You won't see a new button or a new menu. But if you've used an agent in the last day or two and felt it land a bit cleaner — the conversation stays more focused, the edits feel more surgical, the deliverables are easier to actually use — that's intentional. A behind-the-scenes upgrade rolled out across every agent on the platform this week.
What changed, in plain terms
Editing is surgical, not all-or-nothing
Before, asking an agent to "tweak the second paragraph" sometimes meant getting the whole document rewritten — losing your tracked changes and forcing a re-read of the entire output. Now agents edit in place: the change you asked for happens to the bit you asked about, and the rest stays put.
Conversations stay on the question
Agents are now more disciplined about staying inside the brief. If you've asked for a competitive audit, they don't drift into adjacent strategic territory unprompted. If you want them to widen the lens, you ask — and the conversation stays where you want it.
Editable documents alongside every PDF
The biggest practical win. When an agent produces a deliverable, you now get both a polished PDF and an editable version sitting next to it. No more screenshotting a slide because the export was locked, no more re-running the agent to get a Word version, no more "can you send me this in something I can edit?" The editable copy is right there, ready to drop into your own workflow.
Smarter link reuse
Agents now remember the links and outputs from earlier in the same chat. If the agent produced a presentation in turn 3 and you ask in turn 5 to "send the deck again," it returns the same URL — it doesn't regenerate or ask you to point at the file. Small thing; saves real friction.
Why we did it
These are the four nudges that turn an AI that impresses into one that's actually useful in a working day. Surgical edits respect your time. Conversational focus respects your brief. Editable docs respect that the agent's output is the start of your work, not the end. URL reuse respects that conversations are a thread, not a series of one-shots.
Where you'll feel it
Everywhere. Every agent in the Agent Library inherits these defaults — from the Daily Social Idea agent to the SEO Technical Audit to the Value Proposition agent. No setting to change, no toggle to find. The next chat you run will already be running on the new behaviour.
If you spot an agent that doesn't behave this way — a stray "shall I rewrite the whole thing?" or an output that arrives PDF-only — let us know. The platform is meant to apply these defaults uniformly, and edge cases help us tighten the screws.