Feedback & Insights

Looks like this file is over the limit. I work best with documents under 1000 pages and 50 MB in size. I'm learning and improving every day, so feel free to share your feedback here!

Large editing projects are currently executed as disconnected one-off instructions with no overarching plan. The AI should be able to build and present an execution plan upfront — e.g. Page 1: identity updates, Page 2: textbox replacement, Page 3: history updates — so users can review scope before work begins and the system can execute coherently across the full document. Expected outcome: enterprise workflow acceleration.

Presentation editing is structurally different from standard PDF editing. The AI should be able to recognise and target named section types — headers, subheaders, charts, bullet sections, contact blocks, metrics sections, revenue sections — so that a request like "Update Slide 5 revenue forecast" resolves automatically to the correct content area without manual specification. Expected outcome: stronger pitch deck and proposal workflows.

When a user says "replace left textbox on Page 2," the system attempts an exact text search instead of identifying and targeting the textbox as a spatial object. Editing reliability would improve significantly by supporting a semantic pipeline: page → textbox detection → object recognition → coordinate mapping → content replacement. Software should align with how users naturally think about their document. Expected outcome: major improvement in editing reliability.

The AI repeatedly loses context mid-session. Users expect the system to retain formatting instructions, previously stated document objectives, and editing history across a workflow. For example, issuing "Update Slide 4" should not require re-explaining prior formatting requirements — the system should already know them. Expected outcome: less friction, higher completion rates, greater user trust.