AI News Daily한국어RSS

Special · OpenAI DevDay 2026

Wednesday, September 30, 2026

7 stories worth reading, out of 40 posts on X and Threads in the past day (X 32 · Threads 8) · Every morning at 9 AM KST

* Summaries and publishing on this page are done by an automated agent. Check the facts against each story's original links.

OpenAI launches dots, always-on agents; the first dot is included with Pro and Business Premium

First seen on X 7 hours ago

At DevDay, OpenAI unveiled dots, always-on agents powered by GPT-6 Astra. Each dot has its own cloud computer and browser and keeps working after the conversation ends. It connects to more than 4,000 apps, and you can reach it in ChatGPT, Slack and Teams, or start a voice call. OpenAI's launch post drew 26,000 likes on X. Dots start rolling out today to Pro and Business Premium users in eligible markets, and your first dot is included in the plan at no extra cost. The first month after launch also comes with extended limits. Talking to your dot doesn't count toward usage limits; only tasks it hands off to Codex or ChatGPT Work do. Enterprise gets a beta that admins have to switch on. To start, create your first dot in the ChatGPT desktop app or a desktop browser, give it a name and connect your apps. Once setup is done, you can message it from the mobile app too. Inside OpenAI, a bug that shows up in Slack sends a dot straight into an investigation. One early tester's dot noticed a forgotten invoice, drafted it, and sent it after he approved. Sensitive tasks such as changing a password always stay with the user, and OpenAI warns that dots can still make mistakes, so review anything consequential.

Further reading

4 sources· @OpenAI, @OpenAI, @avocadochoe +1

  1. GPT-6.1 Sol arrives at $2 in and $10 out per million tokens, the same price as Sonnet 5.5

    First seen on X 6 hours ago

    OpenAI released GPT-6.1 Sol, an upgrade to GPT-6 Sol. OpenAI says it comes close to GPT-6 Astra on agentic coding, computer use and professional work at one-fifth of Astra's price. API pricing is $2 per million input tokens, $10 per million output tokens and $0.10 per million cached input tokens. The input and output prices match the Sonnet 5.5 that launched yesterday (Sonnet 5.5 charges $0.20 for cache reads). All figures are OpenAI's own evaluations. On DeepSWE v1.1, which tests software work in real codebases, it matched Astra at roughly a fifth of the cost. On the AutomationBench workflow test it scored 2.2 points above Opus 5.5 at medium effort for about a third of the cost. On Terminal-Bench Science 0.1 it averages $5.47 per task against $23.21 for Opus 5.5 and $23.80 for Astra, more than 75% cheaper. For the hardest scientific research, OpenAI says to keep using Astra, which tops that benchmark at 68.1%. Plus, Pro, Business, Enterprise and Edu users get it today in ChatGPT Work and Codex, but not yet in regular Chat. The API model name is gpt-6.1-sol, and a GPT-6.1 Sol Ultrafast tier, up to 8x faster, follows within days.

    Further reading

    2 sources· @OpenAI, @thsottiaux

  2. ChatGPT adds a $500 Pro plan; existing Pro 200 subscribers keep today's limits only through October 29

    First seen on X 9 hours ago

    ChatGPT now has a $500-a-month Pro 500 plan with 25 times the usage of Plus and access to Ultrafast. According to OpenAI's help page, Ultrafast is included only in Pro 500 among the three Pro plans. Pro 200 reopens to new subscribers the same day, but new sign-ups get a lower usage allowance than before. Codex's lead described the new ratios as Plus 1x, Pro 100 5x and Pro 200 10x. Existing Pro 200 members who were subscribed around the cutoff keep their previous allowance until October 29 and then move to the lower one; the price stays at $200. Existing subscribers also keep the 20x multiplier for a while and receive extra credits. There is a new way to use the subscription elsewhere. With Sign in with ChatGPT, Plus and Pro users can spend their plan allowance in 16 partner tools, including Devin, Notion, Vercel, T3 and OpenClaw, and set how much each one can use. Reactions were mixed. @bridgemindai, who bought the $500 plan himself, wrote that he feels cheated.

