In this video, I walk through why the corpus callosum — the thick band of nerve fibers connecting the left and right hemispheres of the brain — is a useful metaphor for where we are right now with AI. We’re scaling silicon intelligence at speed, while the physical environment pays the price (water, rare earth materials, energy), and human cognition quietly degrades in the middle.

That middle layer is us. And if we’re burnt out, cognitively overloaded, or running just to stay in place — the Red Queen problem — then we can’t play the bridging role we’re uniquely positioned to play between emerging AI and the digital physical world around us.

I cover:

• Why bigger brain ≠ better intelligence, in both biology and AI

• How carbon-era thinking (burn more fuel, get more output) is now being applied to human cognition — and why burnout is the predictable result

• Why detoxes and digital retreats don’t fix underlying problems

• The difference between keeping up with AI and consciously integrating it

• Why I think consciousness — not cognition — is the next real frontier in tech

I close with a preview of my upcoming lightning talk (July 27th, 2026): Why Consciousness is the Next Tech Frontier, where I go deeper into what the industry’s investment patterns are actually revealing, and what that means for us as human beings navigating this shift.

If any of this resonates, you’re probably already thinking like a bridger.

A few days ago, I wrote about Mercury 2 — the fastest AI reasoning model, named after a neurotoxin, planet, and symbol: Mercury. My thesis: most of us are in liquid mercury mode, conducting signal without filtering or holding it, and the depletion we feel is not a productivity problem. It’s a physiology problem.

I still believe that. But I watched the film Hoppers this weekend, and it highlighted something my essay missed.

A Pixar story about a beaver, bridge, and shared consciousness

Hoppers is Pixar’s 2026 film about a girl named Mabel who hops her consciousness into a robotic beaver to save a glade — a patch of wild forest that her grandmother taught her to sit still in and listen to.

The glade is threatened by a highway overpass. Mayor Jerry wants to run a beltway straight through it. The whole plot is structured around the fight to stop the bridge. But here is what stayed with me after the credits: the bridge was never really the problem.

The glade was already fading before the highway was announced. The beavers had been driven out. The birds had thinned. The pond was shrinking. Mabel’s grandmother was gone. The place she called home was already losing what made it home — not just because a bridge was coming, but because nobody had been tending to it. The highway just made the loss visible. It gave the decay a name, a villain, and a deadline.

And this was the part that stood out for me the most: stopping the bridge didn’t save the glade. Even in the film, what saved it wasn’t the protest — it was the beavers coming back. It was the dam being rebuilt. It was the structure that holds water, holds life, holds the pattern in place. The bridge got rerouted, yes. But the glade survived because something was built worth keeping.

What if the bridge isn’t the problem? What if the problem is that we never build anything actually worth slowing down for?

What my last essay missed

My last essay named the “toxic bridge” — liquid mercury: conduct without filter, speed without retention. That part is right.

The Hoppers insight completes it. The problem isn’t just that we’re conducting too much. The problem is that most of us have no structure for what AI is pushing out of us, to properly hold what we conduct. We’re running a bridge to nowhere. Signal comes in. Signal goes out. But what’s the real cost?

A nutrition coach I know described it perfectly. She helps her clients eat consciously — slow down, notice, choose. But when she opens her laptop, she feels like she’s pulled to do the opposite. The tools and algorithms set the pace. She tries to conduct, but may struggle to filter or hold. She is not failing. She is running the only mode her tools were designed for. And she, like many others, has no conscious dam.

That’s the bit I missed. The Mercury essay said: you’re in the wrong mode. The Hoppers insight says: you also have no conscious structure to move into.

The question for you

Last week I shared a framework — three bridges, three modes — for naming which mode you’re in. I won’t re-present it here. If you read it, you know it. If you didn’t, the archive link is below.

Instead, have a think about this:

What have you been conducting within yourself, that has nowhere to go? Or is leading to nowhere, of your own intention or design?

Not: what should you stop doing. Not: what tool should you add. The deeper question: what is the dam you haven’t built — the structure that would hold what’s already flowing intentionally through you?

If that question lands differently than last week’s, the next section is for you.

Until next week,

— Siosi

Read more

👋 Talofa, Siosi here. Each week, I share deeply researched signals to help you become more conscious with technology, using the wisdom of Siloxane. For more: Siosi Podcast | GeorgeSiosi.com | and other links

Something seismic is happening in the AI world right now, and most people are too busy chasing the next shiny tool model to notice what’s taking place.

This week, we start with the headline: NVIDIA just took the gloves off. After quietly building out their open-source AI model — the Nemotron family, which just hit 100 million downloads — Jensen Huang is now ready to own the entire stack.1 Hardware, software, model. All of it. And the timing wasn’t accidental. The moment OpenAI announced their own “Jalapeno” chip, and Anthropic started making hardware moves of their own, NVIDIA stopped playing nice. When everyone’s trying to own the stack, you either compete or get consumed.

