OpenAI Dots launched as always-on agents. The access list does not match.

OpenAI introduced OpenAI Dots at DevDay on September 29, 2026, as always-on agents powered by GPT-6 Astra. TechCrunch's piece is timestamped 10:17 AM PDT. The agents are described as running in the background with minimal oversight, which is a very different product shape from a chat box you poke and then close. Two write-ups cover the launch, and they do not agree on Enterprise access.

TL;DR - On OpenAI Dots, TechCrunch says Dots opened Tuesday for Pro and Business Premium in eligible markets, launchable from Codex or ChatGPT, with Slack and Teams messaging and text marked coming soon. - Business Standard says each Dot gets its own cloud computer and browser, can do recurring work, can hand work to sub-agents, and spans ChatGPT, text, email, and Slack. - The two write-ups disagree on Enterprise. Business Standard lists it in initial access. TechCrunch's availability line does not.

What actually happened

DevDay is the same Tuesday stage as the rest of OpenAI's September 29, 2026 slate. Dots are the agent product in that slate, and the model named underneath them is GPT-6 Astra, not the cheaper Sol model from the neighboring announcement. If you blur those posts together, you will brief the wrong stack.

On OpenAI Dots, TechCrunch carries OpenAI's line in this form. File it as a slogan, not as a scope document.

remarkably capable, always-on agents built to handle everything.

"Everything" is not a scope document. The operational description next to it is more useful: they run in the background, and the oversight is supposed to be minimal. Minimal oversight is the part worth arguing about inside a company, because background work is exactly where a bad instruction gets time to compound.

The announcement post, as quoted for this piece, tells a single user where to start. Naming a dot is a product gesture, not a permissions model.

Today, you can start with your primary dot, give it a name, and make it your own.

It is not a description of permissions, logs, or what happens when that dot has tools. Treat the sentence as onboarding copy, and stop there.

TechCrunch's availability line is Pro and Business Premium, in eligible markets, from Tuesday, and you launch one from Codex or ChatGPT. Messaging runs through Slack and Teams. Text is placed at "coming soon," which means the write-up is telling you not to expect it on day one.

On OpenAI Dots, Specialist Dots, in that same account, can get identities, credentials, and tools. OpenAI is also working with Microsoft on Agent 365. This piece does not have a definition of Agent 365 beyond that collaboration line, so do not invent one in a meeting.

Business Standard describes the machinery differently, and more concretely. Each Dot has its own cloud computer and browser. It can do recurring work and hand work to sub-agents.

The surfaces it lists are ChatGPT, text, email, and Slack. Initial access in that write-up is Pro, Business Premium, and Enterprise.

What to do as a reader

Do not merge the access lists. TechCrunch names Pro and Business Premium in eligible markets. Business Standard names Pro, Business Premium, and Enterprise.

Enterprise is the disagreement, and it is a real one, not a wording quibble. If your rollout plan says "Enterprise is in," you are following Business Standard, not TechCrunch. If it says Enterprise was not in the first availability line, you are following TechCrunch. Pick up both pages and look at the plan names before anyone writes a policy.

On OpenAI Dots, there is a second tension, smaller but sloppy if you ignore it. TechCrunch says text is coming soon. Business Standard lists text among the places a Dot spans, alongside ChatGPT, email, and Slack.

Those sentences cannot both be a clean "text works today" without somebody being wrong or using "text" for two different things. The pack does not resolve it. Say they disagree, and wait for the product UI.

Slack shows up in both. Teams shows up in TechCrunch and not in the Business Standard channel list used here. Email shows up in Business Standard and not in TechCrunch's messaging line.

A specialist Dot getting identities, credentials, and tools is TechCrunch's detail. A private cloud computer, a browser, recurring work, and sub-agents are Business Standard's details. None of that is a reason to average the stories into one feature grid.

If you are evaluating this for work, start with the boring questions the pages do answer. Which plan names are you actually on. Which app you can launch from, Codex or ChatGPT, matches TechCrunch.

Whether a background agent with credentials is acceptable is a governance question these articles do not settle for you. "Minimal oversight" plus "credentials and tools" is a combination that deserves a human policy, not a vibe check.

A practical test, if your plan is on a list you trust, is narrow. Name a primary dot, give it one recurring task you can verify, and watch where the messages land. If text is absent, that fits TechCrunch's "coming soon." If email is absent, that fits the fact that email is a Business Standard claim, not a TechCrunch one.

On OpenAI Dots, write down what you saw. A launch-day UI beats a stitched recap.

What this piece will not pretend

There is no price in this account. A monthly Pro figure that has floated around other coverage was not in the TechCrunch Dots article, so it is not here either. Do not backfill it from memory. If a budget conversation needs a number, get it from a page you opened, not from this recap.

Agent 365 is a name attached to a Microsoft collaboration, nothing more, in the material used here. Sub-agents, a cloud computer, and a browser are Business Standard's description of how a Dot is built. They are not a security review. Identities, credentials, and tools are a TechCrunch line about specialist Dots, and that line does not say which credentials or which tools.

Eligible markets are mentioned and not listed. "From Tuesday" means the DevDay date in TechCrunch's framing, not a promise about every region. Background execution with minimal oversight is the product idea.

On OpenAI Dots, it is also the risk sentence. If your bar for an agent is "a person clicks Run," this is a different product, and the slogan about handling everything does not lower that bar.

Read TechCrunch's Dots story next to Business Standard's DevDay recap. Where they disagree, keep both sentences. Where only one of them mentions a channel or a plan, label it as one outlet's line.

Readers who want sourced recaps that are already on the site can read DogeOS public testnet and PUMP token buyback burn as separate live posts.