OpenAI Introduces Dots, AI Agents That Keep Working in the Background

Morning, {{first name | folks}}! AI had a busy stretch. OpenAI launched background AI agents, Perplexity turned AI tasks into automations, Google signed a new AI safety accord, MongoDB introduced infrastructure for production agents, and OpenAI released a cheaper model for coding and computer use.

OpenAI has introduced Dots, always-on AI agents that can work on tasks without needing a new prompt every time. Each Dot gets its own cloud computer and can connect to more than 4,000 apps, with access through ChatGPT, Slack, and Teams.

Dots can keep context across tasks, work in the background, and ask for approval before taking actions that affect accounts or share information. OpenAI is rolling them out to Pro and Business Premium users, with specialist Dots for specific company roles also being tested. The shift is from AI that answers when you ask to AI that keeps working after you leave.

Perplexity has added automations to computers, letting AI agents run tasks on a schedule or when something happens in apps like Slack, Gmail, Outlook, Linear, and GitHub. Each automation can use connected files, tools, and projects while remembering what happened in previous runs.

Automations can handle things like weekly competitor reports, customer reply drafts, and project updates, and can flag tasks when something expected hasn’t happened. Users can also set which actions the agent can take on its own and which need human approval.

Google CEO Sundar Pichai says Google has signed the White House Accord on Superintelligence with the US administration and tech leaders. The agreement covers AI testing, evaluations, red-teaming, and safeguards against misuse and misalignment.

Pichai said Google has invested hundreds of billions of dollars across its AI stack over the past two years and will continue investing. The companies involved also committed to working on shared safety practices as frontier AI systems become more capable.

MongoDB has launched Atlas Agent Engine, a set of tools for building and running AI agents in production. It handles things like memory, retrieval, tool use and governance, while supporting different models and agent frameworks.

The goal is to let teams build agents on top of infrastructure they already use instead of adding another stack. MongoDB says the tools are designed for agents that need to work with real application data and take actions, not just answer questions.

OpenAI has launched GPT-6.1 Sol, a new model for coding, computer use and professional work. OpenAI says it gets close to GPT-6 Astra on several evaluations while costing around one-fifth as much for standard input and output tokens.

GPT-6.1 Sol matches Astra on DeepSWE coding tests at about one-fifth the cost and comes within 2.1 percentage points on OSWorld computer-use tasks at roughly one-seventh the cost. It is available through the API as gpt-6.1-sol for $2 per million input tokens and $10 per million output tokens. The release puts more focus on how much performance you get for each dollar, not just which model scores highest.