FlytBase / Shipped product / FlytClip

Drone footage became a report operators could approve and share.

An outside recipient needed to understand a drone mission without opening FlytBase or assembling the story from raw files. I designed FlytClip around an evidence package that carried media, context, and reviewed findings together. The operator could edit AI drafts, approve the report, and share its evidence through a controlled link.

FlytClip Studio with drone video, selected media, and report review controls.
The shipped Studio is where the operator selects media, adds context, and prepares the report for review.
My role
Product strategy, design, and AI-assisted frontend / backend implementation.
Ownership
Identified the problem and owned the product from definition through shipment.
For
Drone operators and external recipients reviewing mission evidence.
ProblemSharing outside FlytBase

The mission ended. The handoff didn’t.

One customer’s recipient had no FlytBase access. The operator had to find media, select evidence, and assemble a long PDF before anyone outside the platform could understand the mission.

Flight logs and Reports already served record-keeping. I built the missing handoff: selected media, context, and reviewed findings in a recipient view that exposed the supporting evidence.

  1. 01

    Find

    Locate the relevant mission media.

  2. 02

    Curate

    Decide what matters and add context.

  3. 03

    Review

    Own the complete composition.

  4. 04

    Share

    Let the recipient inspect the evidence.

I prioritized the DFR handoff, where a responder may need the evidence without being a FlytBase operator.

Product modelPackaging media with context

The report had to carry the footage’s context.

A clip link left the operator explaining what the recipient was looking at. I made the shareable object an evidence package: selected media, mission context, findings, observations, and review state.

The original media stayed in FlytBase. The same package carried its context from the first working prototype through review and into the recipient’s view.

FlytClip media selection showing mission clips ready for an evidence package.
Select mission media first. The package carries its operational context through to the recipient.

What a link left behind

The recipient gets a clip. The sender still has to explain what it shows and whether its conclusions were reviewed.

What travelled together

Media, context, and review state stay together as the operator prepares and shares the account.

Video playbackKeeping clips tied to the source

The recipient needed the moment and its source.

To inspect a finding, the recipient needed the relevant part of the source. A rendered-trim path briefly existed, but it added processing, storage, and another file to manage.

The flow changed to saving a playback window against the original video. The operator sets the range in Studio; the controlled viewer presents that saved portion of the same source, without generating a separate trimmed file.

FlytClip Studio playback range controls beneath the original mission video.
Studio capture: the orange range handles and Save playback window control define the source segment to present.

What this removed

A second video-processing and storage lifecycle. FlytClip saves the source reference and playback window rather than another rendered clip.

What it depended on

The original video must remain available, and the permitted window must be enforced reliably. The shared experience is view-only, not a downloadable clip.

Expiry and revocation close the share link. They do not delete the original Gallery video.

Report reviewOperator approval before sharing

AI could draft. The operator approved the report.

A generated report could look complete before anyone checked it. I kept findings as candidates and let the operator edit the account while inspecting the exact report prepared for the recipient.

Inline edits shaped the findings; one approval committed the complete report. That removed repeated approval actions, but put more weight on checking every required field and evidence plate before publication.

FlytClip report drawer with draft findings and approval controls.
The report drawer keeps candidate findings in a private review step before approval.
  1. 01

    Private analysis

    Candidates stay in the workspace.

  2. 02

    Review & edit

    The operator shapes the report.

  3. 03

    Approve

    One decision on the composition.

  4. 04

    Share

    Only the approved view is public.

Evidence changes → review again

Changing media or context makes earlier analysis stale. Missing references and incomplete results cannot cross the approval boundary.

Only the approved account

The public payload is explicitly allowlisted. Prompts, provider responses, rejected candidates, and diagnostics remain private.

Claim attributionSeparating operator notes from AI

The report had to show who made each claim.

The first rule kept a note private unless it supported an AI finding. In the running product, that meant the model could silence knowledge the operator brought from the site.

I gave approved Operator Observations their own place and linked evidence. AI Findings stayed separate, so readers could see who was making each claim.

AI Finding

A model-generated candidate, reviewed by the operator. Its source is visible.

Operator Observation

A human assertion, approved independently. Its evidence travels with it.

Shared recipient view with source video and the approved report.
The recipient can inspect source media beside the approved report, with operator observations and AI findings kept distinct.

Independent observations preserved the operator’s voice, at the cost of more provenance and publication rules. Evidence Plates kept both kinds of claim inspectable; weak crops fell back to the analyzed frame.

ImplementationBuilding and testing the report flow

The review rule had to survive the build.

I built on FlytBase’s shared app foundation. Agent-readable rules guided Codex through small working slices, first with mock contracts and then with real media and services.

Codex helped implement the slices; the report AI drafted findings inside FlytClip. I checked the running behavior against the evidence, review, and publication rules, then owned the tests, staging checks, and release judgment.

  1. 01

    Frame the rule

    Define meaning, private states, and scope.

  2. 02

    Build a slice

    Constrain the agent with mock contracts.

  3. 03

    Test the meaning

    Review behavior with real media.

  4. 04

    Verify & ship

    Revise UI, contracts, and checks together.

OutcomeSharing an approved evidence report

Recipients could inspect what the operator approved.

FlytClip shipped a connected path from mission footage to a recipient-facing report. Context, source evidence, and the origin of each claim stayed available for inspection; the operator approved what was shared.

I now start with what the recipient must be able to check: whose claim it is, what supports it, and what approval allows them to see.