1. Define the goal and boundary
Lock the positioning, constraints, and release criteria first so AI execution does not keep rewriting the direction.
Gleam Lab · Workflow
AI-assisted Project Workflow
This page explains how wzk.icu moves with human judgment plus AI execution: define boundaries, ship in batches, and verify with build, links, SEO, and live checks.
Operating Model
Core Principle
AI expands execution bandwidth, reduces repetitive work, and helps catch omissions. Human judgment owns the main narrative, public boundary, factual verification, and release decision. This is not a packaged methodology; it is the site’s practical maintenance workflow.
2. Split work into verifiable batches
Separate pages, data, SEO, bilingual consistency, and check scripts into batches that can each be built and verified.
3. Connect content as project assets
Posts, series, topics, projects, FAQ, and roadmap pages should reference each other so the site becomes a clear evidence system.
4. Verify locally before deployment
Build, link checks, SEO checks, i18n checks, sensitive-content checks, and live curl checks happen before rsync deployment.
Guardrails
Collaboration Boundaries
- Do not present future capabilities as delivered capabilities.
- Do not expose internal endpoints, accounts, secrets, MQTT defaults, production interfaces, or customer information.
- Do not stuff keywords for SEO; core pages should remain truthful, restrained, and explainable.
- Do not let AI replace final judgment; positioning, public boundaries, and release responsibility stay human-owned.