AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: The OpenAI Dev Day 2026 Rundown: What Was Announced on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on office and shipping supplies

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

At DevDay on 29 September 2026, OpenAI announced more than 20 products and features positioning ChatGPT as an agent platform, including always-on ‘Dots’ agents powered by GPT-6 Astra, expanded plugin tooling, a Decisions API in limited preview, and the cheaper GPT-6.1 Sol model.

OpenAI used its DevDay on 29 September 2026 to announce more than 20 products and features that reposition ChatGPT from a chatbot into an agent platform, including always-on agents called Dots, a major expansion of its plugin system, a new Decisions API in limited preview, and the lower-cost GPT-6.1 Sol model. The company pitched ChatGPT as a shared surface for humans and agents, saying it now reaches 1.2 billion weekly users. For anyone who builds or sells software, the announcements change how customers may find and hire services — through agents selecting plugins by outcome rather than by brand.

Dots are OpenAI’s biggest consumer bet: persistent agents powered by GPT-6 Astra, each running on its own cloud computer and browser, able to connect to over 4,000 apps through plugins. A dot works toward a user’s goals around the clock, learns preferences over time and can manage several projects at once. Access is available now for Pro and Business Premium subscribers in eligible markets; Enterprise, Edu and Healthcare customers can use a beta once an administrator enables it, and it ships off by default. The first dot is included in the plan, and chatting with it does not count toward usage limits, though tasks it starts in Codex or ChatGPT Work do. Users can reach dots through ChatGPT on desktop, web and mobile, plus Slack and Teams — voice calls work today, with texting to come.

OpenAI framed control as the design constraint. Each dot runs on a separate cloud computer; dots can use saved passwords without exposing them to the model; and background ‘proactive research’ is limited to read-only tools. Custom Rules let users allow, require approval for or block specific actions, an Activity View shows background work, and monitoring tools can pause or stop a dot. OpenAI nonetheless warns that dots can make mistakes. An enterprise variant, Specialist dots, gives agents their own identity, credentials and IT-provisioned hardware, with pilots starting in procurement, invoice processing, email marketing, customer support and commercial contracting, and a planned Microsoft Agent 365 governance integration.

The plugin system is now the gateway into ChatGPT, Codex and dots. New additions include plugin extensions that live in the ChatGPT sidebar — Figma and Adobe showed integrations — a Plugin Creator with a new submission flow, improved directory discovery, Sites that let teams host plugins with their own connected data, and support for the proposed MCP Events specification, which lets a plugin trigger an automation when something changes in a connected app. An OpenAI Marketplace lets eligible enterprise customers spend part of their OpenAI commitment with 32 launch partners including Figma, Adobe, Salesforce, ServiceNow, Harvey, CrowdStrike and Baseten. The developer docs also include a guide for submitting a Claude Code plugin, which suggests OpenAI intends to accept skills built for other agent ecosystems.

On the model side, the Agents API now supports computer use, letting agents operate software through its interface, with OpenAI running the Codex harness on the developer’s behalf. Sign in with ChatGPT lets users carry their existing plan to 16 launch partners. The Decisions API applies GPT-6 Luna to narrow, repetitive choices — developers define questions with a finite list of allowed answers and receive a selection usable for classifying content, routing requests or choosing an agent’s next step. It is in limited preview. GPT-6.1 Sol offers near-Astra intelligence at one-fifth of Astra’s token prices, according to the roundup of the event.

At a glance
recapWhen: announced 29 September 2026
The developmentOpenAI used its DevDay event on 29 September 2026 to announce its shift from chatbot to agent platform, with more than 20 products and features.

OpenAI DevDay 2026 at a glance

29 September 2026: more than 20 announcements across ChatGPT, Codex, the API and enterprise.

The five launches that matter, plus Sol

Six of the 20-plus announcements decide what builders do next.
DotsAlways-on agents on GPT-6 Astra with their own cloud computerrolling out
PluginsExtensions, Plugin Creator, discovery, Sites and MCP eventslive now
Decisions APILuna answers from a finite list you definelimited preview
Agents API + computer useManaged Codex harness that can drive a browserpublic beta
Sign in with ChatGPTUsers bring their plan to 16 partner appslive now
GPT-6.1 SolNear-Astra intelligence at one-fifth of the priceWork, Codex, API

The rest of the slate

Green means live, amber means preview or beta, blue means coming.

Collaboration

ChatGPT SpacelivePagesliveCollaborative slidesnext few weeksTeams and Team Taskslive@ChatGPT in Slack and TeamsliveMeetings pluginbeta, macOSShareable profileslive

Codex

Codex CloudliveCLI refreshvoice, /agentsCode ReviewliveCodex Security Cloudlive

Plans, enterprise, speed

Pro 500$500, 25× PlusUltrafastAstra now, Sol soonPrivate IntelligencePrivate Inference this fallBedrock Managed Agentsruns in AWSOpenAI Marketplace32 partners, apply
From “OpenAI DevDay 2026: Everything Announced and What It Means” on thorstenmeyerai.com. Sources: OpenAI DevDay recap and product pages, OpenAI developer docs, The New Stack, The Next Web, Axios, Artificial Analysis. Checked 30 September 2026. OpenAI figures are OpenAI’s own claims.

