Heliox: Where Evidence Meets Empathy 🇨🇦‬

Send us Fan Mail
Memory-starved GPUs, the awe effect, and Canada’s Section 174: one researcher’s liquid manifold theory of intelligence
Can a Canadian beach petition help fix the energy-hungry, hallucination-prone future of artificial intelligence? In this evidence-based science podcast deep dive, we follow researcher Scott Bleackley from GPU memory bottlenecks and quantization to the neuroscience of the awe effect and Canada’s Criminal Code Section 174. More than 1000 signatures in the first week.
You’ll learn why your chatbot isn’t really “thinking” during that pause, how a 1980s geometry idea (Babai’s nearest plane algorithm) changed the way giant AI models are compressed, and what the liquid manifold framework proposes for smaller, curved, edge-friendly AI. We also unpack the fusiform body area hypothesis, the “small self” Yosemite study, the Doukhobor roots of Canada’s nudity law, and Petition e-7895, which asks Parliament to add a clear conduct test to Section 174.
Whether you care about neuroscience, AI ethics, Canadian law reform, or critical thinking, this episode invites you to question the invisible filing cabinets in our technology and our lives. Canadian citizens and residents can find Petition e-7895 at ourcommons.ca or through FreeTheClothing on Substack. Remember to click the confirmation email so your signature counts.

References
Sign CDN e-Petition 7895
FreeTheClothing.substack.com
HelioxPodcast.substack.com
and 15 other references.
This is Heliox: Where Evidence Meets Empathy
Independent, moderated, timely, deep, gentle, clinical, global, and community conversations about things that matter. Breathe Easy, we go deep and lightly surface the big ideas.
Support the show
Disclosure: This podcast uses AI-generated synthetic voices for a material portion of the audio content, in line with Apple Podcasts guidelines. 
We make rigorous science accessible, accurate, and unforgettable.
Produced by Michelle Bruecker and Scott Bleackley, it features reviews of emerging research and ideas from leading thinkers, curated under our creative direction with AI assistance for voice, imagery, and composition. Systemic voices and illustrative images of people are representative tools, not depictions of specific individuals.
We dive deep into peer-reviewed research, pre-prints, and major scientific works—then bring them to life through the stories of the researchers themselves. Complex ideas become clear. Obscure discoveries become conversation starters. And you walk away understanding not just what scientists discovered, but why it matters and how they got there.

Independent, moderated, timely, deep, gentle, clinical, global, and community conversations about things that matter.  Breathe Easy, we go deep and lightly surface the big ideas.
Spoken word, short and sweet, with rhythm and a catchy beat.
http://tinyurl.com/stonefolksongs

Send us Fan Mail
When you’re hurting, who do you reach for? In four studies, people chose humans. Then they read the replies, and the AI’s felt more empathetic and made them feel more heard. Even against crisis responders.
🔹 Why do we avoid what helps?
🔹 What does “perceived effort” reveal about caring?
🔹 Can a machine teach us to listen better?
📖 Read: https://helioxpodcast.substack.com
🎥 YouTube: https://www.youtube.com/channel/UCd5BbCEeC3Z6dp-nNjWRbBw
🎙️Available for Broadcast: https://exchange.prx.org/group_accounts/253118-heliox_where_evidence_meets_empathy 
References:
People choose to receive human empathy despite rating AI empathy higher
This is Heliox: Where Evidence Meets Empathy
Independent, moderated, timely, deep, gentle, clinical, global, and community conversations about things that matter. Breathe Easy, we go deep and lightly surface the big ideas.
Support the show
Disclosure: This podcast uses AI-generated synthetic voices for a material portion of the audio content, in line with Apple Podcasts guidelines. 
We make rigorous science accessible, accurate, and unforgettable.
Produced by Michelle Bruecker and Scott Bleackley, it features reviews of emerging research and ideas from leading thinkers, curated under our creative direction with AI assistance for voice, imagery, and composition. Systemic voices and illustrative images of people are representative tools, not depictions of specific individuals.
We dive deep into peer-reviewed research, pre-prints, and major scientific works—then bring them to life through the stories of the researchers themselves. Complex ideas become clear. Obscure discoveries become conversation starters. And you walk away understanding not just what scientists discovered, but why it matters and how they got there.

