Start up companion

Inspiration

Quiet Performance began with a familiar startup problem: important information was scattered across conversations, documents, images, spreadsheets, and ecommerce tools. Product ideas, design references, samples, supplier discussions, tasks, and expenses all existed, but not in one place where they could guide daily decisions.

I wanted to create a calm, founder-focused companion that could turn long-form thinking into organised execution without becoming another complicated business system.

What it does

Quiet Performance Startup Companion brings the operational side of the startup into one connected workspace. It provides:

• A visual library for uploading design ideas and inspiration
• Product-specific design history and current-design tracking
• Sample management and feedback
• Supplier discussions and commitments
• Notes, decisions, and supporting files
• Prioritised tasks and progress tracking
• Editable NZD expenses with receipt evidence and audit history
• A focused overview of current priorities, activity, and spending

The aim is to help a founder understand what is happening, what has changed, and what needs attention next.

How I built it

I developed the project through a Chat-to-Codex co-creation workflow.

Chat was my thinking partner. All startup work took place inside a dedicated Chat project, where I explored ideas, challenged assumptions, refined requirements, and made product decisions through extended conversations.

Those discussions were then transformed into large, contextual, and actionable prompts for Codex. Codex acted as my execution tool: inspecting the existing repository, implementing features, preserving data, testing changes, improving the responsive interface, and preparing deployments through OpenAI Sites.

The application uses React and TypeScript, with Cloudflare D1 for structured information and R2 for attachments. Validation, migrations, permission checks, and important mutations remain server-side.

I used a similar discussion-to-execution method while developing the customer-facing Shopify store, with project-generated prompts and computer use in the in-app browser.

Challenges

The biggest challenge was deciding what the product should not become. It needed to connect designs, samples, suppliers, tasks, and costs without turning into a heavy ERP or accounting platform.

Another challenge was balancing flexibility with traceability. Design ideas needed a standalone visual space while still connecting to products. Expenses needed to remain editable without losing receipt evidence or change history. Existing startup information also had to survive each database and interface revision.

Finally, AI-generated suggestions needed clear boundaries. The companion can assist with context, but it cannot silently create authoritative tasks or overwrite important business records.

What I learned

I learned that AI collaboration becomes much more powerful when thinking and execution are treated as separate but connected activities.

Chat helped me work through uncertainty and develop the reasoning behind each decision. Codex was most effective when it received the resulting context as a clear implementation brief rather than an isolated one-line request.

I also learned that building with AI is still an iterative product discipline. Human judgment remains essential for deciding what matters, resolving trade-offs, reviewing outcomes, and determining the next step.

Accomplishments

I am proud that the project evolved from startup conversations into a functioning operational product with real data structures, migrations, responsive interfaces, automated tests, deployment configuration, and a documented repository that another person can run locally.

More importantly, it demonstrates a repeatable way for a non-traditional builder to move from thoughtful discussion to working software.

What's next

Next, I plan to deepen the connection between product development and the Shopify storefront, using my companion as the basis for creating the shopify tool using computer use in codex, which is already well underway.

Built With

• chatgpt
• cloudflare
• codex
• drizzle
• next.js
• openai
• react
• shopify
• sites
• tailwind
• typescript
• vite
• vitest
• zod