Live · product factory

DivePlan

Closed-loop multi-model product factory.

Grok runs the brain, critic, and score. Your Google AI Chat/Canvas sessions do the high-effort generation — until the work clears the bar.

On-device UI · product visualization

9:41
SYSTEM COCKPIT // RUNTIME BUS_LIVE
ACTIVE MISSION

zacgibson.work

Intake — scope the portfolio island.

01 · Intake
GENERATORSTANDBY
Google AI
consumer · Chat/Canvas
CRITICREADY
Grok
brief · score · gate
ARTIFACTQUEUED
Portfolio island — waiting for cycle

Product truth

What is locked. What is next.

Mobile-first iOS control plane. Official APIs stay secondary. Exceptional is a numeric rubric — not vibes.

Primary generator

Google AI consumer — Chat + Canvas

Gemini API = structured / fallback only.

Critic / scorer

Grok — PRD, style prompts, rubric, gates

Scorecard + threshold stop. Human override retained.

Build state

CHAPTER 1 LOCKED · CHAPTER 2 NEAR-LOCKED

Native iOS foundation modularized. LoopEngine models shipped. Design system preserved.

Next

SUBSCRIPTION SESSION LAYER

Assisted copy/paste path first. Real browser control is owner-gated and ToS-aware.

Skill · prd-perfection

The perfection loop

One closed cycle. Cap the spins. Audit everything. The phone above plays the public stage labels — synced here as it runs.

  1. 01
    Intake → Grok PRD vN Strategic brief, not a vibes dump.
  2. 02
    High-effort expansion prompt Style engine composes the ask for the generator.
  3. 03
    Google AI max expansion Consumer Chat / Canvas as the generator.
  4. 04
    Score against rubric Rewrite or lock. Numeric bar only.
  5. 05
    Decompose → package → remote Per-task loops, then workers.
  6. 06
    Audit trail sealed Cycle complete. Cap the spins.
// audit · sample spine
ok mission · zacgibson.work · viz
ok critic · grok · rubric_bound
ok generator · google_ai_consumer
wait mac_device_proof · browser_path_tos

Built with

Cursor Grok Google Xcode