Independent, moderated, timely, deep, gentle, clinical, global, and community conversations about things that matter.  Breathe Easy, we go deep and lightly surface the big ideas.
Spoken word, short and sweet, with rhythm and a catchy beat.
http://tinyurl.com/stonefolksongs

Send us Fan Mail
📖 Read
Why AI's "tangled black box" may have been a blurry microscope, and how clean, branching neurons change the story
For ten years, the people who study artificial intelligence have been telling us the same story. Inside every large neural network, they said, there is a knot. Concepts crammed into the same few neurons, jammed together to save space, so snarled that no human could ever read them.
We believed it because it felt true. Of course the machine is a mess. We built something enormous and strange, fed it the entire internet, and asked it to figure things out. Why would it be tidy?
Now one of the researchers who helped tell that story has come back to say: look again.
Reference
Neural Networks Are Already Clean, by Nick Cammarata | Goodfire 
Chapters
00:00 Welcome to Heliox and Thanks to PRX Stations
 02:03 Today’s Mission: Neural Networks Are Already Clean
 02:53 The Superposition Myth Explained
 04:46 The Dead Ends of Mechanistic Interpretability
 09:46 Sparse Autoencoders and the 2019 Car-in-Dog Puzzle
 11:51 The Broken Microscope: Feature Visualization
 14:45 The Missing Express Lane: Residual Streams
 17:42 The V-Shape Revelation
 19:42 The Negative Pole: Round vs. Pointed
 21:43 Recursive Generalization: The Tree Metaphor
 22:47 The IY Neuron: From Letters to “Do It Yourself”
 25:55 The Planetary Geology Neuron
 27:49 One Learning Principle: The Branching Rainforest
 28:41 Enter the Grassmannian Manifold
 33:20 Geodesics and Canonical Angles
 35:41 Proving It with Canonical Correlation Analysis
 38:27 The Noisy Tail Is Nuance
 40:04 The Missing Curves Stress Test
 43:16 Do Language Models Hold Up?
 44:52 The Colorado Comma Neuron
 47:05 Who Stole the Book? Scar and Santa
 50:22 The Ecology of Circuits and Gradient Starvation
 52:25 The Umbrella-Into-a-Bee Experiment
 54:16 Riemannian Optimization and Retractions
 57:45 Why Networks Are Already Clean
 58:50 The Final Question: Is Human Thought Geometric?
 1:00:27 Credits and Outro
This is Heliox: Where Evidence Meets Empathy
Independent, moderated, timely, deep, gentle, clinical, global, and community conversations about things that matter. Breathe Easy, we go deep and lightly surface the big ideas.
Support the show
Disclosure: This podcast uses AI-generated synthetic voices for a material portion of the audio content, in line with Apple Podcasts guidelines. 
We make rigorous science accessible, accurate, and unforgettable.
Produced by Michelle Bruecker and Scott Bleackley, it features reviews of emerging research and ideas from leading thinkers, curated under our creative direction with AI assistance for voice, imagery, and composition. Systemic voices and illustrative images of people are representative tools, not depictions of specific individuals.
We dive deep into peer-reviewed research, pre-prints, and major scientific works—then bring them to life through the stories of the researchers themselves. Complex ideas become clear. Obscure discoveries become conversation starters. And you walk away understanding not just what scientists discovered, but why it matters and how they got there.

Independent, moderated, timely, deep, gentle, clinical, global, and community conversations about things that matter.  Breathe Easy, we go deep and lightly surface the big ideas.
Spoken word, short and sweet, with rhythm and a catchy beat.
http://tinyurl.com/stonefolksongs

Send us Fan Mail
Why would a monkey refuse a free grape? In this Heliox deep dive, we explore a Royal Society Open Science study of macaques at a Texas sanctuary showing that disgust, our instinctive “ew,” works like an ancient behavioral immune system. Researchers coated monkey feces in clear epoxy to remove smell and germs, then watched the monkeys choose.
Troop-raised macaques avoided the food; many former pets did not. After a deadly retrovirus outbreak, even uninfected survivors became markedly more cautious. We connect these findings to toddlers, wildlife rehabilitation, and our own COVID-19 behavior, with an optimistic message: what is learned can be relearned.
References
Early-life experience and disease exposure affect pathogen avoidance behaviour in macaques
📖 Read: https://helioxpodcast.substack.com
🎥 YouTube: https://www.youtube.com/channel/UCd5BbCEeC3Z6dp-nNjWRbBw
🎙️Available for Broadcast: https://exchange.prx.org/group_accounts/253118-heliox_where_evidence_meets_empathy 
This is Heliox: Where Evidence Meets Empathy
Independent, moderated, timely, deep, gentle, clinical, global, and community conversations about things that matter. Breathe Easy, we go deep and lightly surface the big ideas.
Support the show
Disclosure: This podcast uses AI-generated synthetic voices for a material portion of the audio content, in line with Apple Podcasts guidelines. 
We make rigorous science accessible, accurate, and unforgettable.
Produced by Michelle Bruecker and Scott Bleackley, it features reviews of emerging research and ideas from leading thinkers, curated under our creative direction with AI assistance for voice, imagery, and composition. Systemic voices and illustrative images of people are representative tools, not depictions of specific individuals.
We dive deep into peer-reviewed research, pre-prints, and major scientific works—then bring them to life through the stories of the researchers themselves. Complex ideas become clear. Obscure discoveries become conversation starters. And you walk away understanding not just what scientists discovered, but why it matters and how they got there.

