How I Built Odd$mith Almost Entirely With OpenClaw

Update: Odd$mith has grown far beyond the version described in this original post. I wrote a follow-up on what changed, what’s automated now, and how much of the product was actually built with AI: How Much of Odd$mith Was Actually Built by AI?

The AI Workflow Gap

Most AI-assisted software development still lives in the realm of isolated tasks: code generation, debugging, or documentation. But what if AI could build a product as a co-developer, operating across the full lifecycle from concept to deployment? That’s the experiment that led to Odd$mith—a sports betting platform built almost entirely through OpenClaw’s stateful agent-based workflow.

From Code Suggestions to Product Iteration

Traditional AI tools excel at discrete tasks: - Generating a function - Debugging a snippet - Writing marketing copy

But these tools often exist in a disconnected layer, unable to maintain context or work within a product’s actual infrastructure. OpenClaw changed that by acting as a persistent agent inside the project’s workspace, interacting with real files, deployment systems, and operational pipelines. This allowed it to:

• Build premium pricing pages with Stripe integration
• Create performance dashboards and watchlists
• Fix production bugs in service workers
• Manage cron jobs and systemd configurations
• Deploy updates via Raspberry Pi and Cloudflare Tunnel

The Stateful Advantage

The key difference was context retention. Instead of starting fresh for each task, OpenClaw could:

1. Inspect existing code to avoid duplication
2. Run commands directly in the project shell
3. Iterate on features without losing context
4. Debug infrastructure misconfigurations

This eliminated the mental overhead of switching between:

• Idea conception
• Code implementation
• Deployment testing
• UX refinement

What Was Built

Odd$mith evolved from a simple picks concept into a full platform with:

• Daily slates and matchup analysis
• Props pages with real-time odds
• Subscription billing and access control
• Personalized watchlists and performance tracking
• PWA support with install prompts
• Blog system with SEO-optimized content

Behind the scenes, OpenClaw handled:

• Infrastructure: Nginx proxying, static/media handling, cron scheduling
• Security: Stripe webhook validation, password reset flows
• Deployment: Raspberry Pi setup, Cloudflare Tunnel for CGNAT

The Human-AI Balance

AI didn’t replace judgment—it amplified it. OpenClaw executed tasks, but human decisions drove:

• What features to prioritize
• When to gate content behind subscriptions
• How to balance complexity vs. usability

This hybrid approach let me iterate rapidly while maintaining control over product direction and quality.

The Future of AI-Driven Development

OpenClaw’s model shifts AI from a tool to a co-builder. By operating within the actual product environment, it becomes a stateful collaborator, not just a code generator. For sports betting and other domains requiring rapid iteration, this represents a paradigm shift—from isolated AI tasks to continuous product evolution.

Explore OpenClaw’s documentation or view the Odd$mith codebase to see how this workflow operates in practice.