What the Agent Shift Means for Builders

The announcements matter because they change how software and services may be discovered and paid for. As the event commentary put it, an agent picks a business by outcome, not by brand: the customer asks for a result, ChatGPT understands the job, and a plugin is selected to do the work. That means a business can effectively be ‘hired before the customer knows your name,’ making plugin selection, activation and completion metrics the new competitive battleground.

The pricing model is also consequential. With the first dot included in existing plans and Sign in with ChatGPT carrying a user’s plan into partner apps, OpenAI is enabling a free core app with paid upsells — a distribution model that lowers the barrier for plugin-based products but increases dependence on OpenAI’s ranking and recommendation systems. GPT-6.1 Sol at one-fifth of Astra’s token price, if it holds up in real workloads, could shift which tasks developers route to which model.

From Chat to Agents: OpenAI’s Platform Path

DevDay has served as OpenAI’s stage for developer-facing launches, and the 2026 edition continued a multi-year move from conversational products toward agentic infrastructure. The plugin announcements build on prior integrations, while the MCP Events support follows the broader industry adoption of the Model Context Protocol for connecting AI systems to external tools. The introduction of a Decisions API marks a further specialization: instead of general-purpose chat, it targets high-volume, narrow classification and routing decisions inside applications.

The event also positioned ChatGPT’s scale — 1.2 billion weekly users, per OpenAI — as the substrate for this agent economy, with dots, plugins and Sign in with ChatGPT all funneling activity through the same interface.

“An agent picks a business by outcome, not by brand — you can be hired before the customer knows your name.”

— Event commentary summarized by ThorstenMeyerAI.com

What DevDay Left Unanswered

Several open questions remain. The Decisions API is in limited preview, with a broad release promised but no confirmed date. OpenAI’s examples of what dots do in practice — such as an early tester’s dot preparing and sending an invoice after approval — come from the company itself or small tester groups, and independent performance data is not yet available. Pricing beyond the first dot, including the cost of added dots, extra speed and higher monthly output, has been outlined but not fully detailed.

It is also unclear how plugin ranking and recommendations will actually behave at scale, how the Microsoft Agent 365 governance integration will work for Specialist dots, and whether accepting Claude Code plugins signals a broader interoperability strategy or a one-way acquisition of skills from rival ecosystems. Real-world reliability of computer use in the Agents API remains to be tested outside demos.

Rollout Timeline and Developer Moves

Dots are available now for Pro and Business Premium subscribers in eligible markets, with Enterprise, Edu and Healthcare betas requiring administrator opt-in. Texting with dots is listed as coming, alongside future options to add dots and buy more speed and output per dot. Developers can expect the broad release of the Decisions API, wider rollout of the OpenAI Marketplace beyond eligible enterprises, and the first Specialist dot pilots in procurement, invoice processing, email marketing, customer support and contracting. Builders evaluating the platform will likely focus next on publishing plugins, integrating Sign in with ChatGPT, and measuring how agent-driven selection affects their conversion and revenue.

Key Questions

What are OpenAI’s new Dots?

Dots are always-on personal agents powered by GPT-6 Astra. Each has its own cloud computer and browser, connects to over 4,000 apps through plugins, and works toward user goals continuously. The first dot is included in Pro and Business Premium plans in eligible markets.

Can dots access my computer and passwords?

Per OpenAI, dots run on their own cloud computers, separate from a user’s machine, and can only reach a laptop if the user grants permission. Dots can use saved passwords without exposing them to the model, background research is read-only, and Custom Rules let users block or require approval for specific actions.

What is the Decisions API and who can use it?

It applies GPT-6 Luna to narrow, repetitive choices: developers define questions with a finite list of allowed answers and receive a selection for tasks like content classification or routing. It is currently in limited preview, with a broad release promised but not yet dated.

What is GPT-6.1 Sol?

GPT-6.1 Sol is a model OpenAI says offers near-Astra intelligence at one-fifth of Astra’s token prices, positioning it as a cheaper option for high-volume workloads. Independent benchmark data has not yet been published.

How do the plugin changes affect businesses?

Plugins are now the distribution channel into ChatGPT, Codex and dots. New extensions, better discovery and the OpenAI Marketplace mean businesses that publish plugins can be selected by agents based on outcomes — before a customer knows their name — making plugin visibility and completion rates commercially important.

Source: ThorstenMeyerAI.com

This content is for general information only and is not financial, tax or legal advice. Consult a qualified professional for decisions about your money.
FALL

Fall Picks

As an affiliate, we earn on qualifying purchases.

You May Also Like

10 Best Content Creator Laptops for Video, Photo, and Design Work in 2026

Discover the 10 best laptops for content creators in 2026, optimized for video editing, photography, and design work, based on processor, display, and performance.

2026 AI Trends: 9 Breakthroughs In Artificial Intelligence

An overview of the nine key AI breakthroughs expected in 2026, highlighting their impact on technology, industry, and society.

The Acoustic Treatment Moves That Improve Office Recording Spaces

Find out how strategic acoustic treatments can transform your office recording space and unlock professional-quality sound results.

Kill-Switch-Proof: How to Build So Washington Can’t Take Your AI Stack Down

Exploring strategies to make AI infrastructure resilient against government shutdowns, emphasizing dependency mapping, gateways, fallback tiers, and open-weight models.