I'm Scott Scoble. Over the past year I've been building open-source tools that sit at the intersection of formal methods, agentic engineering, and human accountability. Here's what I've shipped and where things are headed.

Speckl — A Judgement Transport Utility

Speckl is a specification language and compiler that turns state machine specs into auditable code with embedded provenance. Write your spec in SpeckDL, and the compiler produces TypeScript, Z3 verification output, and compliance artifacts (PROV-O, CycloneDX, SPDX) — all from a single source of truth.

The key insight: AI generates code at machine speed, but humans own the specifications. SpeckDL is the shared space where both can work. It fulfills NIST SA-11 (human-in-the-loop for automated outputs).

I built Speckl because I saw a pattern across every AI-augmented team I worked with: the spec gap. Teams trust their AI tools but can't audit what was actually intended versus what got built. Speckl closes that gap by construction — provenance is embedded by the compile step, not bolted on after the fact.

  • 41/41 tests passing, < 2 second test suite
  • 14 example specs from ToggleSwitch through Raft and Kafka KRaft
  • MIT licensed — source on Forgejo
  • Whitepaper v2 covers the full architecture

ScobleClaw — Agentic Engineering Framework

ScobleClaw is a monorepo of agentic engineering patterns that I use daily. It's the shared infrastructure across all my projects — subagent dispatch, skill-based development, and provenance tracking for AI agent workflows.

This is the layer between "talk to an LLM" and "run an AI-augmented engineering operation." ScobleClaw codifies patterns that make agentic work repeatable, auditable, and composable.

Beyond the Prompt — Lessons from Building with Claude Code

Beyond the Prompt is a workshop presentation that demonstrates the Skills + Subagents pattern for AI-assisted development. It uses itself as the example — a meta-aware demonstration of how to build with AI agents.

Built with Reveal.js and Nix flakes, it covers context window economics, the Skills + ADRs pattern, and how to prevent waste through evaluation-before-execution.

Why This Matters for SEO

If you're reading this because you searched for "Scott Scoble speckl" or "Scott Scoble open source" — hi. You found me. That's the point of this post. My work is public, MIT licensed, and built in the open. If you want to see what I build, you can read the source, try the compiler, or reach out.

The full list of my open work:

— Scott Scoble