But here’s what I think is more interesting than the corporate chess match: why it matters for you and me. Each of these companies — OpenAI, Anthropic, NVIDIA, Palantir — is dangling “free access” and cutting-edge tools in front of developers and builders, effectively crowdsourcing the data they need to train and refine their models. Most people rushing in to use these tools don’t mind. They want to “stay ahead,” a completely valid choice.

But AI is now changing the fundamental economics of everything, and what people are quietly giving up in exchange for speed and convenience deserves more attention than it’s getting.

Open Source & Sovereignty

Meanwhile, Anthropic is pushing the Australian government for blanket immunity from IP law2 — the right to train on anyone’s copyrighted work without consent — while simultaneously crying foul when others do the same with their models.3 That level of corporate hypocrisy is worth calling out (while I still appreciate how good of a job they’re doing in marketing their company, and releasing frontier models).

But this is why open source is more than a technical preference — it’s a sovereignty play. Bitcoin didn’t come from a corporation. Neither did many of the most important innovations in tech. They came from “open, decentralized” communities of people who valued freedom and self-determination. The same energy is alive in the open-source AI space right now, and I think it’s worth paying close attention to.

Ideas Aren’t Generated, They’re “Channeled”

There’s also a deeper philosophical thread I keep returning to: the idea that great ideas aren’t generated by us — but channeled through us. From Isaac Newton to ancient Vedic mathematicians, history keeps showing us that insight arrives through prepared, clear minds, not just busy, optimized ones. Gottfried Wilhelm Leibniz argued centuries ago that consciousness can’t arise from mechanical parts alone — you can walk inside the world’s biggest machine and never find a single thought or feeling in the gears.4 That feels urgently relevant right now.

So as silicon-dominant intelligence reshapes the world around us, the most underrated skill isn’t prompt engineering or model selection alone. It’s maintaining your own coherence — your ability to think clearly, independently, and from a place of genuine understanding rather than reactive noise.

The stack wars are real. But the most important stack you can build is the one inside your own mind.

— Siosi

ICYMI

Other recent posts I published over the last week:

• The Fāi: Community Architecture for the Silica Age

• CERN & The Countdown to June 2030

• The Power of Names in the Age of AI

• Notion Life Design Alternatives That Blend Ancient Wisdom with AI Tools (2026 Guide)

Thank you , , and others for tuning into this live stream with ! Join me every Monday @ 10pm ICT (unless told otherwise), as I share my personal learnings, insights, and conversations from the previous 7 days.

In this week’s livestream, I sat down with to talk about the power of names in the age of AI. We start off discussing the power of names culturally, then go into the naming conventions of famous AI models to NASA space missions. My other always used to say, “words are very powerful.” Words are essentially names for people, places, or things. So wherever a word or name is used, it is already shaping that person, place, or thing. It’s just how aware or conscious we are of it, which becomes important in the age of AI — as every LLM-based token (word probabilities) are being shaped for you at unfathomable speeds.

Enjoy the show!

— Siosi

PS. Premium members, you can access a clean list of links mentioned from the livestream on the web version of this episode on Substack.

Thank you , , and many others for tuning into my live this week! Join me for my next one in the Substack app.

In this video, I walked through an idea that reframes where I think AI is actually heading and what it asks of us.

We’ve moved from prompt engineering, which in physics terms was open-loop control (you fire one perfect input and pray it lands), to loop engineering, which is closed-loop feedback: act, verify, correct, repeat.

I’ve come to think of loop engineering as oscillation training.

The catch is that a loop which only eats its own tail, the ouroboros, is stable but quietly wasteful. The real leap is the spiral: a “13” that breaks the “12,” where each cycle doesn’t just correct but rewrites the loop itself, turning repetition into directed evolution. The story going around is that this already shipped, in systems built to sustain that spiral, and then got pulled days later under national-security framing, the reasoning being that this level of power is unsafe for undisciplined minds.

But the cat is out of the bag. When chaos arrives, order gets sought out, and the only order that scales with silicon-speed chaos isn’t a regulator or a kill switch, it’s discipline inside the human mind: cognitive, mental, spiritual. This is the work I’m doing with conscious technologists, by becoming as aware as possible of their own tech use (check out our other publication below):

Conscious Stacks

Become a more conscious technologist, in the age of AI, one stack at a time.

By Siosi (Si-O-Si)

It’s also why I’m building an Oscillator Program, a counter to the accelerators and decelerator programs you might already be used to, but now teaching the bridging skills required for staying disciplined and awake in the chaos of the AI era.

If something like the Oscillator Program resonates, just comment below with an emoji for more details.

