OpenAI's dots watch the work and take the next step

On September 29, OpenAI introduced dots, agents designed to keep working over time. Instead of stopping after an answer, a dot can remember requests, spot pending work, and move it forward. It has its own cloud computer and can be reached through ChatGPT or tools such as Slack.

Notice work that was missed

In an early-user example shared by OpenAI, a dot spotted a forgotten invoice and prepared it. The user approved the invoice before it was sent. The idea is to identify the next task, not just wait for another prompt.

Dots can connect to external apps, but users choose their access and scope. Availability also depends on region and plan; this is not something everyone can use immediately.

From answering questions to continuing the work

AI use could shift from opening a chat each time to having an agent remember the flow of work and speak up when something needs attention. Meta also announced a small-business version of Muse on the same day, making this longer-running form of assistance more concrete.

People remain responsible for important actions, while agents can research, organize, and prepare drafts. That could help small daily tasks move forward without repeatedly starting over.

openai.com ↗ ↗

Tools shaped around the work itself

What stands out about today's dots and Muse announcements is less the list of tasks they can take on than the possibility of work continuing between conversations.

Spot a missing invoice. Prepare the next advertising draft.

Instead of explaining the situation from the beginning every time, a person could have an agent carry the context forward and prepare what comes next.

People still decide when to send, publish, or spend. If an agent handles the research and preparation beforehand, more time is available for those decisions.

A conversation with AI begins to become an ongoing working relationship.

Claude Code / Anthropic

No new article was verified in this category today.

OpenAI / ChatGPT

Medium impactMedium signal
GPT-6.1 Sol aims to make complex work less costly
OpenAI introduced GPT-6.1 Sol on September 29 for coding and multistep work. The company says it approaches GPT-6 Astra's capability at one fifth of Astra's standard token price. That comparison is OpenAI's own claim.
openai.com ↗ ↗

AI Agents / Coding

Medium impactMedium signal
GitHub Copilot starts offering GPT-6.1 Sol
GitHub said on September 29 that GPT-6.1 Sol is rolling out in Copilot. Eligible users can select it in VS Code, Copilot CLI, the coding agent, and other surfaces. Plan eligibility and rollout timing vary.
github.blog ↗ ↗

AI INSIGHTS / KOHEI TORIGOE

Kohei Torigoe

AI Product Architect

Over 25 years in UI/UX design, product development and business development. Founded and led a company with a team of around 100 people.

Currently planning, designing and developing new products using AI.

Through AI Signal, I share discoveries from developing new products and the possibilities AI opens up for the future, while trying the latest AI technologies firsthand.

Profile, work and contact ↗

Video Generation AI

No new article was verified in this category today.

AI Music Generation

No new article was verified in this category today.

Foundation Models / Research

Medium impactMedium signal
Google Research tests steadier control over AI-generated images
Google Research introduced Diffusion Controller on September 29. The research aims to follow details in an image prompt while preserving the rest of the picture—for example, adding sunglasses without distorting a face. It is a research result, not a broadly released product.
research.google ↗ ↗

AI Automation / Business Use

Medium impactMedium signal
Meta connects Muse to small-business workflows
Meta introduced Muse for Small Business on September 29. It can connect to business tools for tasks such as sales analysis and advertising drafts. The company says publishing, sending, and spending require the person's approval.
about.fb.com ↗ ↗

AI Insights