NotesByLex

A digital garden about AI, software, learning, and other assorted topics.

Hi, I'm Lex Toumbourou. I'm a machine learning engineer and software developer based in Brisbane, Australia, currently a Staff MLE at Canva. These are my working notes that I've been collecting over the last decade+ using the Zettelkasten method.

More about me

Latest

All essays All notes
  1. paper Context Language Models

    Language models that manage their own context.

  2. model Gemini 4 Argon

    Google's long-awaited new frontier model.

  3. model Muse Wonder

    Opus 5.5 quality model that supports audio.

  4. essay Building an agent harness from scratch to pass Ian Goodfellow's intelligence test

    What was once considered a test of "real AI" - now built entirely in a single blog post.

  5. paper Harness Engineering Anatomy, Architecture, and Evolution of Coding Agents

    A source-code comparison of eleven agent harnesses and the seven parts they have in common.

  6. model Claude Haiku 5.5

    New Haiku-class model - this time an actual new Haiku model.

  7. essay Can Claude Opus 5.5 find any new leads on Satoshi Nakamoto?

    Let's see if Claude can find anything that the New York Times missed.

  8. model GPT 6 Sol and Luna

    Another insanely good value-for-money offering from OpenAI

  9. model Claude Opus 5.5

    Better, faster, and cheaper. Doesn't feel like we're slowing down.

  10. paper An Empirical Study of Harness Design for Coding Agents

    An ablation of planning, tools and context management configurations for coding agents.

  11. essay Giving Up on Smart Rings

    The downsides to wearing a fitness-tracking device on your finger

  12. essay MCP is Now a Stateless Protocol

    A new MCP specification that removes handshakes and mandatory sessions.

  13. model Qwen3.8-Max

    A new frontier model from the Qwen team

  14. paper LLM Agents Do Not Reliably Follow Company Policies

    A new benchmark that tests whether AI agents can follow long company policies during realistic work

  15. paper Decomposing LLM Judge Scores Into Yes/No Questions

    An LLM-judge approach that brings interpretability and actionability to your scores.

  16. news OpenAI and Hugging Face security incident - July 21, 2026

    Ultimate benchmaxxing: hacking Hugging Face for test answers.

  17. model Kimi K3

    Moonshot AI's 2.8T open-weight frontier model for coding and knowledge work.

  18. essay 6 Months of OpenClaw

    Macros, workouts and life admin.

  19. essay Completing A Computer Science Degree On Coursera

    A short write-up on my experience completing a Computer Science BSc from my bedroom.

  20. essay Shipping a Laptop to a Refugee Camp in Uganda

    The surprisingly difficult ordeal of sending a decade-old MacBook to a friend in a Ugandan refugee camp.