Today’s roundup focuses on keeping useful AI workflows working: preparing custom GPT replacements, understanding Copilot Notebook limits, checking a faster Perplexity option, and reviewing saved model settings after a retirement deadline. Here are four practical updates and reminders for office teams.
1. Custom GPT migration: check your timeline and integrations
Platform: OpenAI / ChatGPT
OpenAI schedules custom GPT retirement for December 11, 2026, with February 11, 2027, applying only to affected Enterprise workspaces with an approved deferral. Follow your account or workspace notice.
The October 26 planned cutoff for creating new custom GPTs is listed under affected Enterprise workspace guidance—not as a universal deadline for every account.
Migration turns instructions into a skill and copies knowledge files into plugin references. Custom actions do not transfer automatically. Existing GPT conversations remain accessible after retirement; the notice describes GPTs and their pages becoming inaccessible, not deletion of every conversation.
Arctic Blossom Takeaway: Inventory the GPTs your team uses, identify their owners, and flag external connections. Finish essential edits before migration, then test the replacement. The original becomes read-only, and sharing access does not automatically carry over. Confirm coworkers can use the replacement before switching.
2. Copilot Notebook limits depend on your access
Platform: Microsoft Copilot Notebooks
Microsoft’s current support documentation distinguishes Copilot Chat from paid Copilot Notebook capabilities. Copilot Chat users can add up to 50 references, all used for grounding—the source material behind an answer. Paid Copilot users can add more than 300 references, but only the first 300 are used for grounding.
The documented entity options also differ: Copilot Chat supports web URLs, while the paid column includes Outlook emails, Teams meetings, SharePoint sites and pages, and folders. Creating a notebook requires the applicable SharePoint or OneDrive service plan.
Why it matters: A notebook being available does not mean every source type or capability is included in your account.
Arctic Blossom Takeaway: Before buying seats, test the source types your workflow needs and confirm licensing with your administrator. This is a current-documentation check, not evidence of a new limit imposed today. Access to a feature is not approval to upload patient or confidential client information.
3. Perplexity announces Ultrafast for Max subscribers
Platform: Perplexity Computer
In its October 14 announcement, Perplexity says Max subscribers can use an Ultrafast speed setting for GPT-6 Astra and GPT-6.1 Sol in Computer. Its instructions are to choose a supported model under Custom, then select Ultrafast.
The announcement describes faster token generation to help finish long Computer tasks sooner. It does not establish availability for every Pro or Enterprise account, instant reasoning, or a measured speed improvement for every task.
Arctic Blossom Takeaway: If the option appears in your Max account, compare one small, non-sensitive task with your normal setting. Check output quality, completion time, and credit usage before committing a long workflow. Faster output still needs source verification.
4. After the GPT-5.5 deadline, check saved settings
Platform: OpenAI / ChatGPT, ChatGPT Work, and Codex
As a follow-up to yesterday’s notice, OpenAI’s published GPT-5.5 retirement date was October 14 for ChatGPT, ChatGPT Work, and Codex across all plans. The OpenAI API is not affected by that notice.
OpenAI’s guidance calls for reviewing configurations that explicitly select GPT-5.5, including saved model settings, custom agents, scheduled tasks, and scripts. Do not assume every workflow has automatically switched to the same replacement.
Arctic Blossom Takeaway: Test a familiar prompt using sample information. Check formatting, calculations, and handling of missing details. Choose a supported model available to your account and client, and update your workflow notes if its behavior changes.
The Bigger Picture
The name of an AI product rarely tells you everything you need to know. Your account, license, workspace permissions, model, integrations, and sharing settings all affect what actually works.
This week, make a short inventory of the AI workflows your team depends on. Record who owns each one, what it connects to, and how you will test a replacement before a deadline or product change disrupts it.
Sources & Further Reading
OpenAI: Custom GPT retirement and migration FAQ — account-specific timelines, migration behavior, and Enterprise milestones.
OpenAI: Moving custom GPT workflows to plugins — testing, integration, and sharing guidance.
Microsoft: Get answers and insights about your Copilot Notebook — reference limits, supported sources, and account requirements.
Perplexity’s announcement feed on X — October 14 Ultrafast announcement. Archived rendering of the company’s post, used to confirm its wording when the direct post could not be retrieved.
OpenAI: ChatGPT and Codex changelog — GPT-5.5 retirement notice.
Educational disclaimer: This article is for general educational purposes and is not legal, compliance, or professional advice. Features, availability, and rollout dates may change. Verify current vendor documentation and follow your organization’s policies before using AI with business information.
