Erdeniz TuncProduct Engineer
WorkAboutWritingPressWork with me
← All workPrototype complete

Optivideo · Optifeed · 2026

Designing and Shipping a Feature End-to-End

An AI video generation prototype for product feed advertising, built with Claude Code and Codex

01 Status

Current status

Prototype complete and live. Two internal stakeholder feedback sessions. Solution spec handed to engineering. First technical version in development. Not launched; no customer usage claim.

02 Summary

Optivideo is a working prototype for AI product video generation inside Optifeed's product feed platform.

I owned the product direction, built the interactive prototype with Claude Code and Codex, and shaped the flow through two internal stakeholder feedback sessions.

The prototype covers product selection, scenario/template selection, confirmation, review, approval, and channel export. After iteration, I handed engineering a solution spec based on the working prototype.

Optivideo working prototype overview
Working prototypeThe feature direction exists as a navigable product artifact.

03 Problem

E-commerce and marketing teams often have product data and images, but product-level video creative is slow, manual, and hard to scale across catalog ads.

Optifeed already sits at the feed layer: product data, product images, and channel distribution. That made AI video generation a logical product direction, but only if the workflow gave users enough control before anything moved toward an ad channel.

04 Riskiest Assumption

What needed to be tested

The riskiest assumption was not whether AI could generate a video. It was whether a marketing or e-commerce user could understand, trust, review, approve, and export a self-serve AI video flow without losing confidence.

05 My Role

  • Product direction
    Discovery, PRD/spec thinking, and scope decisions for the first testable version.
  • Prototype build
    AI-assisted prototype build with Claude Code and Codex.
  • Feedback and iteration
    Two internal stakeholder feedback sessions and iteration after feedback.
  • Engineering handoff
    Solution spec handed to engineering with flow, decisions, states, and boundaries.

06 Prototype Flow

  • Start
    Product selection and campaign setup.
  • Configure
    Template/scenario selection, including the Textile/Fashion branch.
  • Commit
    Generation confirmation before the user spends tokens or waits for output.
  • Control
    Review, approval, reject/regenerate, and explicit channel export.
Optivideo prototype product selection from catalog data
Product selectionThe flow starts from Optifeed's core advantage: product data and images already in the platform.
Optivideo prototype template and scenario selection including Textile/Fashion
Scenario selectionThe prototype made scope and category decisions visible before engineering build.

07 Key Decisions From Feedback

  • Mandatory approval
    Review and approval before export, so no video moves toward a channel automatically.
  • Generation confirmation
    Token cost and estimated generation time shown before generation begins.
  • Focused first scope
    Textile/Fashion chosen as the first engineering scope.
  • Explicit export control
    Channel selection before export, so users decide where approved videos go.
Optivideo confirmation step with token cost and estimated generation time
ConfirmationUsers see cost and waiting time before committing to generation.

08 Engineering Handoff

The handoff included the live prototype and a solution spec based on the working flow.

The goal was to give engineering a clearer starting point: product logic, key decisions, visible states, and explicit out-of-scope boundaries.

Optivideo prototype channel export selection
Channel exportExport requires explicit user choice and keeps control in the user's hands.

09 Current Status

The prototype is complete and live. The solution spec has been handed to engineering.

Engineering owns the first technical version now, and that version is in development.

The feature has not launched. There are no customer usage, adoption, revenue, ROAS, or performance metrics yet.

The feature is designed for rollout to Optifeed's existing B2B customer base. That customer-base context is not an adoption claim.

10 What This Proves

  • Artifact before build
    I can turn a product idea into a working artifact before engineering commits to the build.
  • AI-assisted first version
    I can use AI coding agents to create a testable first version while keeping product judgment in front.
  • Feedback-driven flow
    I can test and improve product flow through internal stakeholder feedback.
  • Stronger handoff
    I can hand engineering a clearer starting point: prototype, decisions, states, and solution spec.
Get in touch → Explore the prototype: optivideo.vercel.app
Next case study
OptiGTM →
Back to
All work