Gemini builds the file, then lets you fix it
What shipped in October, and the job each release is good for.
October’s most useful release came with the least noise. The Gemini app can create a file and then edit it with you, so the thing you asked for arrives finished instead of as text to paste elsewhere. Most of the month went that way: releases that hand back work, not answers.
The month in brief
GPT‑6 took over ChatGPT’s Chat tab for every tier of user. Google shipped Gemini 4 Argon for long multi‑step work and moved the Gemini API’s Computer Use tool to general availability. Anthropic folded chat and Cowork into one Claude Enterprise, Meta and Mistral published large open weights, and GitHub made a Copilot session portable.
What shipped
Google: File creation and editing in the Gemini app
Ask Gemini for a document, deck, sheet, PDF or code file and it builds the thing in the chat, where you can edit it or have it redone.
Best use: turn a photo of handwritten notes into a study guide, then fix the two sections that came out wrong without leaving the chat.
Who gets it: all Gemini app users, in every language the app supports. PDFs on web and mobile, the other formats on web only.
Catch: on a phone, PDFs are the only files you can edit.
OpenAI: GPT‑6 in ChatGPT Chat with Intelligent UI
The Chat tab runs GPT‑6 Sol for paid users and Luna for Free and Go, and an answer can arrive as tappable buttons, forms or charts.
Best use: compare options or plan a trip, choosing as you go by tapping the form it draws instead of typing three follow‑ups.
Who gets it: Plus, Pro, Business and Enterprise from 7 October, Free and Go the day after, on web and mobile.
Catch: the Chat tab only. Work and Codex keep the models they had.
Google: Gemini 4 Argon
A reasoning model for long multi‑step work in coding, finance, legal, security and science. It reads text, images, video, audio and PDFs.
Best use: give a coding agent in Antigravity a refactor spanning several files, then review the finished change rather than each step.
Who gets it: the Gemini app and AI Mode in Search for Google AI Pro and Ultra, plus AI Studio, Android Studio and Antigravity.
Catch: it writes text only, and input and output share one window.
Anthropic: Claude Dashboards and Claude Motion
Dashboards answers a plain‑language question against a connected warehouse such as BigQuery, Snowflake or Databricks and draws charts that refresh themselves. Motion turns a report into a short MP4.
Best use: an analyst asks for weekly sales by region, shares the live dashboard, then sends the clip into Monday’s meeting.
Who gets it: Dashboards on paid plans, Motion on Team and Enterprise. Docs, Slides and Design left beta the same day.
Catch: both are beta, and an Enterprise admin has to switch them on.
Anthropic: Claude Enterprise unified experience
Chat and Cowork are one place now, and bigger work runs in an isolated cloud machine that carries on after you shut the laptop.
Best use: ask for a competitor comparison off your shared files, close the laptop, and read the finished document from your phone.
Who gets it: Claude Enterprise plans on web, desktop and mobile. Pro and Max got the same experience back in September.
GitHub: Copilot cloud experience
Each Copilot agent session runs in a cloud sandbox that snapshots itself when it stops, so work you start on github.com resumes in Slack, Teams, the CLI, VS Code or GitHub Mobile.
Best use: on call, ask Copilot in Slack to trace a bug, read the proposed fix on your phone, then finish it in the editor.
Who gets it: paid Copilot plans, rolling out from 28 October. Copilot Free keeps Copilot Chat on github.com and GitHub Mobile.
Google: Computer Use tool in the Gemini API
A model you call from your own code can now work a browser or a desktop: it reads the screen, then clicks, types and scrolls.
Best use: have an agent fetch last month’s invoices from a supplier portal that never got an API, so nobody keys them in by hand.
Who gets it: the Interactions and GenerateContent APIs, now GA.
OpenAI: GPT‑6.1 Terra
A fast model for looking at things: text and images in, text out, with computer use, function calling and a 1,050,000‑token window.
Best use: build an agent that reads supplier drawings and photos and fills in the matching catalog entries, so staff check instead of type.
Who gets it: the OpenAI API, on Responses, Chat Completions and Batch.
Use case of the month
IDC cuts code review from 5 days to about 30 minutes with Claude
IDC, a technology research firm sixty years old, connected its Quanta intelligence platform to Claude through an MCP connector, then rolled Claude out internally, engineering first. Its engineers rebuilt the data repository their analysts work from in three weeks, against an estimated three months. A review that once filled five days now takes about half an hour, and three‑week planning sprints run in three days.