About Silver Wizard Software

Steve Deighton

The Founder

Steve Deighton
74 years old. Retired real-time software engineer, consultant, and program manager.

Connect on LinkedIn →

The Unusual Part

Steve hasn't written or read a single line of code for any Silver Wizard product.

Every application, every tool, every line of infrastructure — built entirely with AI collaboration through Claude Code.

But don't mistake "AI-built" for "amateur hour."

The Foundation

20 years as a real-time software engineer: Telco-grade systems. The kind that can't go down. The kind where "good enough" means lives are at risk.

DevOps and cloud experience at EMSA: Modern infrastructure. Production operations. What it takes to ship and maintain systems at scale.

Consulting and program management: Decades of seeing what works, what breaks, and why projects fail.

"Got a bit bored of paradise. And AI is truly interesting, remarkable, and a software engineering game changer."

All the languages I once used — Assembler, Algol 60, Algol 68, CHILL, Pascal, Borland OO Pascal — are out of date. But software engineering principles, build-to-test methodologies, and telco reliability standards are timeless.

So I deliberately do NOT touch the code. I ensure telco principles are followed and my modules stay in the 400-800 lines of code "sweet spot" — the size where complexity stays manageable and testing stays comprehensive.

The Philosophy

AI as co-developer, not replacement

Claude Code isn't autocomplete. It's a collaborator that brings implementation speed while Steve brings architecture, requirements, and standards.

Telco-grade principles still apply

  • Systems that don't crash
  • Error handling that assumes failure
  • Logging that helps debug production issues
  • Documentation that survives the original author

Build AI into apps where it adds value

Not because it's trendy. Because specific problems (content creation, UI testing, memory persistence) are genuinely better solved with AI capabilities.

Mac-only. For now.

Mac users expect quality. They value polish. They're willing to pay for tools that work well. That's the market worth building for.

The Result

Production software with decades of engineering discipline and zero lines of human-written code.

Applications that feel professional because the standards behind them are professional.

Proof that AI-native development doesn't mean compromising on quality. It means leveraging AI where it's strong (implementation, iteration) while maintaining human judgment where it matters (architecture, user experience, reliability).

What We Build

🏗️ Multi-Agent Infrastructure

Production systems for coordinating multiple AI agents: messaging, conferencing, reincarnation, fleet management. Built for ourselves, available for licensing.

🛠️ Developer Tools

MCP servers, PyQt testing frameworks, persistent memory systems. Tools that make AI-assisted development reliable and repeatable.

🐕 End-User Applications

Mac apps for real problems: email search across multiple accounts, AI-powered publishing pipelines. Currently in alpha with family users.

🔌 MCP Ecosystem

Model Context Protocol servers that extend Claude Code with specialized capabilities. Open source contributions to the AI developer community.

Get in Touch

Interested in our infrastructure? Have questions about AI-native development? Want to discuss licensing?