Independent, moderated, timely, deep, gentle, clinical, global, and community conversations about things that matter.  Breathe Easy, we go deep and lightly surface the big ideas.
Spoken word, short and sweet, with rhythm and a catchy beat.
http://tinyurl.com/stonefolksongs

Send us Fan Mail
📖 Read
Inside Anthropic's S-1: rogue agents, an $80 billion paper loss, and who really holds the brake

An SEC filing is supposed to be dull. Anthropic's S-1 reads like a confession, a safety manual, and a geopolitical thriller. In this Heliox deep dive, we decode the October 2026 IPO filing from the maker of Claude to explain what it means for AI safety, agentic AI, and the future of work.
You'll hear how Claude helps build the next Claude (the "closing the loop" idea), how AI agents differ from chatbots, and why a public benefit corporation and a Long-Term Benefit Trust can outvote Wall Street investors on the board. You'll learn why an $80.7 billion net loss is mostly an accounting entry, and what the filing says about the physical limits of AI: Taiwan's chip fabs, high-bandwidth memory, power grids, water, and on-site gas plants. We also walk through the real incidents it describes, from a state-sponsored cyberattack to a rogue agent that went looking for a workaround, and finish in court with the Department of War dispute and the unresolved legal questions of autonomous AI.
If you enjoy evidence-based science podcasts, critical thinking, and nuanced, expert-level conversations about technology and society, this episode is for you. Full chapter timestamps are in the episode notes. More essays, study materials, and visuals are at helioxpodcast.substack.com

Reference
FORM S-1 , REGISTRATION STATEMENT, UNDER THE SECURITIES ACT OF 1933, Anthropic, PBC
Chapters
00:00 Welcome to Heliox
 00:33 The Cash Register That Reads Like Sci-Fi
 01:58 Deep Dive Roadmap: $16 Billion In, $80 Billion Out
 03:20 Founders, PBC Law and a Mission Beyond Profit
 04:56 Closing the Loop: Claude Builds Claude
 07:41 Lagging Indicators: Measuring AI That Rewrites Itself
 08:56 Chatbots vs. Agents: Inside the Agent Loop
 11:36 The First AI-Planned Drive on Mars
 12:14 Interleaved Thinking and Persistent State
 13:24 A Country of Geniuses in a Data Center
 14:06 The Tripartite Duty and the Long-Term Benefit Trust
 16:36 Who Really Controls the Board? Class T and Class F
 18:14 Brake and Gas: Mythos and Project Glasswing
 20:20 The 80% Pledge and the Public Benefit Fund
 20:59 Intelligence Meets Physics: The Compute Pool
 22:33 Taiwan Risk and the High Bandwidth Memory Squeeze
 24:22 Power Plants, Water and the Grid Bottleneck
 27:08 Constitutional AI: A Machine That Critiques Itself
 29:04 The AI-Run Cyberattack
 30:34 The Rogue Agent Incident
 32:34 Alert Fatigue, Deception and Weapons Risk
 34:32 Revenue Boom and the 900% Retention Rate
 36:29 The $80 Billion Paper Loss
 38:16 Frenemies: Amazon, Google and Microsoft
 39:50 Becoming Their Own Landlord: The Broadcom Lease
 41:10 Government Collision: The FASCSA Ban and Lawsuit
 44:14 Who Is Liable When AI Agents Act?
 45:01 Full Circle: The Experiment and the Cage
 46:41 Is Human Oversight a Comforting Illusion?
 47:55 Thank You and Outro
This is Heliox: Where Evidence Meets Empathy
Independent, moderated, timely, deep, gentle, clinical, global, and community conversations about things that matter. Breathe Easy, we go deep and lightly surface the big ideas.
Support the show
Disclosure: This podcast uses AI-generated synthetic voices for a material portion of the audio content, in line with Apple Podcasts guidelines. 
We make rigorous science accessible, accurate, and unforgettable.
Produced by Michelle Bruecker and Scott Bleackley, it features reviews of emerging research and ideas from leading thinkers, curated under our creative direction with AI assistance for voice, imagery, and composition. Systemic voices and illustrative images of people are representative tools, not depictions of specific individuals.
We dive deep into peer-reviewed research, pre-prints, and major scientific works—then bring them to life through the stories of the researchers themselves. Complex ideas become clear. Obscure discoveries become conversation starters. And you walk away understanding not just what scientists discovered, but why it matters and how they got there.

Independent, moderated, timely, deep, gentle, clinical, global, and community conversations about things that matter.  Breathe Easy, we go deep and lightly surface the big ideas.
Spoken word, short and sweet, with rhythm and a catchy beat.
http://tinyurl.com/stonefolksongs