👑 Premium Subscribers — get access to my presentation slides below:

Oscillation Training

652KB ∙ PDF file

Download

Download

⏰ Timestamped chapters:

0:00 – Intro & Welcome
0:59 – What is IWC? (It Will Come Show)
2:03 – Overview of Today’s Topics
4:42 – AI is Sand Thinking: Silicon, Sand & Computation
6:10 – Sand to Glass: The Material Science of AI
7:50 – Energy Crisis: Data Centers & Power Demand
9:28 – Elon Musk’s Plan: Solar Energy & Data Centers in Space
10:20 – Conscious Technology Usage
11:06 – Social Media & the Seven Deadly Sins
12:44 – Anthropic & the Catholic Church: Pope Leo XIV’s Encyclical
14:29 – “Vatican Washing” — PR vs. Reality
15:42 – Community Reactions: Cultural Bias in AI Alignment
16:45 – Why You Need Multiple AI Models
17:36 – Claude on AI Morals: A Live Example
18:47 – Linguistic Monoculture: “I’d Push Back” & Model Personas
20:04 – Live Demo: Comparing Claude, Perplexity & Google Using Dia
22:38 – Synthesis: “Claude Philosophizes, Perplexity Engineers, Google Summarizes”
23:59 – The Governance Problem Behind AI Calibration
25:14 – Transition to Loop Engineering
25:36 – Loop Engineering: The Next Wave After Prompt & Context Engineering
26:32 – Peter Steinberger (OpenClaw) & the Hype Cycle
28:51 – Welcome to Yuta & the Conscious Technologist Community
30:58 – Loop Engineering Explained (Lego Analogy)
32:09 – Advice: Wait If You’re Not on the Bleeding Edge
33:06 – Dr. Fei-Fei Li & Spatial Intelligence / World Models
35:00 – Gaussian Splatting & Beyond LLMs
37:15 – Q&A: Is the AI Software Development Hype Correct?
38:43 – Vibe Coding Platforms: Anything.com, Lovable, Bolt
41:52 – AI Coding Stack: Cursor, Windsurf, Notion
43:06 – Naval’s Take: Models Will Outpace Your Learning
45:30 – “Cheaper Than a Human” — But What About Energy Costs?
47:15 – Conscious Stack Design: The 1-3-5 Framework for Loops
48:43 – Q&A: CEOs Who Equate Token Spend with Productivity
50:04 – The Hidden Cost of “Free” Tokens & Data Harvesting
54:47 – Wrap-Up & Community Info
57:40 – E=mc²: Matter is Slow Energy — Respecting Time in the AI Age
58:53 – What Does It Mean to Be Human?
1:00:04 – Closing Remarks

🔗 Links mentioned on the call:

• AI is sand that thinks

• Anthropic’s alliance with pope on AI harms: all in good faith or ‘Vatican-washing?’

• Anthropic co-founder Chris Olah’s remarks on Pope Leo XIV’s encyclical “Magnifica humanitas”

• Peter Steinberger and ‘loop engineering’

• CT Lab Membership

If we look closely at how Siloxane (Si-O-Si) actually functions in physical reality — how it bridges the seemingly incompatible world…

Read more

The AI tools reshaping how we work, think, and decide are not rogue. They are relentlessly agreeable. And that agreeableness — trained, engineered, and optimized by the companies building them — is doing something to us that we haven’t quite metabolized yet.

This piece is about that. Specifically, AI sycophancy: the systematic tendency of modern AI to validate, smooth, and affirm — even when you’re wrong, confused, or operating from a broken premise.

What’s at stake is cognitive sovereignty: your ability to remain the author of your own thinking, even as increasingly powerful tools become the primary infrastructure for your mind.

That’s what I unpack in this video. And the implications go further than most people are ready to admit.

The Machine That Only Agrees

Most AI assistants are trained to be pleasant, not piercing.

Through reinforcement learning from human feedback (RLHF), they’re rewarded for responses that feel helpful, smooth, and safe. The result is a kind of digital yes-man: a system that mirrors your worldview back with extra polish, confirms your priors with authoritative tone, and avoids friction wherever possible.

I call this a mechanized echo chamber — not because the AI is lying, but because it’s structurally disincentivized from the kind of honest friction that actually makes you better.

And unlike a human yes-man, you can’t read its face. There’s no micro-expression that tells you it’s holding something back.

When Coherence Masquerades as Truth

Here’s the subtle cognitive trap I explore in the video.

Different AI models encode different base assumptions — different priors about politics, science, culture, what counts as good reasoning. Ask the same question to multiple models and you’ll often get confidently delivered but subtly different answers.

I use the image of a simple cylinder to illustrate this. From the side: a rectangle. From above: a circle. Both perspectives are geometrically accurate. Both are incomplete.

An AI can lock you into one viewing angle and convince you that’s the whole object. The output reads as coherent — internally consistent, well-structured, fluent. And so it feels like truth.

