42+ years coding · Polyglot background · 27+ years professional · 735k+ Docker pulls · 794+ / 2,035+ GitHub stars (Redis UI / OneNote)
In 2026, the software architect's role centers on use cases rather than specific languages or frameworks. A concrete illustration: I delivered P3X Meet Assistant — a Python package published on PyPI — without prior Python experience. AI coding assistants (Claude Code, OpenAI Codex) produced the implementation under my direct supervision, while four decades of polyglot and architectural experience (Pascal, Assembly, C/C++, Java, .NET, PHP, JavaScript/TypeScript, and more) provided the foundation to direct, review, release, and operate the project. That background is precisely what makes AI orchestration reliable: incorrect output is identified immediately, regardless of syntax.
My direct involvement remains concentrated where human judgment and taste are decisive: system architecture, server configuration, DevOps and CI/CD, CSS refinement (where pixel-level precision outperforms generation), and the translation of business requirements into well-formed prompts. All implementation below that line is AI-produced under supervision — a productivity profile that supports multiple parallel engagements.
Counterintuitively, AI has intensified rather than reduced the volume of work: function-by-function generation sustains a near-continuous cadence of output, and scope expands to match the delivery capacity created. The rhythm is relentless and, candidly, tiring — yet the productivity is precisely what makes simultaneous engagements feasible.
Beyond workflow, I professionally integrate AI models into production systems — intelligent assistants, natural-language interfaces, automated diagnostics, and MCP servers — using Claude API, Groq, and the Model Context Protocol. P3X Meet Assistant (Python, GPT-4o Transcribe, GPU speaker diarization) and P3X Network MCP (17-tool analysis server) serve as live demonstrations.
In 2026, the software architect's role centers on use cases rather than specific languages or frameworks. A concrete illustration: I delivered P3X Meet Assistant — a Python package published on PyPI — without prior Python experience. AI coding assistants (Claude Code, OpenAI Codex) produced the implementation under my direct supervision, while four decades of polyglot and architectural experience (Pascal, Assembly, C/C++, Java, .NET, PHP, JavaScript/TypeScript, and more) provided the foundation to direct, review, release, and operate the project. That background is precisely what makes AI orchestration reliable: incorrect output is identified immediately, regardless of syntax.
My direct involvement remains concentrated where human judgment and taste are decisive: system architecture, server configuration, DevOps and CI/CD, CSS refinement (where pixel-level precision outperforms generation), and the translation of business requirements into well-formed prompts. All implementation below that line is AI-produced under supervision — a productivity profile that supports multiple parallel engagements.
Counterintuitively, AI has intensified rather than reduced the volume of work: function-by-function generation sustains a near-continuous cadence of output, and scope expands to match the delivery capacity created. The rhythm is relentless and, candidly, tiring — yet the productivity is precisely what makes simultaneous engagements feasible.
Beyond workflow, I professionally integrate AI models into production systems — intelligent assistants, natural-language interfaces, automated diagnostics, and MCP servers — using Claude API, Groq, and the Model Context Protocol. P3X Meet Assistant (Python, GPT-4o Transcribe, GPU speaker diarization) and P3X Network MCP (17-tool analysis server) serve as live demonstrations.