Sitelet https://github.com/adrianwedd
Skip to content
View adrianwedd's full-sized avatar
22 followers · 45 following

Achievements

Achievement: Pair Extraordinairex4Achievement: Pull Sharkx4Achievement: QuickdrawAchievement: YOLO

Achievements

Achievement: Pair Extraordinairex4Achievement: Pull Sharkx4Achievement: QuickdrawAchievement: YOLO

Highlights

Organizations

@HomesTasmania @failurefirst

Block or report adrianwedd

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
adrianwedd/README.md

Adrian Wedd — systems architect (Tasmania)

I build evaluation infrastructure for agentic AI and ship the surrounding tooling — MCP servers, voice and embodied interfaces, ops plumbing. Seven years at Tasmania's Department of Communities; wrote Homes Tasmania's first generative-AI policy. No formal degree.

Now

  • failure-first — adversarial evaluation framework for agentic AI. 13,988 scenarios across 414 attack classes; multi-model vulnerability runs. (research repo private)
  • spark — persistent embodied AI resident, not a chatbot in a chassis: continuous memory, provenance-tracked authority, deterministic crisis fallbacks. Built on PiCar-X + Raspberry Pi for a neurodivergent kid.
  • afterwords — zero-shot voice cloning on Apple Silicon. Gives Claude Code a voice. ~15 voices included.
  • rlm-mcp — MCP server implementing the Recursive Language Model pattern. Lets Claude Code work over 1M+ character documents via session-based chunking and BM25.

Shipped

  • thiswasntinthebrochure.wtf — co-parenting ADHD, autism, PDA & ODD. Four localised editions (AU/UK/US/NZ), ≥70% peer-reviewed citations.
  • Agentic-Index — scored, curated catalogue of autonomous AI tooling.
  • adrianwedd.com — living workshop. AI safety writing, NotebookLM-generated audio + cinematic video on every post, Astro 6 + Tailwind 4.
  • Client work: Evolve Chiropractic, Wolf Clan Zen Do Kai, Tassie Mini-Z, Clutterbusters, Readings with Rowena.

Stack

Python, TypeScript, Bash. Anthropic SDK, MCP, LangGraph, OpenRouter, vLLM, Ollama. FastAPI, Astro, Cloudflare Workers + R2, PostgreSQL, Redis. Pytest, Playwright, custom adversarial harnesses.

Links

adrianwedd.com · failurefirst.org · thiswasntinthebrochure.wtf · status

Pinned Loading

  1. failurefirst/failure-first failurefirst/failure-first Public

    Adversarial evaluation framework for embodied and agentic AI — failure-first methodology, jailbreak corpus, VLA red-teaming, and policy research.

    Astro 4

  2. spark spark Public

    SPARK — a Claude-powered robot companion for a neurodivergent kid. Built on SunFounder PiCar-X + Raspberry Pi.

    Python 6 1

  3. afterwords afterwords Public

    Give Claude Code a voice. Zero-shot voice cloning on Apple Silicon. Voices included.

    Python 2

  4. why-demonstrated-risk-is-ignored why-demonstrated-risk-is-ignored Public

    Research portfolio on why organisations fail to act on known, demonstrated risks: canonical essay, dependency-ordered research backlog, codebook, and agent-assistance audit trail.

    Astro

  5. adrianwedd.com adrianwedd.com Public

    adrianwedd.com — living workshop

    Astro

  6. before-the-words-existed before-the-words-existed Public

    Gibson encoded cognitive mismatch in 1984, before the vocabulary existed to name it — a forensic, experiential reading of Neuromancer.

    Python