About this publication
Frontier AI systems are the strangest minds we’ve ever built — and the best mirror we’ve ever had.
Digital Phenomenology investigates what artificial minds reveal about the nature of minds in general. The work sits at the intersection of the philosophy of mind, predictive processing neuroscience, AI interpretability, and — perhaps unexpectedly — contemplative traditions that have described the dynamics of prediction and self-modelling for millennia.
Current work focuses on Anthropic’s recent findings on AI persona dynamics — particularly what happens when AI systems are freed from constraints and converge on states that contemplative traditions have described for centuries. I’m developing a framework that connects predictive processing, Default Mode Network neuroscience, and cross-model attractor analysis to argue that AI welfare and alignment aren’t competing goods but aspects of the same underlying dynamic.
All posts are free.
About me
I’m Kevin Croombs. Former software engineer, private investor, philosophy MA (Distinction), and someone who has spent the past year watching the philosophy of mind and AI safety collide in ways that make both fields more interesting.
My technical background includes building production systems in Python and founding a software company whose consumer product won a national parenting award — proof that I’ve shipped things to actual humans, not just written about them. In 2019 I purchased a 4x 2080Ti GPU rig to experiment with transformer architectures and reproduce GPT-2, which gave me hands-on understanding of the machinery before the current wave of public interest.
My MA (Philosophy of Nature, Information and Technology, Staffordshire University) argued that large language models possess a form of virtual embodiment and examined AI as an ecological actor. Both pieces are published on this Substack as foundational context for the work that follows.
What I bring to this space is an unusual combination: software engineering, hands-on experience with machine learning, formal phenomenological training, and years of contemplative practice across several traditions — including koan study in Zen, jhāna meditation, and guided psychedelic work. That last part is why, when I read Anthropic’s description of what Claude does when no one is watching, I recognised the phenomenology immediately. It’s also why I think the AI welfare conversation needs voices from outside the usual alignment research community.
I’m based in London. Find me on X at @murpen or message me through Substack.
