Available for opportunities

Hi, I'm John Henderson

I'm a senior software engineer focused on beneficial AI, working to ensure harms from AI are minimized and benefits are shared widely.

Latest writing View my work Get in touch

Writing

Recent posts

Writings on AI safety.

Research Substack

Is interaction mode a property mode of the task?

Whether people use AI to automate their work or to augment it could shape much of AI's economic impact, but is that choice a property of the task itself? I ranked 491 O*NET tasks by augmentation share in the public WildChat dataset and in the Anthropic Economic Index. The two orderings barely agree (Spearman's rho = 0.065), suggesting the task alone doesn't determine how people work with AI. Written as my project for the BlueDot Impact Technical AI Safety course.

Published August 15, 2026

Read the post View the code

Explainer

Civic AI Canada AI Policy for All Canadians

How to Read a Model Card

Model cards (sometimes called system cards) are documents developers release alongside new AI models, covering what a system can do, what it can't, and how it was tested. They can be tough to read without some background, so this guide walks through them for a general audience. Easy to share with friends and family.

Published August 7, 2026

Read the article

Work

Selected Projects

A few things I've worked on. Click to expand each one.

AISafety.info is a free, volunteer-run resource that answers hundreds of questions about AI safety and the existential risks posed by advanced artificial intelligence. It serves everyone from curious newcomers to researchers, pairing human-written explanations of core concepts with an AI-powered distiller that summarizes the broader alignment literature.

GitHub Live site

A chronological timeline of Canadian AI governance events and policy developments. The tracker maintains a curated record of significant milestones — Senate hearings, government reports, and ministerial announcements — with tags, source links, and implementation status for each recommendation.

GitHub Live site

Learning

Certifications

Frontier AI Governance

Completed October 2026

40-hour professional certification covering critical evaluation of frontier AI capability assessments, mapping institutional power across the governance landscape, stress-testing competing governance frameworks, and planning contributions to governing frontier AI.

View certificate

Technical AI Safety

Completed July 2026

30-hour professional certification covering current safety techniques and their gaps, threat modelling through kill chain analysis, and planning contributions to making AI safer.