You are currently renting your own intelligence. And Big Tech is making you pay for the privilege.

Every time you use a centralized AI, your data is vacuumed up by a few giants, accelerating a dangerous consolidation of power.

But there is an even bigger physical threat: energy.

If Artificial General Intelligence (AGI) requires centralizing computing in massive server farms, it will concentrate massive energy demand in just a few locations. As AI pioneer Dr. Ben Goertzel warns, this creates a nasty synergy that will turn the AI race into an energy conflict.

Decentralized, open infrastructure is the key to safely distributing both our intelligence and our power consumption.

Many people don't realize there is a realistic solution to balance the scales: truly open, decentralized AI. In my latest Cafe Zai interview with Reza Rassool, Founder and Chair of Kwaai, he noted that while companies like OpenAI call themselves open, they should probably be called "ClosedAI" since they rarely release anything truly open anymore.

To fight this, Kwaai's 1,000-person volunteer "Rebel Alliance" is building a decentralized, open-source AI infrastructure so we can finally be the owners, not the renters, of our own intelligence.

For those of us in the built environment, this is how we protect our physical infrastructure from being held hostage by "AI landlords."

A lack of documentation almost cost Peter Lasensky, founder of Command Post, $20 million. His response? He invented a simple, contemporaneous documentation tool using a Nextel walkie-talkie in 2001. Hear his 30-year journey from being one of San Diego's largest builders to creating Command Post Rev Ops and the Daily Report App—tools that manage the business of construction and have helped customers resolve claims up to $20 million. Discover how his "minimum viable product" evolved from human transcribers to full automation.

Then why can’t you keep ice cream at the right temperature?

Everyone imagines ice cream melting , but the real operational killer is often the opposite: the freezer goes too cold, product gets freezer burn, and inventory is lost.

This episode is a field note as I prep to fly to AHR in Vegas — reflecting on a week where we proved something important with Facil.ai + Onuma:

Smart buildings are not defined by dashboards. They’re defined by outcomes.

If you have sensors and “smarts,” but your ice cream is still outside the right temperature range, you don’t have a smart building, you have a disconnected system. And this is not just about ice cream.

We connected:

• live IoT telemetry (Facil.ai)

• to CloudBIM context (Onuma)

• to operational action (work orders)

And we did it fast , not because it was “easy,” but because the foundation has been built for years, and AI is now accelerating the last mile: synthesis, storytelling, and iteration.

If you have smarts and your ice cream isn’t right… you have a problem.
This episode is about fixing that problem , with connection, context, and meaning. Join us!

Cross-Pollinating AutomatedBuildings: Mentorship & Bumblebees

After more than five decades in the building automation industry, Ken Sinclair has seen it all, from brass pneumatic thermostats and refrigerator-sized computers to cloud platforms, digital twins, and AI. As the founder of AutomatedBuildings.com and a fixture at AHR for nearly 30 years, Ken’s real superpower isn’t prediction; it’s connection.

In this Café ZAI conversation, Ken shares his “bumblebee” philosophy: cross-pollinating ideas between people, disciplines, and generations. We explore how mentorship really works, why innovation dies in silos, what’s changed at AHR over the decades, and why trust, not technology, remains the industry’s most valuable currency.

A conversation about history, humility, and why the future of automated buildings depends on sharing the honey.

Data Sharing, IP, and the Future of Buildings

John Turner brings decades of hard-earned experience across major global organizations, working from the owner side, leading digital transformation, and pushing real-world BIM and digital twin strategies at scale. In this Café Zai session, John Turner and I talk about what’s holding the industry back: siloed thinking, IP fear, and “hill protection.” We explore the full building lifecycle, why AI will amplify the cost of locked-up data, and how owners can shift from fragmented systems to connected, value-producing intelligence. 

Because the real world hasn’t fully come to the table yet.

At NeurIPS Urban AI, I spoke with Fabian Raisch (TU Rosenheim/TU Munich), whose research targets scalable AI-based building control with significant potential: improving efficiency by 10–50%, depending on building conditions and systems.

Fabian shared the most important transition happening right now: his team has been working in the ideal world of simulated data, and now they’re stepping toward real deployment, working directly with heat pump operators and practical constraints.

That’s the pivot point for the entire industry. The AI is moving fast. The research is impressive. The agents are ready.

What’s missing is living, accessible, trustworthy building data, because most buildings are still “quiet. Not every building AI approach needs huge training datasets; some (like what Facil describes) are closer to real-time control and continuous learning through intervention. But either way, we still need buildings to become accessible, observable, and safely controllable.

In this Café ZAi conversation, Kimon Onuma speaks with Zheng (Thomas) Tang from NVIDIA about how synthetic data, computer vision, and digital twins are teaching AI to understand buildings, warehouses, and cities. The discussion explores “outside-in” safety, multi-camera perception, and why seeing environments accurately is foundational to the future of urban AI.

Donna Vakalis, a postdoc researcher at Mila, Quebec AI Institute, focuses on reducing climate emissions from buildings using AI. Her work involves optimizing HVAC systems without extensive data, employing data-efficient reinforcement learning. Vakalis' research aims to create agents that can control building systems effectively from day one, even without prior experience in the specific building. She uses in-context reinforcement learning, a novel method that adapts to different building conditions. Vakalis' mentors, Joshua Bengio and David Ronik, are leaders in the field. The conversation also touched on the broader implications of AI in urban settings and the high demand for AI talent.

A Columbia robotics researcher and a Google AI engineer join Café Zai to ask why our buildings, giant robots in disguise, are still asleep.

About the Guest:

Judah Goldfeder is a researcher at Columbia University’s robotics and dynamic systems lab and an AI researcher at Google, where he works on reinforcement learning for real buildings. He bridges two worlds—cutting-edge AI research and the messy, real physics of HVAC, buildings, and cities—which makes his viewpoint uniquely valuable.

This episode explores why buildings behave like giant robots, why most of them are still “asleep,” and what it will take to wake them up.

Connect with Judah on LinkedIn:

🔗 https://www.linkedin.com/in/judah-goldfeder/

Explore the UrbanAI Workshop at NeurIPS (where I will be presenting at his invitation):

🔗 https://urbanai2025.github.io/

Kimon Onuma and Cyril Verley trade confessions from decades inside the BIM and digital twin world. From hoarding data and chasing flashy visuals to selling LOD 500 models that nobody could use, they name the “Seven Deadly BIMs” and discuss how the industry can find redemption through open data, APIs, and collaboration.

This episode of Café Zai invites listeners to reflect, and rethink how architectural intelligence can finally move from drawings to data.