Skip to content
View cclawton's full-sized avatar

Block or report cclawton

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
cclawton/README.md

Craig Lawton

I build agentic systems, evaluate foundation models for Australian regulated sectors, and use AI as a creative instrument. The three connect: I evaluate the models, orchestrate the agents, and put the output to work.


Agentic systems

Headless autonomous loops running on a Mac mini — workers that produce, a supervisor that triages, Signal for delivery. No dashboards, no polling.

  • supervisor — two-layer producer/verifier architecture. Workers write run files; a separate triage agent classifies them and signals only when something actually matters. Integrates with any tool that writes to the filesystem — Obsidian, CI, scrapers, task managers.
  • bullpen — profile-driven content pipeline using Claude Code subagents. Researcher, drafter, rhythm polisher, trimmer, safety reviewer, image prompter — each with a distinct role. Voice and constraints are config, not code.
  • podcastindex-mcp-server — MCP server for the Podcast Index API. Exposes search, episode lookup, and feed discovery as tools any MCP-aware agent can call.

AI evaluation

Leaderboards are useful, but they do not tell you how a model behaves in a specific operating context. I am building tools that measure model performance against local tasks, constraints, and evidence.

  • hexapla: early evaluation harness for running repeatable model tests, scoring outputs, and producing comparable result matrices.

AI creative

Writing and music are the same problem: structure, voice, and knowing when to stop. I use AI as a collaborator on both.

  • bullpen — the writing side. Multi-agent pipeline from research through publication, with per-profile voice configuration.
  • asian-sentry-techniques — the music side. Orchestration and composition notes from working in Logic Pro with foundation models. AI as co-composer, not autocomplete.
  • music21-mcp — MCP server exposing 16 music21 MIDI tools to AI agents. Key detection, chord analysis, transposition, velocity, quantization, reharmonization, melody extraction, form analysis, and pattern search. Spec-driven, TDD, 120 tests.
  • reascript-mcp — MCP server for Reaper/ReaScript project control and state readback. Parses .rpp projects, generates Lua ReaScripts, and writes smoke-test scripts for REAPER execution. The DAW execution layer for the agentic music stack.

Writing

I write about AI deployment in Australian regulated sectors, sovereign infrastructure, and building with agents at craiglawton.com. AWS-affiliated — I disclaim my role, not my opinions.


Connect

Pinned Loading

  1. podcastindex-mcp-server podcastindex-mcp-server Public

    MCP server exposing 25+ PodcastIndex.org tools to AI agents — search, discovery, episode lookup, trending feeds, Value4Value data

    TypeScript 3

  2. simple-guitar-tuner simple-guitar-tuner Public

    Android instrument tuner built with AI-assisted development

    Kotlin 1

  3. asian-sentry-techniques asian-sentry-techniques Public

    AI-assisted orchestration and composition techniques — Logic Pro + foundation models

  4. bullpen bullpen Public

    Profile-driven AI writing pipeline using Claude Code subagents

    Python

  5. supervisor supervisor Public

    Autonomous supervisor agent — two-layer producer/verifier architecture with Hermes cron and Signal delivery

    Shell