Notes on
engineering decisions.
The questions behind the work: what a model is actually for, how to evaluate it, how far to trust an agent, and where the computing should happen.
More capability should mean more room to act.
A thread from a 2015 essay on human possibility to today’s experiments with voice, visual context and agents.
Accuracy is only part of the decision.
Why false alarms, missed events and the review workflow belong in the same conversation.
Agent systems need a memory, and a boundary.
Durable handoffs, explicit authority and evidence of completion are part of the engineering.
Where the model runs is a product decision.
Latency, connectivity, data movement and operating constraints shape useful AI.
The questions came
before the current tools.
2024Synthetic conversations, directed by a human.
A published demonstration of expressive voices, editable scripts and control over the flow of AI-generated conversation.
The rise of AI: they will be us, should we worry?
An early essay on AI and human agency.
A human-centred view of technological innovation.
A philosophical exploration of technology as an extension of human capability: the starting point for a continuing line of thought.