OpenAI teases Dots: A new class of always-on agents
OpenAI just teased a new product called Dots, described as always-on agents. Details are limited, but the implications are massive.
Tools · Source: Hacker News
What happened
OpenAI just published a new page on their site introducing a product called Dots. The official URL confirms this is a first-party release directly from the company. The only text available right now describes them as always-on agents.
Hard details are completely unavailable at this moment. The company has not released technical specifications, context window limits, or pricing structures. We only have the product name and the core premise of persistent background execution.
The tech community spotted the URL on Hacker News. Speculation is running high, but OpenAI is keeping the actual documentation locked down. Builders are left waiting for the full announcement to understand the exact capabilities.
Key facts
- Dots — The official name of the new OpenAI product
- Always-on agents — The exact phrase used by OpenAI to describe the system
Why it matters
The phrase always-on agents represents a massive shift in how we build AI applications. Until now, models only acted when prompted by a user or a scheduled script. An always-on agent implies persistent state, continuous monitoring, and autonomous action without manual triggers. This means builders can stop writing complex scaffolding just to keep their agents alive.
This is a clear signal that OpenAI is moving further up the software stack. They are no longer content just providing the raw intelligence layer via an API. By hosting the agents themselves, OpenAI wants to own the runtime environment. This threatens startups building agent hosting platforms and orchestration tools.
For builders
Stop building custom agent hosting
If OpenAI provides native persistent execution, your custom polling loops become technical debt. Wait for the full release before investing more engineering hours into agent orchestration. You will save time but lose control over the infrastructure.
Prepare for new billing models
Always-on means continuous compute cycles. Expect a shift from simple pay-per-token API calls to pay-per-hour or subscription models for these agents. Founders need to start modeling how persistent compute will impact their unit economics.
Rethink your product moats
If anyone can deploy an always-on agent with a single API call, basic automation is no longer a competitive advantage. You need to focus on proprietary data and unique workflows. The infrastructure alone will not protect your margins.
My take
OpenAI wants to eat your entire backend. Dots proves they are tired of just being a dumb API and want to be the operating system for the next generation of software. I love the idea of deleting my orchestration code, but I am terrified of the ecosystem lock-in.
Original reporting: Hacker News. This is my rewrite and opinion.