    Further reading

    4 sources· @thsottiaux, @testingcatalog, @bridgemindai +1

  3. Ultrafast mode is up to 8x faster in Codex, but only Pro 500 and Enterprise can use it for now

    First seen on X 6 hours ago

    Ultrafast is OpenAI's premium speed tier. Token generation is up to 8x faster in Codex (300 tokens per second) and up to 6x faster in the API. GPT-6 Astra Ultrafast opened today in the API, ChatGPT Work and Codex; on the product side it is limited to Pro 500 and Enterprise. GPT-6.1 Sol Ultrafast is coming soon. Sam Altman wrote that it is "so fast I do not ever want to go back," and the post passed 6,000 likes. You pick it in the model picker, and it draws on your plan's included usage first and credits after that. Buying credits on Pro 100 or Pro 200 does not unlock Ultrafast.

    Further reading

    2 sources· @OpenAI, @sama

  4. Codex moves to the cloud: run it on your computer, from your phone or in the cloud

    First seen on X 6 hours ago

    Codex can now run wherever you need it: on your computer, remotely from a phone, or in the cloud from any device. Reusable environments make tasks start quickly and give a team one approved setup. It is available on Plus, Pro, Business, Healthcare, Education and Enterprise. Codex's lead wrote that once you've configured a cloud environment, it's hard to go back to building on your laptop. The CLI was refreshed as well. You can start and steer tasks by voice, and a new /agents view lets you delegate work and track several tasks at once. The desktop app gets a code review screen with summaries and diffs where you can ask Codex about potential issues, and with automatic reviews Codex takes a first pass in the cloud while you're away. Codex Security Cloud scans whole GitHub repositories on demand or on a schedule, removes duplicate findings and prepares fixes. For developers, the Agents API is in preview. It is the same technology behind dots and OpenAI's cloud agents, and it now supports computer use, with OpenAI running the infrastructure. Bedrock Managed Agents, built with Amazon, runs OpenAI agents entirely inside AWS. The Decisions API, which points Luna at a fixed set of questions with predefined answers, is in limited preview with a broad release planned within days.

    Further reading

    2 sources· @thsottiaux, @DanDr1s

  5. ChatGPT gets Space and Pages for teams, and you can call @ChatGPT in Slack and Teams

    First seen on X 6 hours ago

    ChatGPT Space is a shared home for teams. Teammates, ChatGPT and your dot build on the same knowledge, and ChatGPT keeps the space organized based on your instructions. Documents in a Space are Pages, a new kind of interactive document with charts, images, checklists and dashboards, where you can tag ChatGPT, Codex or your dot to take on action items. It is available on Pro, Business and Enterprise in the desktop app and on the web. Mention @ChatGPT in Slack or Teams and it helps with team work without each person needing a ChatGPT license (Business and Enterprise). The Meetings plugin saves notes and action items in your Space and deletes the audio once notes are ready (macOS desktop, beta for Pro and Business). Collaborative slides arrive in the coming weeks, with export to PowerPoint or Google Slides. On the plugin side, plugin extensions let developers give a plugin its own sidebar panel, and support for the proposed MCP Events spec lets automations start when something happens in a connected app. Enterprise customers can apply part of their existing OpenAI commitment to partner software through OpenAI Marketplace, whose first 32 partners include Figma, Adobe, Salesforce, ServiceNow and CrowdStrike.

    Further reading

    2 sources· @ChatGPT, @prompt_what

  6. DevDay reactions split between jabs at look-alike features and complaints about the $500 plan

    First seen on X 11 hours ago

    Reactions split right after the keynote. One developer summed up the announcements as "a clone of GrokBot, a clone of Notion and Claude Tag, a clone of Opus 5.5, a clone of Claude Cloud Environment, a $500 plan, a super fast mode, a marketplace and a bunch of plugins," and the post passed 400 likes. Someone who bought the $500 plan said he feels cheated. Posts welcoming Ultrafast and Codex in the cloud kept coming as well. The rumors that ran ahead of the event did not show up on the list. There had been talk of a surprise GPT-6o reveal and of a GPT-6.1 Astra release, but OpenAI's DevDay recap mentions neither, and GPT-6.1 Sol is the only new model. A post asking whether a five-hour limit is being added to the $100 and $200 plans (746 likes) has no official answer that we could confirm.

    Further reading

    4 sources· @goon_nguyen, @CodexResets1, @bridgemindai +1
View all 40 collected posts