September 30, 2026
On Tuesday, OpenAI didn't ship one product -- it shipped an ecosystem. Dots, a new agentic avatar meant to work independently of any app or device, launched alongside expanded ChatGPT plug-ins with sidebar homes, interactive panels, and automations, plus a cheaper new model, GPT-6.1 Sol, that OpenAI says nearly matches the flagship GPT-6 Astra on professional tasks like code review and multistep workflows. TechCrunch framed this correctly: it's a direct shot at the traditional app store model, where ChatGPT becomes the place software gets discovered and used, by humans and by other agents.
For a business buyer, the interesting part isn't Dots itself -- it's the plumbing underneath it. If ChatGPT becomes the default surface where employees find and run software, the vendors who win are the ones who build for that surface early, not the ones who wait for it to mature. That's the same bet we've made with our own AI App Builder: the interface war is being fought at the layer above individual apps, and tools that can plug into multiple agent ecosystems will outlast ones locked to a single app store's rules. It's also worth remembering OpenAI has been burned before by moving fast on trust and safety -- a pattern we've flagged in OpenAI's Apology Tour and the Price of Moving Fast. An agent that pursues goals continuously with minimal oversight is exactly the kind of feature that looks great in a keynote and shaky in an incident report six months later.
Buried under the product launches is a quieter fact: OpenAI is not a public supporter of Nvidia's Open Agent Safety Platform, an industry-wide effort aimed at reining in rogue AI agents. TechCrunch reports OpenAI is privately working with Nvidia anyway, which suggests this is less a philosophical objection than a branding decision -- OpenAI would rather set its own agent-safety terms than sign onto someone else's coalition, especially while it's simultaneously launching an agent, Dots, designed to run with minimal human oversight.
That's a defensible business move, but it's a strange one to make quietly on the same day you're launching a more autonomous agent. If you're a company evaluating agentic tools for internal workflows, the lesson isn't that OpenAI is being reckless -- it's that public commitments to safety standards and actual deployment practices can diverge, and you should ask vendors which one you're actually getting. We've written before about how quickly bolt-on agents can go sideways in production, most recently in SalesBleed: What a Salesforce Agentforce Bug Reveals About Bolt-On AI Agents. Autonomy sells well in a demo. It's a liability question in a security review.
TechCrunch also reports OpenAI is in talks to raise $30 billion at a $1.4 trillion valuation, reportedly its last round before a delayed 2027 IPO. Read alongside Tuesday's launches, the message is clear: Dots, the plug-in ecosystem, and GPT-6.1 Sol aren't side experiments, they're the product line OpenAI is betting its public debut on. A cheaper, nearly-as-capable model matters more than it sounds -- it's the piece that makes agentic features affordable enough for a mid-size business to actually run at scale, rather than just pilot. If you're comparing that against building or buying your own internal tooling, this is worth weighing alongside our own thinking in Buy vs. Build vs. ViibeStack: a cheaper frontier model lowers the cost of experimentation everywhere, not just inside ChatGPT.
None of this is settled yet. App-store-style ecosystems live or die on developer adoption, and OpenAI is asking developers to bet on a platform it substantially controls -- the same tension Apple and Google have never fully resolved with their own app stores. Whether OpenAI's version earns more trust than theirs is an open question, not a foregone conclusion.
If your team is already using ChatGPT day to day, are you seeing plug-ins and agents like Dots as genuinely useful new surfaces, or just more infrastructure to secure and manage? Tell us where you land.
Sources