenAI's Roadmap

We have to look at the timing. Bringing Embiricos onto the stage days after a major launch isn't a coincidence. It is a marketing victory lap, but for founders, it is a chance to look at the plumbing. Dots isn't just a new UI. It represents a shift in how OpenAI treats memory and context. The biggest hurdle for AI agents has always been the "goldfish problem"—they forget what they were doing three steps ago.

If Dots solves for persistent context across team environments, the barrier to entry for complex automation drops. But it also raises the stakes for data privacy and sovereignty. This is where the crypto founders usually perk up. If OpenAI is building a centralized hub for all team interactions and project logic, the demand for decentralized, private alternatives will skyrocket. You cannot have a global work coordinator that is a black box owned by a single entity without eventually facing a massive backlash regarding data ownership.

Infrastructure vs. Application

I talk to a lot of founders who are still trying to build the "next big AI app." My advice is usually to stop. The application layer is being eaten by the incumbents. Google, Microsoft, and OpenAI are building the applications. Where the real opportunity lies is in the infrastructure that makes these agents reliable.

If you listen to the way Embiricos talks about product, it is always about friction. The friction in current AI models is high. You have to prompt them, check them, correct them, and then copy-paste their output. Dots is an attempt to remove the copy-paste. It is an attempt to make the AI a participant rather than a consultant. As a builder, your goal should be to find the places where this participation breaks down. Where does an agent fail? It fails at high-trust financial transactions. It fails at hardware integration. It fails at cross-chain verification in the Web3 world.

The Reality of Agentic Workflows

Let's be honest: most AI agents today are still pretty bad at multi-step reasoning. They get distracted. They hallucinate mid-process. By the time Disrupt 2026 rolls around, we will have seen the first wave of real-world failures from this new batch of tools. We will see where the guardrails were too thin.

The founder's perspective here should be one of cautious adoption. Use these tools to lean out your operations, but do not build your core value proposition on top of a feature that OpenAI can replicate in a weekend. Dots is a feature, not a company. However, the technology behind it—the ability to maintain state across complex workflows—is the new baseline. If your software feels more static than an OpenAI agent, you are already legacy code.

Takeaways for the Builder Community

• Verticalization is the only shield: If you build a horizontal tool, you are competing with Dots. If you build a tool specifically for high-frequency trading or legal compliance in specific jurisdictions, you have a moat.
• Focus on State, not just Chat: The future is about maintaining context over long periods. If your tool doesn't remember what the user did last week, it's useless.
• Watch the API Pricing: As these agents become more active (doing more tasks automatically), the cost of compute will either become a bottleneck or a competitive advantage.

The appearance of Alexander Embiricos at Disrupt is going to be a litmus test. We will see if OpenAI is ready to be an enterprise OS or if they are still just a research lab with a very popular demo. For those of us on the ground, the mission remains the same: ignore the hype, watch the git commits, and build things that actually work when the API goes down.

Read the original at TechCrunch Startups →