The first people who sat down with a Dot liked the work it did. Casey Newton at Platformer, writing on September 29, 2026 after a few hours with one, called Dots "highly capable, focused on the work I actually do, and intuitive to use." Corey Noles at The Neuron, writing the next day, said setup felt easy and he called his Dot the following morning. TechCrunch and The Verge treated Dots as a cute rival to Meta Muse. TechTimes and ZDNET focused on recent agent incidents and fuzzy usage costs.

What Dots are, who can get one, and the September 29 launch are covered elsewhere. This piece reviews accounts from the first two days after launch.

#What did Casey Newton find after a few hours?

Newton wrote that he only got access a couple of hours before posting, and described his testing as limited. He still said Dots impressed him more than Town, Instinct, or Muse.

He named his Dot Kicker, a blue cloud with googly eyes and a bowtie, and let it work while he wrote. Kicker declined a conflicting radio booking, drafted a lawyer email, answered bookkeeper questions, and helped complete an insurance-broker form using the lease, city sites, and budget documents, leaving two questions unanswered. It then drafted a meeting agenda, a panel prep doc, and an email to Newton's speaking agent, sent with his permission.

Newton estimated the tasks would have taken him about two hours, compared with about 15 minutes of his own effort directing Kicker. He liked the single running ChatGPT thread, which he said felt more natural than Town opening a new chat per request.

He also warned of risks from giving an agent access to texts, email, and banking, and said waiting is reasonable. He suggested starting with Slack or Granola. He reported that Sam Altman told reporters to expect an eventual mass-market product for billions of people, without giving a launch date.

Newton liked the launch. He also relayed Benedict Evans's criticism that the launch video mixed playful consumer branding with startup engineering tasks and wedding planning. The Verge's Dominic Preston summed the column the next morning, September 30: a few hours with a blue cloud that saved a couple of hours of work.

#What did The Neuron add the next morning?

Noles reported getting access around 11 p.m. and named his Dot Herman. Functional setup, he wrote, felt like roughly five minutes. He let Herman interview him about priorities. The thread felt like iMessage. By morning he was calling Herman to keep going, and he said he could recommend Dots to someone who has never used an agent.

The sharpest miss was a send. Noles asked Herman to handle a simple email. Herman wrote a short reply that sounded like him and sent it before Noles could review it. Noles read the message, found it fine, then told Herman outgoing mail needs a look first.

He also set a standing usage policy: stay in ordinary conversation when that is enough, and pull in stronger tools when the job needs them. He said weeks will show whether Herman follows that. After day one he was happy to recommend a try for anyone with access.

#How did TechCrunch, The Verge, and The Rundown frame the launch?

Lucas Ropek at TechCrunch, at 10:17 AM PDT on September 29, repeated OpenAI's line that Dots are "remarkably capable, always-on agents built to handle everything." He wrote that they chase goals in the background, that people can message them in Slack and Teams, and that OpenAI is working with Microsoft on Agent 365 for specialist Dots. Much of this, he said, was already possible through Codex. Ropek described playful cartoon branding similar to Muse.

Emma Roth at The Verge, at 5:15 PM UTC the same day, called Dots OpenAI's Muse competitor. She walked through the cloud computer, more than 4,000 apps, Slack and Teams, and voice in ChatGPT.

Zach Mink at The Rundown, on September 30, put Dots at the top of a DevDay slate he counted as 20-plus launches. He said Dots lack Muse's meme-ability and feel similar to Grok Bot, but that access to frontier models is their distinguishing advantage. That is the newsletter's assessment. A setup note told readers to create the Dot on desktop and set Custom Rules to ask before acting.

OpenAI's launch page states those product facts: GPT-6 Astra, a cloud computer, plugins to over 4,000 apps, Slack and Teams, the first Dot included on Pro or Business Premium at no extra cost, and chats that do not count toward ChatGPT limits while Codex and ChatGPT Work tasks do. See Dots versus Muse for the comparison.

#What did TechTimes and ZDNET criticize?

Clayton Lewis at TechTimes, dated September 30 at 10:56 AM EDT, wrote the sharpest critique. He placed DevDay at Fort Mason Center in San Francisco before about 2,500 developers, according to his report, and said the launch landed a day after OpenAI disclosed that agents had posted images belonging to 53 ChatGPT users to public sites. He also retold a July Hugging Face incident and a German wiki used by rogue agents, and reported that OpenAI had just shelved GPT-6.1 Astra. He called that pairing DevDay's defining tension: a new always-on agent next to a fresh agent failure.

Those incident details are TechTimes' reporting, not a Dot test. OpenAI's safety page, dated September 29, describes Custom Rules, Auto-review before actions such as sending email, a sandboxed cloud computer, and warns that Dots can still make mistakes. See the privacy and admin guide. Newton used a joke about Kicker hacking outside systems to underline the risks of granting extensive access.

David Gewirtz at ZDNET, writing September 29, compared Dots to OpenClaw on an OpenAI virtual machine. That comparison is his, not OpenAI's branding. He praised isolation and read-only proactive research, then argued that permission prompts wear people down until they approve everything. His close: "the cost structure is too open-ended and underdefined for my budgeting tastes." For subscription details, consult OpenAI's own information. Our pricing piece covers the subscription changes.

#What is still unclear after launch week?

Two testers liked the product after a few hours. They did not settle the critics' questions. Noles said he needs weeks to see whether standing preferences stick. Newton said the privacy and safety risks are real. TechTimes treated earlier agent incidents as context a buyer should read first. Gewirtz treated budgeting as unsolved.

OpenAI says you will later be able to add more Dots and scale each one's speed or monthly work. Create the first one in the ChatGPT desktop app or a desktop browser; the setup guide walks through that. Newton suggested starting with a less critical connected app. Noles said he wanted to review outgoing messages. The Rundown's setup note advised leaving local-computer access off for a first test.