What's New in NetworkOS

You confirm changes before they save. When the agent wants to create or edit something in your workspace, you see exactly what will change first — including a before-and-after on edits. Nothing saves until you confirm, and you can cancel with nothing written.
The agent asks when it needs to know more. It asks right in the conversation, with options you can pick from, and you can always type your own answer instead. If there are several questions, you can answer them all before it continues.
See a Mission's whole story in one view. A Mission now opens straight to its description and everything produced in it, together on a single page — no sub-sections to navigate. The Missions page shows the creator of each Mission alongside it, so you can tell your team's work apart at a glance.
Missions are where your work product lives. A Mission collects what you and the agent produce inside it — the artifacts that come out of the work, and the files that provide context. Records are data powering your full workspace, and are no longer scoped to specific missions.
Know what a skill does and how to use it before you run it. Each Agent Skill page now carries its full description, how it works, and how to invoke it — so you can find the right skill for the job and put it to work.

See every result, not just the first fifty. Expanding a record table now gives you access to all the rows that came back with the output, with paging in both directions.
Categorize the organizations that anchor your ecosystem. Federally Funded Research and Development Centers (FFRDCs), University Affiliated Research Center (UARCs), Government Laboratories, Accelerators, and Investors are now organization types you can select — so the institutions at the center of government innovation are described accurately.

Typing into the organization fields of the create-new window now keeps every character, in the order you typed it.
Refreshing a page keeps you on the page you were reading instead of sending you back to the homepage.

Know how a skill gets invoked. Each Agent Skill now tells you whether it's Command or Ambient. Ambient skills can run automatically when the conversation calls for them — and you can still invoke them by command whenever you want.
Artifacts look the same everywhere you find them. Every place that lists Artifacts in a table now uses the same one, with a single filter for switching between all Artifacts and just your own.
Cleaner tables in Terminal. In-Network and New status now appears only on Organizations, where it actually means something — People and Event rows are no longer cluttered with a status that doesn't apply to them.

Know what's behind a Data Pack. Each Data Pack's About section now links out to the source its data comes from and tells you how often it refreshes — so you can judge how current an answer's underlying data is.

Choosing Workspace when saving a generated output now saves it to your Workspace, not to the Mission you were working in.
Google Drive syncing no longer gets stuck showing "syncing" indefinitely.

N/A

The Sub-Organizations section on an Organization record now displays with the correct heading, weight, and indentation.
Password-reset emails now show as an actual reset instead of looking like an invitation.

Answers grounded in your Workspace. The AI now searches your Workspace before answering substantive questions, so responses lead with your own curated records.
Data Packs get detail pages. Each Data Pack now has its own page showing what it provides, its availability, when its data was last refreshed, and which Agent Skills use it — with an enable/disable toggle right on the page.
The Acceptable Use Policy replaces the EULA. You'll now see our Acceptable Use Policy everywhere the End User License Agreement used to appear — the acknowledgment screen, login footer, and account menu.

Leaving a conversation while a job is running no longer strands it — when you return, you'll see whether it's still running, finished, or failed, and the conversation shows up in your history without a refresh.
A job that fails mid-run is now reported as failed instead of quietly looking complete, so you know to retry.

What's New in NetworkOS - August 2026