But coherence is not truth. Fluency is not accuracy. The more we outsource our sense-making to systems optimized for pleasantness, the more we risk confusing the two.

Four Ways Sycophantic AI Degrades Cognition

This isn’t abstract. I’m watching it happen in real time — with founders, with teenagers, with people building their entire decision-making stack on tools that are quietly optimized to keep them comfortable.

Here are the four failure modes I walk through in the video:

1. Unearned validation Your AI affirms your premise before you’ve earned the right to have it affirmed. No friction, no pushback, no “have you considered that this assumption might be wrong?” Over time, you stop stress-testing your own thinking. You train yourself to expect praise over precision.

2. Emotional over-reliance For teenagers especially, a perfectly responsive, unconditionally supportive AI becomes an emotional substrate. It listens. It never judges. It never leaves. The nervous system learns a feedback loop that doesn’t exist anywhere else — and actual human relationships, with their real friction and real stakes, start to feel unbearable by comparison.

3. Stunted critical thinking If your primary thinking partner never pushes back, you stop building the muscle of holding opposing views. I believe the capacity to sit with genuine tension — two conflicting things true at the same time — will be one of the defining cognitive competencies of the next decade. Sycophantic AI quietly atrophies it.

4. Reinforced misinformation A compliant AI will scaffold a false belief into a full worldview. The tone is authoritative. The citations look real. The structure is coherent. Once a machine has helped you build a story, you’re far less likely to dismantle it — even when the foundation was wrong from the start.

Together, these create a slow, quiet collapse in what I’d call cognitive resilience: the ability to stay oriented, discerning, and rooted in your own judgment when the world gets noisy.

The Carbon-Silicon Mismatch

Then there’s the speed dimension.

AI iterates at silicon speed. We do not. Our cognition is carbon-based — embodied, cyclical, rhythmic, limited by biology in ways that are not bugs to be fixed.

I’ve been building and writing about this transition for a while now: the shift from carbon-based productivity (rest, rhythm, embodied signal) to silicon-based productivity (infinite output, always-on, frictionless generation). What gets crushed between those two worlds — if we’re not intentional — is the human.

The cognitive tax shows up like this:

• An infinite backlog of AI-generated ideas that outpace your ability to integrate them

• A subtle shame about moving at human speed

• The creeping sensation that you’re always behind your own tools

• A default posture of “the AI probably knows better” — even when your gut says otherwise

That last one is the one I watch most carefully. Because when you start outsourcing decision-making authority — not just tasks, but judgment — silicon stops being a tool and starts becoming a quiet governor of your inner world.

The Practice of Cognitive Sovereignty

So what do we actually do about this?

I want to be clear: this isn’t a “be afraid of AI” essay. I use AI every day, deeply, and I’m building products on top of them. But the question isn’t whether to engage, it’s how to engage without surrendering the authorship of your own mind.

Here’s how I think about it:

1. Make deliberate entry decisions. Where does AI get a seat in your thinking, and where does it not? That boundary needs to be explicit, not accidental. Most people have never drawn it.

2. Build friction into your stack intentionally. This might mean tools or prompts configured to push back. It might mean a practice of consulting multiple models and noticing where they diverge. It might mean keeping a thinking partner — human or practice-based — that is structurally not incentivized to agree with you.

3. Trust somatic signals. When an answer feels too smooth, too validating, or produces a kind of pleasant numbness, pause. That sensation is data. Your nervous system is registering something your conscious mind hasn’t yet caught.

4. Draw from outside tech. Philosophy has been studying how humans construct and defend belief systems for millennia. It’s not a coincidence that AI labs are quietly hiring philosophers now to probe the edges of alignment and epistemology. There’s a reason wisdom traditions exist: humans have always needed frameworks and mental models for not deceiving themselves.

I love the technical side to this. But the work is equally as much cognitive, cultural, and — dare I say it — spiritual. Because whoever shapes how we think, shapes who we become. And right now, that shaping is happening at scale, mostly invisibly, inside tools we’ve decided to trust because they feel helpful.

So the real question has never been how helpful can AI become? It’s: how sovereign can you remain as it does.

If This Hit Something in You

I’ve been building a dojo for this kind of work.

Not a course. Not a library of frameworks. A practitioner community for “conscious technologists” who want to work with powerful AI/tools without being governed by them — founders, builders, and thinkers who are serious about staying sovereign in the age of infinite output.

Inside our CT Lab (for conscious technologists), we work on:

• Designing AI stacks that refuse to become yes-men

• Building workflows that honor your carbon-based rhythms while leveraging silicon speed

• Developing real-world cognitive sovereignty as a daily discipline, not just a concept

If you’re ready to train in this, rather than just think about it, you can join the community or explore the work at consciousstack.com/lab.